id
int64
0
458k
file_name
stringlengths
4
119
file_path
stringlengths
14
227
content
stringlengths
24
9.96M
size
int64
24
9.96M
language
stringclasses
1 value
extension
stringclasses
14 values
total_lines
int64
1
219k
avg_line_length
float64
2.52
4.63M
max_line_length
int64
5
9.91M
alphanum_fraction
float64
0
1
repo_name
stringlengths
7
101
repo_stars
int64
100
139k
repo_forks
int64
0
26.4k
repo_open_issues
int64
0
2.27k
repo_license
stringclasses
12 values
repo_extraction_date
stringclasses
433 values
27,800
cli.py
kovidgoyal_calibre/src/calibre/devices/usbms/cli.py
__license__ = 'GPL v3' __copyright__ = '2009, John Schember <john@nachtimwald.com>' __docformat__ = 'restructuredtext en' import os import shutil import time from calibre import fsync from calibre.devices.errors import PathError from calibre.utils.filenames import case_preserving_open_file class File: def __init__(self, path): stats = os.stat(path) self.is_dir = os.path.isdir(path) self.is_readonly = not os.access(path, os.W_OK) self.ctime = stats.st_ctime self.wtime = stats.st_mtime self.size = stats.st_size if path.endswith(os.sep): path = path[:-1] self.path = path self.name = os.path.basename(path) def check_transfer(infile, dest): infile.seek(0) dest.seek(0) return infile.read() == dest.read() class CLI: def get_file(self, path, outfile, end_session=True): path = self.munge_path(path) with open(path, 'rb') as src: shutil.copyfileobj(src, outfile) def put_file(self, infile, path, replace_file=False, end_session=True): path = self.munge_path(path) close = False if not hasattr(infile, 'read'): infile, close = open(infile, 'rb'), True infile.seek(0) if os.path.isdir(path): path = os.path.join(path, infile.name) if not replace_file and os.path.exists(path): raise PathError('File already exists: ' + path) dest, actual_path = case_preserving_open_file(path) with dest: try: shutil.copyfileobj(infile, dest) except OSError: print('WARNING: First attempt to send file to device failed') time.sleep(0.2) infile.seek(0) dest.seek(0) dest.truncate() shutil.copyfileobj(infile, dest) fsync(dest) # if not check_transfer(infile, dest): raise Exception('Transfer failed') if close: infile.close() return actual_path def munge_path(self, path): if path.startswith('/') and not (path.startswith(self._main_prefix) or (self._card_a_prefix and path.startswith(self._card_a_prefix)) or (self._card_b_prefix and path.startswith(self._card_b_prefix))): path = self._main_prefix + path[1:] elif path.startswith('carda:'): path = path.replace('carda:', self._card_a_prefix[:-1]) elif path.startswith('cardb:'): path = path.replace('cardb:', self._card_b_prefix[:-1]) return path def list(self, path, recurse=False, end_session=True, munge=True): if munge: path = self.munge_path(path) if os.path.isfile(path): return [(os.path.dirname(path), [File(path)])] entries = [File(os.path.join(path, f)) for f in os.listdir(path)] dirs = [(path, entries)] for _file in entries: if recurse and _file.is_dir: dirs[len(dirs):] = self.list(_file.path, recurse=True, munge=False) return dirs def mkdir(self, path, end_session=True): if self.SUPPORTS_SUB_DIRS: path = self.munge_path(path) os.mkdir(path) def rm(self, path, end_session=True): path = self.munge_path(path) self.delete_books([path]) def touch(self, path, end_session=True): path = self.munge_path(path) if not os.path.exists(path): open(path, 'wb').close() if not os.path.isdir(path): os.utime(path, None)
3,609
Python
.py
90
30.788889
85
0.592349
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,801
books.py
kovidgoyal_calibre/src/calibre/devices/usbms/books.py
__license__ = 'GPL 3' __copyright__ = '2009, John Schember <john@nachtimwald.com>' __docformat__ = 'restructuredtext en' import os import re import sys import time from functools import cmp_to_key from calibre import force_unicode, isbytestring from calibre.constants import preferred_encoding from calibre.devices.interface import BookList as _BookList from calibre.devices.mime import mime_type_ext from calibre.ebooks.metadata import title_sort from calibre.ebooks.metadata.book.base import Metadata from calibre.prints import debug_print from calibre.utils.config_base import tweaks from calibre.utils.icu import sort_key from polyglot.builtins import cmp, iteritems, itervalues, string_or_bytes def none_cmp(xx, yy): x = xx[1] y = yy[1] if x is None and y is None: # No sort_key needed here, because defaults are ascii try: return cmp(xx[2], yy[2]) except TypeError: return 0 if x is None: return 1 if y is None: return -1 if isinstance(x, string_or_bytes) and isinstance(y, string_or_bytes): x, y = sort_key(force_unicode(x)), sort_key(force_unicode(y)) try: c = cmp(x, y) except TypeError: c = 0 if c != 0: return c # same as above -- no sort_key needed here try: return cmp(xx[2], yy[2]) except TypeError: return 0 class Book(Metadata): def __init__(self, prefix, lpath, size=None, other=None): from calibre.ebooks.metadata.meta import path_to_ext Metadata.__init__(self, '') self._new_book = False self.device_collections = [] self.path = os.path.join(prefix, lpath) if os.sep == '\\': self.path = self.path.replace('/', '\\') self.lpath = lpath.replace('\\', '/') else: self.lpath = lpath self.mime = mime_type_ext(path_to_ext(lpath)) self.size = size # will be set later if None try: self.datetime = time.gmtime(os.path.getctime(self.path)) except: self.datetime = time.gmtime() if other: self.smart_update(other) def __eq__(self, other): # use lpath because the prefix can change, changing path return self.lpath == getattr(other, 'lpath', None) @property def db_id(self): '''The database id in the application database that this file corresponds to''' match = re.search(r'_(\d+)$', self.lpath.rpartition('.')[0]) if match: return int(match.group(1)) return None @property def title_sorter(self): '''String to sort the title. If absent, title is returned''' return title_sort(self.title) @property def thumbnail(self): return None class BookList(_BookList): def __init__(self, oncard, prefix, settings): _BookList.__init__(self, oncard, prefix, settings) self._bookmap = {} def supports_collections(self): return False def add_book(self, book, replace_metadata): return self.add_book_extended(book, replace_metadata, check_for_duplicates=True) def add_book_extended(self, book, replace_metadata, check_for_duplicates): ''' Add the book to the booklist, if needed. Return None if the book is already there and not updated, otherwise return the book. ''' try: b = self.index(book) if check_for_duplicates else None except (ValueError, IndexError): b = None if b is None: self.append(book) return book if replace_metadata: self[b].smart_update(book, replace_metadata=True) return self[b] return None def remove_book(self, book): self.remove(book) def get_collections(self): return {} class CollectionsBookList(BookList): def supports_collections(self): return True def in_category_sort_rules(self, attr): sorts = tweaks['sony_collection_sorting_rules'] for attrs,sortattr in sorts: if attr in attrs or '*' in attrs: return sortattr return None def compute_category_name(self, field_key, field_value, field_meta): from calibre.utils.formatter import EvalFormatter renames = tweaks['sony_collection_renaming_rules'] field_name = renames.get(field_key, None) if field_name is None: if field_meta['is_custom']: field_name = field_meta['name'] else: field_name = '' cat_name = EvalFormatter().safe_format( fmt=tweaks['sony_collection_name_template'], kwargs={'category':field_name, 'value':field_value}, error_value='GET_CATEGORY', book=None) return cat_name.strip() def get_collections(self, collection_attributes): from calibre.utils.config import device_prefs debug_print('Starting get_collections:', device_prefs['manage_device_metadata']) debug_print('Renaming rules:', tweaks['sony_collection_renaming_rules']) debug_print('Formatting template:', tweaks['sony_collection_name_template']) debug_print('Sorting rules:', tweaks['sony_collection_sorting_rules']) # Complexity: we can use renaming rules only when using automatic # management. Otherwise we don't always have the metadata to make the # right decisions use_renaming_rules = device_prefs['manage_device_metadata'] == 'on_connect' collections = {} # get the special collection names all_by_author = '' all_by_title = '' ca = [] all_by_something = [] for c in collection_attributes: if c.startswith('aba:') and c[4:].strip(): all_by_author = c[4:].strip() elif c.startswith('abt:') and c[4:].strip(): all_by_title = c[4:].strip() elif c.startswith('abs:') and c[4:].strip(): name = c[4:].strip() sby = self.in_category_sort_rules(name) if sby is None: sby = name if name and sby: all_by_something.append((name, sby)) else: ca.append(c.lower()) collection_attributes = ca for book in self: tsval = book.get('_pb_title_sort', book.get('title_sort', book.get('title', 'zzzz'))) asval = book.get('_pb_author_sort', book.get('author_sort', '')) # Make sure we can identify this book via the lpath lpath = getattr(book, 'lpath', None) if lpath is None: continue # Decide how we will build the collections. The default: leave the # book in all existing collections. Do not add any new ones. attrs = ['device_collections'] if getattr(book, '_new_book', False): if device_prefs['manage_device_metadata'] == 'manual': # Ensure that the book is in all the book's existing # collections plus all metadata collections attrs += collection_attributes else: # For new books, both 'on_send' and 'on_connect' do the same # thing. The book's existing collections are ignored. Put # the book in collections defined by its metadata. attrs = collection_attributes elif device_prefs['manage_device_metadata'] == 'on_connect': # For existing books, modify the collections only if the user # specified 'on_connect' attrs = collection_attributes for attr in attrs: attr = attr.strip() # If attr is device_collections, then we cannot use # format_field, because we don't know the fields where the # values came from. if attr == 'device_collections': doing_dc = True val = book.device_collections # is a list else: doing_dc = False ign, val, orig_val, fm = book.format_field_extended(attr) if not val: continue if isbytestring(val): val = val.decode(preferred_encoding, 'replace') if isinstance(val, (list, tuple)): val = list(val) elif fm['datatype'] == 'series': val = [orig_val] elif fm['datatype'] == 'text' and fm['is_multiple']: val = orig_val elif fm['datatype'] == 'composite' and fm['is_multiple']: val = [v.strip() for v in val.split(fm['is_multiple']['ui_to_list'])] else: val = [val] sort_attr = self.in_category_sort_rules(attr) for category in val: is_series = False if doing_dc: # Attempt to determine if this value is a series by # comparing it to the series name. if category == book.series: is_series = True elif fm['is_custom']: # is a custom field if fm['datatype'] == 'text' and len(category) > 1 and \ category[0] == '[' and category[-1] == ']': continue if fm['datatype'] == 'series': is_series = True else: # is a standard field if attr == 'tags' and len(category) > 1 and \ category[0] == '[' and category[-1] == ']': continue if attr == 'series' or \ ('series' in collection_attributes and book.get('series', None) == category): is_series = True if use_renaming_rules: cat_name = self.compute_category_name(attr, category, fm) else: cat_name = category if cat_name not in collections: collections[cat_name] = {} if use_renaming_rules and sort_attr: sort_val = book.get(sort_attr, None) collections[cat_name][lpath] = (book, sort_val, tsval) elif is_series: if doing_dc: collections[cat_name][lpath] = \ (book, book.get('series_index', sys.maxsize), tsval) else: collections[cat_name][lpath] = \ (book, book.get(attr+'_index', sys.maxsize), tsval) else: if lpath not in collections[cat_name]: collections[cat_name][lpath] = (book, tsval, tsval) # All books by author if all_by_author: if all_by_author not in collections: collections[all_by_author] = {} collections[all_by_author][lpath] = (book, asval, tsval) # All books by title if all_by_title: if all_by_title not in collections: collections[all_by_title] = {} collections[all_by_title][lpath] = (book, tsval, asval) for (n, sb) in all_by_something: if n not in collections: collections[n] = {} collections[n][lpath] = (book, book.get(sb, ''), tsval) # Sort collections result = {} for category, lpaths in iteritems(collections): books = sorted(itervalues(lpaths), key=cmp_to_key(none_cmp)) result[category] = [x[0] for x in books] return result def rebuild_collections(self, booklist, oncard): ''' For each book in the booklist for the card oncard, remove it from all its current collections, then add it to the collections specified in device_collections. oncard is None for the main memory, carda for card A, cardb for card B, etc. booklist is the object created by the :method:`books` call above. ''' pass
12,760
Python
.py
285
31.470175
88
0.540534
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,802
driver.py
kovidgoyal_calibre/src/calibre/devices/usbms/driver.py
__license__ = 'GPL v3' __copyright__ = '2009, John Schember <john at nachtimwald.com>' __docformat__ = 'restructuredtext en' ''' Generic USB Mass storage device driver. This is not a complete stand alone driver. It is intended to be subclassed with the relevant parts implemented for a particular device. ''' import json import os import shutil from itertools import cycle from calibre import fsync, isbytestring, prints from calibre.constants import filesystem_encoding, ismacos, numeric_version from calibre.devices.usbms.books import Book, BookList from calibre.devices.usbms.cli import CLI from calibre.devices.usbms.device import Device from calibre.ebooks.metadata.book.json_codec import JsonCodec from calibre.prints import debug_print from polyglot.builtins import itervalues, string_or_bytes def safe_walk(top, topdown=True, onerror=None, followlinks=False, maxdepth=128): ' A replacement for os.walk that does not die when it encounters undecodeable filenames in a linux filesystem' if maxdepth < 0: return islink, join, isdir = os.path.islink, os.path.join, os.path.isdir # We may not have read permission for top, in which case we can't # get a list of the files the directory contains. os.path.walk # always suppressed the exception then, rather than blow up for a # minor reason when (say) a thousand readable directories are still # left to visit. That logic is copied here. try: names = os.listdir(top) except OSError as err: if onerror is not None: onerror(err) return dirs, nondirs = [], [] for name in names: if isinstance(name, bytes): try: name = name.decode(filesystem_encoding) except UnicodeDecodeError: debug_print('Skipping undecodeable file: %r' % name) continue if isdir(join(top, name)): dirs.append(name) else: nondirs.append(name) if topdown: yield top, dirs, nondirs for name in dirs: new_path = join(top, name) if followlinks or not islink(new_path): yield from safe_walk(new_path, topdown, onerror, followlinks, maxdepth-1) if not topdown: yield top, dirs, nondirs # CLI must come before Device as it implements the CLI functions that # are inherited from the device interface in Device. class USBMS(CLI, Device): ''' The base class for all USBMS devices. Implements the logic for sending/getting/updating metadata/caching metadata/etc. ''' description = _('Communicate with an e-book reader.') author = 'John Schember' supported_platforms = ['windows', 'osx', 'linux'] # Store type instances of BookList and Book. We must do this because # a) we need to override these classes in some device drivers, and # b) the classmethods seem only to see real attributes declared in the # class, not attributes stored in the class booklist_class = BookList book_class = Book FORMATS = [] CAN_SET_METADATA = [] METADATA_CACHE = 'metadata.calibre' DRIVEINFO = 'driveinfo.calibre' SCAN_FROM_ROOT = False def _update_driveinfo_record(self, dinfo, prefix, location_code, name=None): import uuid from calibre.utils.date import isoformat, now if not isinstance(dinfo, dict): dinfo = {} if dinfo.get('device_store_uuid', None) is None: dinfo['device_store_uuid'] = str(uuid.uuid4()) if dinfo.get('device_name', None) is None: dinfo['device_name'] = self.get_gui_name() if name is not None: dinfo['device_name'] = name dinfo['location_code'] = location_code dinfo['last_library_uuid'] = getattr(self, 'current_library_uuid', None) dinfo['calibre_version'] = '.'.join([str(i) for i in numeric_version]) dinfo['date_last_connected'] = isoformat(now()) dinfo['prefix'] = prefix.replace('\\', '/') return dinfo def _update_driveinfo_file(self, prefix, location_code, name=None): from calibre.utils.config import from_json, to_json if os.path.exists(os.path.join(prefix, self.DRIVEINFO)): with open(os.path.join(prefix, self.DRIVEINFO), 'rb') as f: try: driveinfo = json.loads(f.read(), object_hook=from_json) except: driveinfo = None driveinfo = self._update_driveinfo_record(driveinfo, prefix, location_code, name) data = json.dumps(driveinfo, default=to_json) if not isinstance(data, bytes): data = data.encode('utf-8') with open(os.path.join(prefix, self.DRIVEINFO), 'wb') as f: f.write(data) fsync(f) else: driveinfo = self._update_driveinfo_record({}, prefix, location_code, name) data = json.dumps(driveinfo, default=to_json) if not isinstance(data, bytes): data = data.encode('utf-8') with open(os.path.join(prefix, self.DRIVEINFO), 'wb') as f: f.write(data) fsync(f) return driveinfo def get_device_information(self, end_session=True): self.report_progress(1.0, _('Get device information...')) self.driveinfo = {} def raise_os_error(e): raise OSError(_('Failed to access files in the main memory of' ' your device. You should contact the device' ' manufacturer for support. Common fixes are:' ' try a different USB cable/USB port on your computer.' ' If you device has a "Reset to factory defaults" type' ' of setting somewhere, use it. Underlying error: %s') % e) from e if self._main_prefix is not None: try: self.driveinfo['main'] = self._update_driveinfo_file(self._main_prefix, 'main') except PermissionError as e: if ismacos: raise PermissionError(_( 'Permission was denied by macOS trying to access files in the main memory of' ' your device. You will need to grant permission explicitly by looking under' ' System Preferences > Security and Privacy > Privacy > Files and Folders.' ' Underlying error: %s' ) % e) from e raise_os_error(e) except OSError as e: raise_os_error(e) try: if self._card_a_prefix is not None: self.driveinfo['A'] = self._update_driveinfo_file(self._card_a_prefix, 'A') if self._card_b_prefix is not None: self.driveinfo['B'] = self._update_driveinfo_file(self._card_b_prefix, 'B') except OSError as e: raise OSError(_('Failed to access files on the SD card in your' ' device. This can happen for many reasons. The SD card may be' ' corrupted, it may be too large for your device, it may be' ' write-protected, etc. Try a different SD card, or reformat' ' your SD card using the FAT32 filesystem. Also make sure' ' there are not too many files in the root of your SD card.' ' Underlying error: %s') % e) return (self.get_gui_name(), '', '', '', self.driveinfo) def set_driveinfo_name(self, location_code, name): if location_code == 'main': self._update_driveinfo_file(self._main_prefix, location_code, name) elif location_code == 'A': self._update_driveinfo_file(self._card_a_prefix, location_code, name) elif location_code == 'B': self._update_driveinfo_file(self._card_b_prefix, location_code, name) def formats_to_scan_for(self): return set(self.settings().format_map) | set(self.FORMATS) def is_allowed_book_file(self, filename, path, prefix): return True def books(self, oncard=None, end_session=True): from calibre.ebooks.metadata.meta import path_to_ext debug_print('USBMS: Fetching list of books from device. Device=', self.__class__.__name__, 'oncard=', oncard) dummy_bl = self.booklist_class(None, None, None) if oncard == 'carda' and not self._card_a_prefix: self.report_progress(1.0, _('Getting list of books on device...')) return dummy_bl elif oncard == 'cardb' and not self._card_b_prefix: self.report_progress(1.0, _('Getting list of books on device...')) return dummy_bl elif oncard and oncard != 'carda' and oncard != 'cardb': self.report_progress(1.0, _('Getting list of books on device...')) return dummy_bl prefix = self._card_a_prefix if oncard == 'carda' else \ self._card_b_prefix if oncard == 'cardb' \ else self._main_prefix ebook_dirs = self.get_carda_ebook_dir() if oncard == 'carda' else \ self.EBOOK_DIR_CARD_B if oncard == 'cardb' else \ self.get_main_ebook_dir() debug_print('USBMS: dirs are:', prefix, ebook_dirs) # get the metadata cache bl = self.booklist_class(oncard, prefix, self.settings) need_sync = self.parse_metadata_cache(bl, prefix, self.METADATA_CACHE) # make a dict cache of paths so the lookup in the loop below is faster. bl_cache = {} for idx, b in enumerate(bl): bl_cache[b.lpath] = idx all_formats = self.formats_to_scan_for() def update_booklist(filename, path, prefix): changed = False # Ignore AppleDouble files if filename.startswith("._"): return False if path_to_ext(filename) in all_formats and self.is_allowed_book_file(filename, path, prefix): try: lpath = os.path.join(path, filename).partition(self.normalize_path(prefix))[2] if lpath.startswith(os.sep): lpath = lpath[len(os.sep):] lpath = lpath.replace('\\', '/') idx = bl_cache.get(lpath, None) if idx is not None: bl_cache[lpath] = None if self.update_metadata_item(bl[idx]): # print 'update_metadata_item returned true' changed = True else: if bl.add_book(self.book_from_path(prefix, lpath), replace_metadata=False): changed = True except: # Probably a filename encoding error import traceback traceback.print_exc() return changed if isinstance(ebook_dirs, string_or_bytes): ebook_dirs = [ebook_dirs] for ebook_dir in ebook_dirs: ebook_dir = self.path_to_unicode(ebook_dir) if self.SCAN_FROM_ROOT: ebook_dir = self.normalize_path(prefix) else: ebook_dir = self.normalize_path( os.path.join(prefix, *(ebook_dir.split('/'))) if ebook_dir else prefix) debug_print('USBMS: scan from root', self.SCAN_FROM_ROOT, ebook_dir) if not os.path.exists(ebook_dir): continue # Get all books in the ebook_dir directory if self.SUPPORTS_SUB_DIRS or self.SUPPORTS_SUB_DIRS_FOR_SCAN: # build a list of files to check, so we can accurately report progress flist = [] for path, dirs, files in safe_walk(ebook_dir): for filename in files: if filename != self.METADATA_CACHE: flist.append({'filename': self.path_to_unicode(filename), 'path':self.path_to_unicode(path)}) for i, f in enumerate(flist): self.report_progress(i/float(len(flist)), _('Getting list of books on device...')) changed = update_booklist(f['filename'], f['path'], prefix) if changed: need_sync = True else: paths = os.listdir(ebook_dir) for i, filename in enumerate(paths): self.report_progress((i+1) / float(len(paths)), _('Getting list of books on device...')) changed = update_booklist(self.path_to_unicode(filename), ebook_dir, prefix) if changed: need_sync = True # Remove books that are no longer in the filesystem. Cache contains # indices into the booklist if book not in filesystem, None otherwise # Do the operation in reverse order so indices remain valid for idx in sorted(itervalues(bl_cache), reverse=True, key=lambda x: -1 if x is None else x): if idx is not None: need_sync = True del bl[idx] debug_print('USBMS: count found in cache: %d, count of files in metadata: %d, need_sync: %s' % (len(bl_cache), len(bl), need_sync)) if need_sync: # self.count_found_in_bl != len(bl) or need_sync: if oncard == 'cardb': self.sync_booklists((None, None, bl)) elif oncard == 'carda': self.sync_booklists((None, bl, None)) else: self.sync_booklists((bl, None, None)) self.report_progress(1.0, _('Getting list of books on device...')) debug_print('USBMS: Finished fetching list of books from device. oncard=', oncard) return bl def upload_books(self, files, names, on_card=None, end_session=True, metadata=None): debug_print('USBMS: uploading %d books'%(len(files))) path = self._sanity_check(on_card, files) paths = [] names = iter(names) metadata = iter(metadata) for i, infile in enumerate(files): mdata, fname = next(metadata), next(names) filepath = self.normalize_path(self.create_upload_path(path, mdata, fname)) if not hasattr(infile, 'read'): infile = self.normalize_path(infile) filepath = self.put_file(infile, filepath, replace_file=True) paths.append(filepath) try: self.upload_cover(os.path.dirname(filepath), os.path.splitext(os.path.basename(filepath))[0], mdata, filepath) except: # Failure to upload cover is not catastrophic import traceback traceback.print_exc() self.report_progress((i+1) / float(len(files)), _('Transferring books to device...')) self.report_progress(1.0, _('Transferring books to device...')) debug_print('USBMS: finished uploading %d books'%(len(files))) return list(zip(paths, cycle([on_card]))) def upload_cover(self, path, filename, metadata, filepath): ''' Upload book cover to the device. Default implementation does nothing. :param path: The full path to the folder where the associated book is located. :param filename: The name of the book file without the extension. :param metadata: metadata belonging to the book. Use metadata.thumbnail for cover :param filepath: The full path to the e-book file ''' pass def add_books_to_metadata(self, locations, metadata, booklists): debug_print('USBMS: adding metadata for %d books'%(len(metadata))) metadata = iter(metadata) locations = tuple(locations) for i, location in enumerate(locations): self.report_progress((i+1) / float(len(locations)), _('Adding books to device metadata listing...')) info = next(metadata) blist = 2 if location[1] == 'cardb' else 1 if location[1] == 'carda' else 0 # Extract the correct prefix from the pathname. To do this correctly, # we must ensure that both the prefix and the path are normalized # so that the comparison will work. Book's __init__ will fix up # lpath, so we don't need to worry about that here. path = self.normalize_path(location[0]) if self._main_prefix: prefix = self._main_prefix if \ path.startswith(self.normalize_path(self._main_prefix)) else None if not prefix and self._card_a_prefix: prefix = self._card_a_prefix if \ path.startswith(self.normalize_path(self._card_a_prefix)) else None if not prefix and self._card_b_prefix: prefix = self._card_b_prefix if \ path.startswith(self.normalize_path(self._card_b_prefix)) else None if prefix is None: prints('in add_books_to_metadata. Prefix is None!', path, self._main_prefix) continue lpath = path.partition(prefix)[2] if lpath.startswith('/') or lpath.startswith('\\'): lpath = lpath[1:] book = self.book_class(prefix, lpath, other=info) if book.size is None: book.size = os.stat(self.normalize_path(path)).st_size b = booklists[blist].add_book(book, replace_metadata=True) if b: b._new_book = True self.report_progress(1.0, _('Adding books to device metadata listing...')) debug_print('USBMS: finished adding metadata') def delete_single_book(self, path): os.unlink(path) def delete_extra_book_files(self, path): filepath = os.path.splitext(path)[0] for ext in self.DELETE_EXTS: for x in (filepath, path): x += ext if os.path.exists(x): if os.path.isdir(x): shutil.rmtree(x, ignore_errors=True) else: os.unlink(x) if self.SUPPORTS_SUB_DIRS: try: os.removedirs(os.path.dirname(path)) except: pass def delete_books(self, paths, end_session=True): debug_print('USBMS: deleting %d books'%(len(paths))) for i, path in enumerate(paths): self.report_progress((i+1) / float(len(paths)), _('Removing books from device...')) path = self.normalize_path(path) if os.path.exists(path): # Delete the ebook self.delete_single_book(path) self.delete_extra_book_files(path) self.report_progress(1.0, _('Removing books from device...')) debug_print('USBMS: finished deleting %d books'%(len(paths))) def remove_books_from_metadata(self, paths, booklists): debug_print('USBMS: removing metadata for %d books'%(len(paths))) for i, path in enumerate(paths): self.report_progress((i+1) / float(len(paths)), _('Removing books from device metadata listing...')) for bl in booklists: for book in bl: if path.endswith(book.path): bl.remove_book(book) self.report_progress(1.0, _('Removing books from device metadata listing...')) debug_print('USBMS: finished removing metadata for %d books'%(len(paths))) # If you override this method and you use book._new_book, then you must # complete the processing before you call this method. The flag is cleared # at the end just before the return def sync_booklists(self, booklists, end_session=True): debug_print('USBMS: starting sync_booklists') json_codec = JsonCodec() if not os.path.exists(self.normalize_path(self._main_prefix)): os.makedirs(self.normalize_path(self._main_prefix)) def write_prefix(prefix, listid): if (prefix is not None and len(booklists) > listid and isinstance(booklists[listid], self.booklist_class)): if not os.path.exists(prefix): os.makedirs(self.normalize_path(prefix)) with open(self.normalize_path(os.path.join(prefix, self.METADATA_CACHE)), 'wb') as f: json_codec.encode_to_file(f, booklists[listid]) fsync(f) write_prefix(self._main_prefix, 0) write_prefix(self._card_a_prefix, 1) write_prefix(self._card_b_prefix, 2) # Clear the _new_book indication, as we are supposed to be done with # adding books at this point for blist in booklists: if blist is not None: for book in blist: book._new_book = False self.report_progress(1.0, _('Sending metadata to device...')) debug_print('USBMS: finished sync_booklists') @classmethod def build_template_regexp(cls): from calibre.devices.utils import build_template_regexp return build_template_regexp(cls.save_template()) @classmethod def path_to_unicode(cls, path): if isbytestring(path): path = path.decode(filesystem_encoding) return path @classmethod def normalize_path(cls, path): 'Return path with platform native path separators' if path is None: return None if os.sep == '\\': path = path.replace('/', '\\') else: path = path.replace('\\', '/') return cls.path_to_unicode(path) @classmethod def parse_metadata_cache(cls, bl, prefix, name): json_codec = JsonCodec() need_sync = False cache_file = cls.normalize_path(os.path.join(prefix, name)) if os.access(cache_file, os.R_OK): try: with open(cache_file, 'rb') as f: json_codec.decode_from_file(f, bl, cls.book_class, prefix) except: import traceback traceback.print_exc() bl = [] need_sync = True else: need_sync = True return need_sync @classmethod def update_metadata_item(cls, book): changed = False size = os.stat(cls.normalize_path(book.path)).st_size if size != book.size: changed = True mi = cls.metadata_from_path(book.path) book.smart_update(mi) book.size = size return changed @classmethod def metadata_from_path(cls, path): return cls.metadata_from_formats([path]) @classmethod def metadata_from_formats(cls, fmts): from calibre.customize.ui import quick_metadata from calibre.ebooks.metadata.meta import metadata_from_formats with quick_metadata: return metadata_from_formats(fmts, force_read_metadata=True, pattern=cls.build_template_regexp()) @classmethod def book_from_path(cls, prefix, lpath): from calibre.ebooks.metadata.book.base import Metadata if cls.settings().read_metadata or cls.MUST_READ_METADATA: mi = cls.metadata_from_path(cls.normalize_path(os.path.join(prefix, lpath))) else: from calibre.ebooks.metadata.meta import metadata_from_filename mi = metadata_from_filename(cls.normalize_path(os.path.basename(lpath)), cls.build_template_regexp()) if mi is None: mi = Metadata(os.path.splitext(os.path.basename(lpath))[0], [_('Unknown')]) size = os.stat(cls.normalize_path(os.path.join(prefix, lpath))).st_size book = cls.book_class(prefix, lpath, other=mi, size=size) return book
24,387
Python
.py
485
37.552577
114
0.581847
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,803
deviceconfig.py
kovidgoyal_calibre/src/calibre/devices/usbms/deviceconfig.py
__license__ = 'GPL 3' __copyright__ = '2009, John Schember <john@nachtimwald.com>' __docformat__ = 'restructuredtext en' from calibre.utils.config_base import Config, ConfigProxy class DeviceConfig: HELP_MESSAGE = _('Configure device') #: Can be None, a string or a list of strings. When it is a string #: that string is used for the help text and the actual customization value #: can be read from ``dev.settings().extra_customization``. #: If it a list of strings, then dev.settings().extra_customization will #: also be a list. In this case, you *must* ensure that #: EXTRA_CUSTOMIZATION_DEFAULT is also a list. The list can contain either #: boolean values or strings, in which case a checkbox or line edit will be #: used for them in the config widget, automatically. #: If a string contains ::: then the text after it is interpreted as the #: tooltip EXTRA_CUSTOMIZATION_MESSAGE = None #: The default value for extra customization. If you set #: EXTRA_CUSTOMIZATION_MESSAGE you *must* set this as well. EXTRA_CUSTOMIZATION_DEFAULT = None #: A dictionary providing choices for options that should be displayed as a #: combo box to the user. The dictionary maps extra #: customization indexes #: to a set of choices. EXTRA_CUSTOMIZATION_CHOICES = None SUPPORTS_SUB_DIRS = False SUPPORTS_SUB_DIRS_FOR_SCAN = False # This setting is used when scanning for books when SUPPORTS_SUB_DIRS is False SUPPORTS_SUB_DIRS_DEFAULT = True MUST_READ_METADATA = False SUPPORTS_USE_AUTHOR_SORT = False #: If None the default is used SAVE_TEMPLATE = None #: If True the user can add new formats to the driver USER_CAN_ADD_NEW_FORMATS = True @classmethod def _default_save_template(cls): from calibre.library.save_to_disk import config return cls.SAVE_TEMPLATE if cls.SAVE_TEMPLATE else \ config().parse().send_template @classmethod def _config_base_name(cls): klass = cls if isinstance(cls, type) else cls.__class__ return klass.__name__ @classmethod def _config(cls): name = cls._config_base_name() c = Config('device_drivers_%s' % name, _('settings for device drivers')) c.add_opt('format_map', default=cls.FORMATS, help=_('Ordered list of formats the device will accept')) c.add_opt('use_subdirs', default=cls.SUPPORTS_SUB_DIRS_DEFAULT, help=_('Place files in sub-folders if the device supports them')) c.add_opt('read_metadata', default=True, help=_('Read metadata from files on device')) c.add_opt('use_author_sort', default=False, help=_('Use author sort instead of author')) c.add_opt('save_template', default=cls._default_save_template(), help=_('Template to control how books are saved')) c.add_opt('extra_customization', default=cls.EXTRA_CUSTOMIZATION_DEFAULT, help=_('Extra customization')) return c @classmethod def _configProxy(cls): return ConfigProxy(cls._config()) @classmethod def config_widget(cls): from calibre.gui2.device_drivers.configwidget import ConfigWidget cw = ConfigWidget(cls.settings(), cls.FORMATS, cls.SUPPORTS_SUB_DIRS, cls.MUST_READ_METADATA, cls.SUPPORTS_USE_AUTHOR_SORT, cls.EXTRA_CUSTOMIZATION_MESSAGE, cls, extra_customization_choices=cls.EXTRA_CUSTOMIZATION_CHOICES) return cw @classmethod def save_settings(cls, config_widget): proxy = cls._configProxy() proxy['format_map'] = config_widget.format_map() if cls.SUPPORTS_SUB_DIRS: proxy['use_subdirs'] = config_widget.use_subdirs() if not cls.MUST_READ_METADATA: proxy['read_metadata'] = config_widget.read_metadata() if cls.SUPPORTS_USE_AUTHOR_SORT: proxy['use_author_sort'] = config_widget.use_author_sort() if cls.EXTRA_CUSTOMIZATION_MESSAGE: if isinstance(cls.EXTRA_CUSTOMIZATION_MESSAGE, list): ec = [] for i in range(0, len(cls.EXTRA_CUSTOMIZATION_MESSAGE)): if config_widget.opt_extra_customization[i] is None: ec.append(None) continue if hasattr(config_widget.opt_extra_customization[i], 'isChecked'): ec.append(config_widget.opt_extra_customization[i].isChecked()) elif hasattr(config_widget.opt_extra_customization[i], 'currentText'): ec.append(str(config_widget.opt_extra_customization[i].currentText()).strip()) else: ec.append(str(config_widget.opt_extra_customization[i].text()).strip()) else: ec = str(config_widget.opt_extra_customization.text()).strip() if not ec: ec = None proxy['extra_customization'] = ec st = str(config_widget.opt_save_template.text()) proxy['save_template'] = st @classmethod def migrate_extra_customization(cls, vals): return vals @classmethod def settings(cls): opts = cls._config().parse() if isinstance(cls.EXTRA_CUSTOMIZATION_DEFAULT, list): if opts.extra_customization is None: opts.extra_customization = [] if not isinstance(opts.extra_customization, list): opts.extra_customization = [opts.extra_customization] for i,d in enumerate(cls.EXTRA_CUSTOMIZATION_DEFAULT): if i >= len(opts.extra_customization): opts.extra_customization.append(d) opts.extra_customization = cls.migrate_extra_customization(opts.extra_customization) return opts @classmethod def save_template(cls): st = cls.settings().save_template if st: return st else: return cls._default_save_template() @classmethod def customization_help(cls, gui=False): return cls.HELP_MESSAGE
6,191
Python
.py
126
39.238095
118
0.639312
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,804
__init__.py
kovidgoyal_calibre/src/calibre/web/__init__.py
__license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' class Recipe: pass def get_download_filename_from_response(response): from polyglot.urllib import unquote, urlparse filename = last_part_name = '' try: purl = urlparse(response.geturl()) last_part_name = unquote(purl.path.split('/')[-1]) disposition = response.info().get('Content-disposition', '') if isinstance(disposition, bytes): disposition = disposition.decode('utf-8', 'replace') for p in disposition.split(';'): if 'filename' in p: if '*=' in disposition: parts = disposition.split('*=')[-1] filename = parts.split('\'')[-1] else: filename = disposition.split('=')[-1] if filename[0] in ('\'', '"'): filename = filename[1:] if filename[-1] in ('\'', '"'): filename = filename[:-1] filename = unquote(filename) break except Exception: import traceback traceback.print_exc() return filename or last_part_name def get_download_filename(url, cookie_file=None): ''' Get a local filename for a URL using the content disposition header Returns empty string if an error occurs. ''' from contextlib import closing from calibre import browser filename = '' br = browser() if cookie_file: from mechanize import MozillaCookieJar cj = MozillaCookieJar() cj.load(cookie_file) br.set_cookiejar(cj) try: with closing(br.open(url)) as r: filename = get_download_filename_from_response(r) except: import traceback traceback.print_exc() return filename
1,856
Python
.py
51
27.019608
71
0.578595
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,805
nytimes.py
kovidgoyal_calibre/src/calibre/web/site_parsers/nytimes.py
#!/usr/bin/env python # License: GPL v3 Copyright: 2022, Kovid Goyal <kovid at kovidgoyal.net> import json import re import sys from pprint import pprint from xml.sax.saxutils import escape, quoteattr from calibre.utils.iso8601 import parse_iso8601 module_version = 11 # needed for live updates pprint def parse_image(i): crop = i.get('crops') or i.get('spanImageCrops') if crop: yield f'<div><img src="{crop[0]["renditions"][0]["url"]}" title="{i.get("altText", "")}">' if i.get('caption'): yield f'<div class="cap">{"".join(parse_types(i["caption"]))}' if i.get('credit'): yield f'<span class="cred"> {i["credit"]}</span>' yield '</div>' elif i.get('legacyHtmlCaption'): if i['legacyHtmlCaption'].strip(): yield f'<div class="cap">{i["legacyHtmlCaption"]}</div>' yield '</div>' def parse_img_grid(g): for grd in g.get('gridMedia', {}): yield ''.join(parse_image(grd)) if g.get('caption'): yield f'<div class="cap">{g["caption"]}' if g.get('credit'): yield f'<span class="cred"> {g["credit"]}</span>' yield '</div>' def parse_vid(v): if v.get('promotionalMedia'): headline = v.get('headline', {}).get('default', '') rendition = v.get('renditions') yield ( f'<div><b><a href="{rendition[0]["url"]}">Video</a>: {headline}</b></div>' if rendition else f'<div><b>{headline}</b></div>' ) yield ''.join(parse_types(v['promotionalMedia'])) if v.get('promotionalSummary'): yield f'<div class="cap">{v["promotionalSummary"]}</div>' def parse_emb(e): if e.get('html') and 'datawrapper.dwcdn.net' in e.get('html', ''): dw = re.search(r'datawrapper.dwcdn.net/(.{5})', e['html']).group(1) yield f'<div><img src="https://datawrapper.dwcdn.net/{dw}/full.png"></div>' elif e.get('promotionalMedia'): if e.get('headline'): yield f'<div><b>{e["headline"]["default"]}</b></div>' yield ''.join(parse_types(e['promotionalMedia'])) if e.get('note'): yield f'<div class="cap">{e["note"]}</div>' def parse_byline(byl): for b in byl.get('bylines', {}): yield f'<div><b>{b["renderedRepresentation"]}</b></div>' yield '<div><i>' for rl in byl.get('role', {}): if ''.join(parse_cnt(rl)).strip(): yield ''.join(parse_cnt(rl)) yield '</i></div>' def iso_date(x): dt = parse_iso8601(x, as_utc=False) return dt.strftime('%b %d, %Y at %I:%M %p') def parse_header(h): if h.get('label'): yield f'<div class="lbl">{"".join(parse_types(h["label"]))}</div>' if h.get('headline'): yield ''.join(parse_types(h['headline'])) if h.get('summary'): yield f'<p><i>{"".join(parse_types(h["summary"]))}</i></p>' if h.get('ledeMedia'): yield ''.join(parse_types(h['ledeMedia'])) if h.get('byline'): yield ''.join(parse_types(h['byline'])) if h.get('timestampBlock'): yield ''.join(parse_types(h['timestampBlock'])) def parse_fmt_type(fm): for f in fm.get('formats', {}): ftype = f.get('__typename', '') if ftype == 'BoldFormat': yield '<strong>' if ftype == 'ItalicFormat': yield '<em>' if ftype == 'LinkFormat': hrf = f['url'] yield f'<a href="{hrf}">' yield fm.get('text', '') for f in reversed(fm.get('formats', {})): ftype = f.get('__typename', '') if ftype == 'BoldFormat': yield '</strong>' if ftype == 'ItalicFormat': yield '</em>' if ftype == 'LinkFormat': yield '</a>' def parse_cnt(cnt): for k in cnt: if isinstance(cnt[k], list): if k == 'formats': yield ''.join(parse_fmt_type(cnt)) else: for cnt_ in cnt[k]: yield ''.join(parse_types(cnt_)) if isinstance(cnt[k], dict): yield ''.join(parse_types(cnt[k])) if cnt.get('text') and 'formats' not in cnt and 'content' not in cnt: if isinstance(cnt['text'], str): yield cnt['text'] def parse_types(x): typename = x.get('__typename', '') align = '' if x.get('textAlign'): align = f' style="text-align: {x["textAlign"].lower()};"' if 'Header' in typename: yield '\n'.join(parse_header(x)) elif typename.startswith('Heading'): htag = 'h' + re.match(r'Heading([1-6])Block', typename).group(1) yield f'<{htag}{align}>{"".join(parse_cnt(x))}</{htag}>' elif typename == 'ParagraphBlock': yield f'<p>{"".join(parse_cnt(x))}</p>' elif typename in {'DetailBlock', 'TextRunKV'}: yield f'<p style="font-size: small;">{"".join(parse_cnt(x))}</p>' elif typename == 'BylineBlock': yield f'<div class="byl"><br/>{"".join(parse_byline(x))}</div>' elif typename == 'LabelBlock': yield f'<div class="sc">{"".join(parse_cnt(x))}</div>' elif typename == 'BlockquoteBlock': yield f'<blockquote>{"".join(parse_cnt(x))}</blockquote>' elif typename == 'TimestampBlock': yield f'<div class="time">{iso_date(x["timestamp"])}</div>' elif typename == 'LineBreakInline': yield '<br/>' elif typename == 'RuleBlock': yield '<hr/>' elif typename == 'Image': yield ''.join(parse_image(x)) elif typename == 'GridBlock': yield ''.join(parse_img_grid(x)) elif typename == 'Video': yield ''.join(parse_vid(x)) elif typename == 'EmbeddedInteractive': yield ''.join(parse_emb(x)) elif typename == 'ListBlock': yield f'\n<ul>{"".join(parse_cnt(x))}</ul>' elif typename == 'ListItemBlock': yield f'\n<li>{"".join(parse_cnt(x))}</li>' elif typename and typename not in { 'RelatedLinksBlock', 'EmailSignupBlock', 'Dropzone', 'AudioBlock', }: yield ''.join(parse_cnt(x)) def article_parse(data): yield '<html><body>' for d in data: yield from parse_types(d) yield '</body></html>' def json_to_html(raw): data = json.loads(raw.replace(':undefined', ':null')) # open('/t/raw.json', 'w').write(json.dumps(data, indent=2)) try: data = data['initialData']['data'] except TypeError: data = data['initialState'] return live_json_to_html(data) content = data['article']['sprinkledBody']['content'] return '\n'.join(article_parse(content)) def add_live_item(item, item_type, lines): a = lines.append if item_type == 'text': a('<p>' + item['value'] + '</p>') elif item_type == 'list': a('<li>' + item['value'] + '</li>') elif item_type == 'bulletedList': a('<ul>') for x in item['value']: a('<li>' + x + '</li>') a('</ul>') elif item_type == 'items': for x in item['value']: a('<h5>' + x['subtitle'] + '</h5>') add_live_item({'value': x['text']}, 'text', lines) elif item_type == 'section': for item in item['value']: add_live_item(item, item['type'], lines) elif item_type == '': b = item if b.get('title'): a('<h3>' + b['title'] + '</h3>') if b.get('imageUrl'): a('<div><img src=' + quoteattr(b['imageUrl']) + '/></div>') if b.get('leadIn'): a('<p>' + b['leadIn'] + '</p>') if 'items' in b: add_live_item({'value': b['items']}, 'items', lines) return if 'bulletedList' in b: add_live_item({'value': b['bulletedList']}, 'bulletedList', lines) return if 'sections' in b: for section in b['sections']: add_live_item({'value': section['section']}, 'section', lines) return raise Exception('Unknown item: %s' % b) else: raise Exception('Unknown item: %s' % b) def live_json_to_html(data): for k, v in data['ROOT_QUERY'].items(): if isinstance(v, dict) and 'id' in v: root = data[v['id']] s = data[root['storylines'][0]['id']] s = data[s['storyline']['id']] title = s['displayName'] lines = ['<h1>' + escape(title) + '</h1>'] for b in json.loads(s['experimentalJsonBlob'])['data'][0]['data']: b = b['data'] if isinstance(b, list): for x in b: add_live_item(x, x['type'], lines) else: add_live_item(b, '', lines) return '<html><body>' + '\n'.join(lines) + '</body></html>' def extract_html(soup, url): if '/interactive/' in url: return ( '<html><body><p><em>' + 'This is an interactive article, which is supposed to be read in a browser.' + '</p></em></body></html>' ) script = soup.findAll('script', text=lambda x: x and 'window.__preloadedData' in x)[0] script = str(script) raw = script[script.find('{') : script.rfind(';')].strip().rstrip(';') return json_to_html(raw) def download_url_from_wayback(category, url, br=None): from mechanize import Request host = 'http://localhost:8090' host = 'https://wayback1.calibre-ebook.com' rq = Request( host + '/' + category, data=json.dumps({'url': url}), headers={'User-Agent': 'calibre', 'Content-Type': 'application/json'}, ) if br is None: from calibre import browser br = browser() br.set_handle_gzip(True) return br.open_novisit(rq, timeout=3 * 60).read() def download_url(url=None, br=None): # Get the URL from the Wayback machine if url is None: url = sys.argv[-1] return download_url_from_wayback('nytimes', url, br) if __name__ == '__main__': f = sys.argv[-1] raw = open(f).read() if f.endswith('.html'): from calibre.ebooks.BeautifulSoup import BeautifulSoup soup = BeautifulSoup(raw) print(extract_html(soup)) else: print(json_to_html(raw))
10,118
Python
.py
263
30.745247
98
0.54999
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,806
natgeo.py
kovidgoyal_calibre/src/calibre/web/site_parsers/natgeo.py
#!/usr/bin/env python from __future__ import absolute_import, division, print_function, unicode_literals import json from pprint import pprint from calibre import prepare_string_for_xml as escape from calibre.utils.iso8601 import parse_iso8601 module_version = 1 # needed for live updates pprint def extract_json(raw): s = raw.find("window['__natgeo__']") script = raw[s : raw.find('</script>', s)] content = json.loads(script[script.find('{') :].rstrip(';'))['page']['content'] if content.get('prismarticle'): return content['prismarticle'] if content.get('article'): return content['article'] def parse_contributors(grp): for item in grp: line = '<div class="auth">' + escape(item['title']) + ' ' for c in item['contributors']: line += escape(c['displayName']) yield line + '</div>' def parse_lead_image(media): if 'image' in media: yield '<p>' if 'dsc' in media['image']: yield ( f'<div><img src="{escape(media["image"]["src"], True)}" ' f'alt="{escape(media["image"]["dsc"], True)}"></div>' ) else: yield f'<div><img src="{escape(media["image"]["src"], True)}"></div>' if 'caption' in media and 'credit' in media: yield ( '<div class="cap">' + media['caption'] + '<span class="cred"> ' + media['credit'] + '</span></div>' ) elif 'caption' in media: yield '<div class="cap">' + media['caption'] + '</div>' yield '</p>' def parse_inline(inl): if inl.get('content', {}).get('name', '') == 'Image': props = inl['content']['props'] yield '<p>' if 'image' in props: yield f'<div class="img"><img src="{props["image"]["src"]}"></div>' if 'caption' in props: yield ( f'<div class="cap">{props["caption"].get("text", "")}<span ' f'class="cred"> {props["caption"].get("credit", "")}</span></div>' ) yield '</p>' if inl.get('content', {}).get('name', '') == 'ImageGroup': if 'images' in inl['content']['props']: for imgs in inl['content']['props']['images']: yield '<p>' if 'src' in imgs: yield f'<div class="img"><img src="{imgs["src"]}"></div>' if 'caption' in imgs: yield ( f'<div class="cap">{imgs["caption"].get("text", "")}<span ' f'class="cred"> {imgs["caption"].get("credit", "")}</span></div>' ) yield '</p>' def parse_cont(content): for cont in content.get('content', {}): if isinstance(cont, dict): yield from parse_body(cont) if isinstance(cont, str): yield cont def parse_body(x): if isinstance(x, dict): if 'type' in x: tag = x['type'] if tag == 'inline': yield ''.join(parse_inline(x)) elif 'attrs' in x and 'href' in x.get('attrs', ''): yield '<' + tag + f' href="{x["attrs"]["href"]}">' yield from parse_cont(x) yield '</' + tag + '>' else: yield '<' + tag + '>' yield from parse_cont(x) yield '</' + tag + '>' elif isinstance(x, list): for y in x: if isinstance(y, dict): yield from parse_body(y) def parse_bdy(item): c = item['cntnt'] if item.get('type') == 'inline': if c.get('cmsType') == 'listicle': if 'title' in c: yield '<h3>' + escape(c['title']) + '</h3>' yield c['text'] elif c.get('cmsType') == 'image': yield from parse_lead_image(c) elif c.get('cmsType') == 'imagegroup': for imgs in c['images']: yield from parse_lead_image(imgs) elif c.get('cmsType') == 'pullquote': if 'quote' in c: yield '<blockquote>' + c['quote'] + '</blockquote>' elif c.get('cmsType') == 'editorsNote': if 'note' in c: yield '<blockquote>' + c['note'] + '</blockquote>' else: if c['mrkup'].strip().startswith('<'): yield c['mrkup'] else: yield '<{tag}>{markup}</{tag}>'.format( tag=item['type'], markup=c['mrkup']) def parse_article(edg): sc = edg['schma'] yield '<div class="sub">' + escape(edg['sctn']) + '</div>' yield '<h1>' + escape(sc['sclTtl']) + '</h1>' if sc.get('sclDsc'): yield '<div class="byline">' + escape(sc['sclDsc']) + '</div>' yield '<p>' yield from parse_contributors(edg.get('cntrbGrp', {})) ts = parse_iso8601(edg['mdDt'], as_utc=False).strftime('%B %d, %Y') yield '<div class="time">Published: ' + escape(ts) + '</div>' if 'readTime' in edg: yield '<div class="time">' + escape(edg['readTime']) + '</div>' yield '</p>' if edg.get('ldMda', {}).get('cmsType') == 'image': yield from parse_lead_image(edg['ldMda']) if edg.get('prismData'): for main in edg['prismData']['mainComponents']: if main['name'] == 'Body': for item in main['props']['body']: if isinstance(item, dict): if item.get('type', '') == 'inline': yield ''.join(parse_inline(item)) elif isinstance(item, list): for line in item: yield ''.join(parse_body(line)) elif edg.get('bdy'): for item in edg['bdy']: yield from parse_bdy(item) def article_parse(data): yield '<html><body>' for frm in data['frms']: if not frm: continue for mod in frm.get('mods', ()): for edg in mod.get('edgs', ()): if edg.get('cmsType') == 'ImmersiveLeadTile': if 'image' in edg.get('cmsImage', {}): yield from parse_lead_image(edg['cmsImage']) if edg.get('cmsType') == 'ArticleBodyTile': yield from parse_article(edg) yield '</body></html>' def extract_html(raw_html): data = extract_json(raw_html) return '\n'.join(article_parse(data))
6,472
Python
.py
159
29.805031
89
0.492926
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,807
utils.py
kovidgoyal_calibre/src/calibre/web/fetch/utils.py
#!/usr/bin/env python # License: GPLv3 Copyright: 2016, Kovid Goyal <kovid at kovidgoyal.net> from calibre.utils.img import blend_on_canvas, image_from_data, image_to_data, scale_image def rescale_image(data, scale_news_images, compress_news_images_max_size, compress_news_images_auto_size): orig_data = data # save it in case compression fails img = image_from_data(data) orig_w, orig_h = img.width(), img.height() if scale_news_images is not None: wmax, hmax = scale_news_images if wmax < orig_w or hmax < orig_h: orig_w, orig_h, data = scale_image(img, wmax, hmax, compression_quality=95) if compress_news_images_max_size is None: if compress_news_images_auto_size is None: # not compressing return data maxsizeb = (orig_w * orig_h)/compress_news_images_auto_size else: maxsizeb = compress_news_images_max_size * 1024 if len(data) <= maxsizeb: # no compression required return data scaled_data = data # save it in case compression fails quality = 90 while len(data) >= maxsizeb and quality >= 5: data = image_to_data(image_from_data(scaled_data), compression_quality=quality) quality -= 5 if len(data) >= len(scaled_data): # compression failed return orig_data if len(orig_data) <= len(scaled_data) else scaled_data if len(data) >= len(orig_data): # no improvement return orig_data return data def prepare_masthead_image(path_to_image, out_path, mi_width, mi_height): with open(path_to_image, 'rb') as f: img = image_from_data(f.read()) img = blend_on_canvas(img, mi_width, mi_height) with open(out_path, 'wb') as f: f.write(image_to_data(img)) if __name__ == '__main__': import sys data = sys.stdin.read() sys.stdout.write(rescale_image(data, (768, 1024), None, 8))
1,885
Python
.py
39
41.974359
106
0.66503
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,808
simple.py
kovidgoyal_calibre/src/calibre/web/fetch/simple.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' ''' Fetch a webpage and its links recursively. The webpages are saved to disk in UTF-8 encoding with any charset declarations removed. ''' import os import re import socket import sys import threading import time import traceback from base64 import standard_b64decode from urllib.request import urlopen from calibre import browser, relpath, unicode_path from calibre.constants import filesystem_encoding, iswindows from calibre.ebooks.BeautifulSoup import BeautifulSoup from calibre.ebooks.chardet import xml_to_unicode from calibre.utils.config import OptionParser from calibre.utils.filenames import ascii_filename from calibre.utils.imghdr import what from calibre.utils.localization import _ from calibre.utils.logging import Log from calibre.web.fetch.utils import rescale_image from polyglot.http_client import responses from polyglot.urllib import URLError, quote, url2pathname, urljoin, urlparse, urlsplit, urlunparse, urlunsplit class AbortArticle(Exception): pass class FetchError(Exception): pass class closing: 'Context to automatically close something at the end of a block.' def __init__(self, thing): self.thing = thing def __enter__(self): return self.thing def __exit__(self, *exc_info): try: self.thing.close() except Exception: pass def canonicalize_url(url): # mechanize does not handle quoting automatically if re.search(r'\s+', url) is not None: purl = list(urlparse(url)) for i in range(2, 6): purl[i] = quote(purl[i]) url = urlunparse(purl) return url bad_url_counter = 0 def basename(url): try: parts = urlsplit(url) path = url2pathname(parts.path) res = os.path.basename(path) except: global bad_url_counter bad_url_counter += 1 return 'bad_url_%d.html'%bad_url_counter if not os.path.splitext(res)[1]: return 'index.html' return res def save_soup(soup, target): for meta in soup.findAll('meta', content=True): if 'charset' in meta['content'].lower(): meta.extract() for meta in soup.findAll('meta', charset=True): meta.extract() head = soup.find('head') if head is not None: nm = soup.new_tag('meta', charset='utf-8') head.insert(0, nm) selfdir = os.path.dirname(target) for tag in soup.findAll(['img', 'link', 'a']): for key in ('src', 'href'): path = tag.get(key, None) if path and os.path.isfile(path) and os.path.exists(path) and os.path.isabs(path): tag[key] = unicode_path(relpath(path, selfdir).replace(os.sep, '/')) html = str(soup) with open(target, 'wb') as f: f.write(html.encode('utf-8')) class response(bytes): def __new__(cls, *args): obj = super().__new__(cls, *args) obj.newurl = None return obj def default_is_link_wanted(url, tag): raise NotImplementedError() class RecursiveFetcher: LINK_FILTER = tuple(re.compile(i, re.IGNORECASE) for i in ('.exe\\s*$', '.mp3\\s*$', '.ogg\\s*$', '^\\s*mailto:', '^\\s*$')) # ADBLOCK_FILTER = tuple(re.compile(i, re.IGNORECASE) for it in # ( # # ) # ) CSS_IMPORT_PATTERN = re.compile(r'\@import\s+url\((.*?)\)', re.IGNORECASE) default_timeout = socket.getdefaulttimeout() # Needed here as it is used in __del__ def __init__(self, options, log, image_map={}, css_map={}, job_info=None): bd = options.dir if not isinstance(bd, str): bd = bd.decode(filesystem_encoding) self.base_dir = os.path.abspath(os.path.expanduser(bd)) if not os.path.exists(self.base_dir): os.makedirs(self.base_dir) self.log = log self.verbose = options.verbose self.timeout = options.timeout self.encoding = options.encoding self.browser = options.browser if hasattr(options, 'browser') else browser() self.max_recursions = options.max_recursions self.match_regexps = [re.compile(i, re.IGNORECASE) for i in options.match_regexps] self.filter_regexps = [re.compile(i, re.IGNORECASE) for i in options.filter_regexps] self.max_files = options.max_files self.delay = options.delay self.last_fetch_at = 0. self.filemap = {} self.imagemap = image_map self.imagemap_lock = threading.RLock() self.stylemap = css_map self.image_url_processor = None self.stylemap_lock = threading.RLock() self.downloaded_paths = [] self.current_dir = self.base_dir self.files = 0 self.preprocess_regexps = getattr(options, 'preprocess_regexps', []) self.remove_tags = getattr(options, 'remove_tags', []) self.remove_tags_after = getattr(options, 'remove_tags_after', None) self.remove_tags_before = getattr(options, 'remove_tags_before', None) self.keep_only_tags = getattr(options, 'keep_only_tags', []) self.preprocess_html_ext = getattr(options, 'preprocess_html', lambda soup: soup) self.preprocess_raw_html = getattr(options, 'preprocess_raw_html', lambda raw, url: raw) self.prepreprocess_html_ext = getattr(options, 'skip_ad_pages', lambda soup: None) self.postprocess_html_ext = getattr(options, 'postprocess_html', None) self.preprocess_image_ext = getattr(options, 'preprocess_image', None) self._is_link_wanted = getattr(options, 'is_link_wanted', default_is_link_wanted) self.compress_news_images_max_size = getattr(options, 'compress_news_images_max_size', None) self.compress_news_images = getattr(options, 'compress_news_images', False) self.compress_news_images_auto_size = getattr(options, 'compress_news_images_auto_size', 16) self.scale_news_images = getattr(options, 'scale_news_images', None) self.get_delay = getattr(options, 'get_delay', lambda url: self.delay) self.download_stylesheets = not options.no_stylesheets self.show_progress = True self.failed_links = [] self.job_info = job_info self.preloaded_urls = {} def get_soup(self, src, url=None): nmassage = [] nmassage.extend(self.preprocess_regexps) # Remove comments as they can leave detritus when extracting tags leaves # multiple nested comments nmassage.append((re.compile(r'<!--.*?-->', re.DOTALL), lambda m: '')) usrc = xml_to_unicode(src, self.verbose, strip_encoding_pats=True)[0] usrc = self.preprocess_raw_html(usrc, url) for pat, repl in nmassage: usrc = pat.sub(repl, usrc) soup = BeautifulSoup(usrc) replace = self.prepreprocess_html_ext(soup) if replace is not None: replace = xml_to_unicode(replace, self.verbose, strip_encoding_pats=True)[0] for pat, repl in nmassage: replace = pat.sub(repl, replace) soup = BeautifulSoup(replace) if self.keep_only_tags: body = soup.new_tag('body') try: if isinstance(self.keep_only_tags, dict): self.keep_only_tags = [self.keep_only_tags] for spec in self.keep_only_tags: for tag in soup.find('body').findAll(**spec): body.insert(len(body.contents), tag) soup.find('body').replaceWith(body) except AttributeError: # soup has no body element pass def remove_beyond(tag, next): while tag is not None and getattr(tag, 'name', None) != 'body': after = getattr(tag, next) while after is not None: ns = getattr(tag, next) after.extract() after = ns tag = tag.parent if self.remove_tags_after is not None: rt = [self.remove_tags_after] if isinstance(self.remove_tags_after, dict) else self.remove_tags_after for spec in rt: tag = soup.find(**spec) remove_beyond(tag, 'nextSibling') if self.remove_tags_before is not None: rt = [self.remove_tags_before] if isinstance(self.remove_tags_before, dict) else self.remove_tags_before for spec in rt: tag = soup.find(**spec) remove_beyond(tag, 'previousSibling') for kwds in self.remove_tags: for tag in soup.findAll(**kwds): tag.extract() return self.preprocess_html_ext(soup) def fetch_url(self, url): data = None q = self.preloaded_urls.pop(url, None) if q is not None: ans = response(q) ans.newurl = url return ans st = time.monotonic() is_data_url = url.startswith('data:') if not is_data_url: self.log.debug('Fetching', url) # Check for a URL pointing to the local filesystem and special case it # for efficiency and robustness. Bypasses delay checking as it does not # apply to local fetches. Ensures that unicode paths that are not # representable in the filesystem_encoding work. if is_data_url: payload = url.partition(',')[2] return standard_b64decode(payload) is_local = 0 if url.startswith('file://'): is_local = 7 elif url.startswith('file:'): is_local = 5 if is_local > 0: url = url[is_local:] if iswindows and url.startswith('/'): url = url[1:] with open(url, 'rb') as f: data = response(f.read()) data.newurl = 'file:'+url # This is what mechanize does for # local URLs self.log.debug(f'Fetched {url} in {time.monotonic() - st:.1f} seconds') return data delta = time.monotonic() - self.last_fetch_at delay = self.get_delay(url) if delta < delay: time.sleep(delay - delta) url = canonicalize_url(url) open_func = getattr(self.browser, 'open_novisit', self.browser.open) try: with closing(open_func(url, timeout=self.timeout)) as f: data = response(f.read()+f.read()) data.newurl = f.geturl() except URLError as err: if hasattr(err, 'code') and err.code in responses: raise FetchError(responses[err.code]) is_temp = getattr(err, 'worth_retry', False) reason = getattr(err, 'reason', None) if isinstance(reason, socket.gaierror): # see man gai_strerror() for details if getattr(reason, 'errno', None) in (socket.EAI_AGAIN, socket.EAI_NONAME): is_temp = True if is_temp: # Connection reset by peer or Name or service not known self.log.debug('Temporary error, retrying in 1 second') time.sleep(1) with closing(open_func(url, timeout=self.timeout)) as f: data = response(f.read()+f.read()) data.newurl = f.geturl() else: raise err finally: self.last_fetch_at = time.monotonic() self.log.debug(f'Fetched {url} in {time.monotonic() - st:f} seconds') return data def start_fetch(self, url): soup = BeautifulSoup('<a href="'+url+'" />') res = self.process_links(soup, url, 0, into_dir='') self.log.debug(url, 'saved to', res) return res def is_link_ok(self, url): for i in self.__class__.LINK_FILTER: if i.search(url): return False return True def is_link_wanted(self, url, tag): try: return self._is_link_wanted(url, tag) except NotImplementedError: pass except: return False if self.filter_regexps: for f in self.filter_regexps: if f.search(url): return False if self.match_regexps: for m in self.match_regexps: if m.search(url): return True return False return True def process_stylesheets(self, soup, baseurl): diskpath = unicode_path(os.path.join(self.current_dir, 'stylesheets')) if not os.path.exists(diskpath): os.mkdir(diskpath) for c, tag in enumerate(soup.findAll(name=['link', 'style'])): try: mtype = tag['type'] except KeyError: mtype = 'text/css' if tag.name.lower() == 'style' else '' if mtype.lower() != 'text/css': continue if tag.has_attr('href'): iurl = tag['href'] if not urlsplit(iurl).scheme: iurl = urljoin(baseurl, iurl, False) found_cached = False with self.stylemap_lock: if iurl in self.stylemap: tag['href'] = self.stylemap[iurl] found_cached = True if found_cached: continue try: data = self.fetch_url(iurl) except Exception: self.log.exception('Could not fetch stylesheet ', iurl) continue stylepath = os.path.join(diskpath, 'style'+str(c)+'.css') with self.stylemap_lock: self.stylemap[iurl] = stylepath with open(stylepath, 'wb') as x: x.write(data) tag['href'] = stylepath else: for ns in tag.findAll(text=True): src = str(ns) m = self.__class__.CSS_IMPORT_PATTERN.search(src) if m: iurl = m.group(1) if not urlsplit(iurl).scheme: iurl = urljoin(baseurl, iurl, False) found_cached = False with self.stylemap_lock: if iurl in self.stylemap: ns.replaceWith(src.replace(m.group(1), self.stylemap[iurl])) found_cached = True if found_cached: continue try: data = self.fetch_url(iurl) except Exception: self.log.exception('Could not fetch stylesheet ', iurl) continue c += 1 stylepath = os.path.join(diskpath, 'style'+str(c)+'.css') with self.stylemap_lock: self.stylemap[iurl] = stylepath with open(stylepath, 'wb') as x: x.write(data) ns.replaceWith(src.replace(m.group(1), stylepath)) def rescale_image(self, data): return rescale_image(data, self.scale_news_images, self.compress_news_images_max_size, self.compress_news_images_auto_size) def process_images(self, soup, baseurl): diskpath = unicode_path(os.path.join(self.current_dir, 'images')) if not os.path.exists(diskpath): os.mkdir(diskpath) c = 0 for tag in soup.findAll('img', src=True): iurl = tag['src'] if iurl.startswith('data:'): try: data = urlopen(iurl).read() except Exception: self.log.exception('Failed to decode embedded image') continue else: if callable(self.image_url_processor): iurl = self.image_url_processor(baseurl, iurl) if not iurl: continue if not urlsplit(iurl).scheme: iurl = urljoin(baseurl, iurl, False) found_in_cache = False with self.imagemap_lock: if iurl in self.imagemap: tag['src'] = self.imagemap[iurl] found_in_cache = True if found_in_cache: continue try: data = self.fetch_url(iurl) if data == b'GIF89a\x01': # Skip empty GIF files as PIL errors on them anyway continue except Exception: self.log.exception('Could not fetch image ', iurl) continue c += 1 fname = ascii_filename('img'+str(c)) data = self.preprocess_image_ext(data, iurl) if self.preprocess_image_ext is not None else data if data is None: continue itype = what(None, data) if itype == 'svg' or (itype is None and b'<svg' in data[:1024]): # SVG image imgpath = os.path.join(diskpath, fname+'.svg') with self.imagemap_lock: self.imagemap[iurl] = imgpath with open(imgpath, 'wb') as x: x.write(data) tag['src'] = imgpath else: from calibre.utils.img import image_from_data, image_to_data try: # Ensure image is valid img = image_from_data(data) if itype not in {'png', 'jpg', 'jpeg'}: itype = 'png' if itype == 'gif' else 'jpeg' data = image_to_data(img, fmt=itype) if self.compress_news_images and itype in {'jpg','jpeg'}: try: data = self.rescale_image(data) except Exception: self.log.exception('failed to compress image '+iurl) # Moon+ apparently cannot handle .jpeg files if itype == 'jpeg': itype = 'jpg' imgpath = os.path.join(diskpath, fname+'.'+itype) with self.imagemap_lock: self.imagemap[iurl] = imgpath with open(imgpath, 'wb') as x: x.write(data) tag['src'] = imgpath except Exception: traceback.print_exc() continue def absurl(self, baseurl, tag, key, filter=True): iurl = tag[key] parts = urlsplit(iurl) if not parts.netloc and not parts.path and not parts.query: return None if not parts.scheme: iurl = urljoin(baseurl, iurl, False) if not self.is_link_ok(iurl): self.log.debug('Skipping invalid link:', iurl) return None if filter and not self.is_link_wanted(iurl, tag): self.log.debug('Filtered link: '+iurl) return None return iurl def normurl(self, url): parts = list(urlsplit(url)) parts[4] = '' return urlunsplit(parts) def localize_link(self, tag, key, path): parts = urlsplit(tag[key]) suffix = ('#'+parts.fragment) if parts.fragment else '' tag[key] = path+suffix def process_return_links(self, soup, baseurl): for tag in soup.findAll('a', href=True): iurl = self.absurl(baseurl, tag, 'href') if not iurl: continue nurl = self.normurl(iurl) if nurl in self.filemap: self.localize_link(tag, 'href', self.filemap[nurl]) def process_links(self, soup, baseurl, recursion_level, into_dir='links'): res = '' diskpath = os.path.join(self.current_dir, into_dir) if not os.path.exists(diskpath): os.mkdir(diskpath) prev_dir = self.current_dir try: self.current_dir = diskpath tags = list(soup.findAll('a', href=True)) for c, tag in enumerate(tags): if self.show_progress: print('.', end=' ') sys.stdout.flush() sys.stdout.flush() iurl = self.absurl(baseurl, tag, 'href', filter=recursion_level != 0) if not iurl: continue nurl = self.normurl(iurl) if nurl in self.filemap: self.localize_link(tag, 'href', self.filemap[nurl]) continue if self.files > self.max_files: return res linkdir = 'link'+str(c) if into_dir else '' linkdiskpath = os.path.join(diskpath, linkdir) if not os.path.exists(linkdiskpath): os.mkdir(linkdiskpath) try: self.current_dir = linkdiskpath dsrc = self.fetch_url(iurl) newbaseurl = dsrc.newurl if len(dsrc) == 0 or \ len(re.compile(b'<!--.*?-->', re.DOTALL).sub(b'', dsrc).strip()) == 0: raise ValueError('No content at URL %r'%iurl) if callable(self.encoding): dsrc = self.encoding(dsrc) elif self.encoding is not None: dsrc = dsrc.decode(self.encoding, 'replace') else: dsrc = xml_to_unicode(dsrc, self.verbose)[0] st = time.monotonic() soup = self.get_soup(dsrc, url=iurl) self.log.debug(f'Parsed {iurl} in {time.monotonic() - st:.1f} seconds') base = soup.find('base', href=True) if base is not None: newbaseurl = base['href'] self.log.debug('Processing images...') self.process_images(soup, newbaseurl) if self.download_stylesheets: self.process_stylesheets(soup, newbaseurl) _fname = basename(iurl) if not isinstance(_fname, str): _fname.decode('latin1', 'replace') _fname = _fname.replace('%', '').replace(os.sep, '') _fname = ascii_filename(_fname) _fname = os.path.splitext(_fname)[0][:120] + '.xhtml' res = os.path.join(linkdiskpath, _fname) self.downloaded_paths.append(res) self.filemap[nurl] = res if recursion_level < self.max_recursions: self.log.debug('Processing links...') self.process_links(soup, newbaseurl, recursion_level+1) else: self.process_return_links(soup, newbaseurl) self.log.debug('Recursion limit reached. Skipping links in', iurl) if newbaseurl and not newbaseurl.startswith('/'): for atag in soup.findAll('a', href=lambda x: x and x.startswith('/')): atag['href'] = urljoin(newbaseurl, atag['href'], True) if callable(self.postprocess_html_ext): soup = self.postprocess_html_ext(soup, c==0 and recursion_level==0 and not getattr(self, 'called_first', False), self.job_info) if c==0 and recursion_level == 0: self.called_first = True save_soup(soup, res) self.localize_link(tag, 'href', res) except Exception as err: if isinstance(err, AbortArticle): raise self.failed_links.append((iurl, traceback.format_exc())) self.log.exception('Could not fetch link', iurl) finally: self.current_dir = diskpath self.files += 1 finally: self.current_dir = prev_dir if self.show_progress: print() return res def option_parser(usage=_('%prog URL\n\nWhere URL is for example https://google.com')): parser = OptionParser(usage=usage) parser.add_option('-d', '--base-dir', help=_('Base folder into which URL is saved. Default is %default'), default='.', type='string', dest='dir') parser.add_option('-t', '--timeout', help=_('Timeout in seconds to wait for a response from the server. Default: %default s'), default=10.0, type='float', dest='timeout') parser.add_option('-r', '--max-recursions', default=1, help=_('Maximum number of levels to recurse i.e. depth of links to follow. Default %default'), type='int', dest='max_recursions') parser.add_option('-n', '--max-files', default=sys.maxsize, type='int', dest='max_files', help=_('The maximum number of files to download. This only applies to files from <a href> tags. Default is %default')) parser.add_option('--delay', default=0, dest='delay', type='float', help=_('Minimum interval in seconds between consecutive fetches. Default is %default s')) parser.add_option('--encoding', default=None, help=_('The character encoding for the websites you are trying to download. The default is to try and guess the encoding.')) parser.add_option('--match-regexp', default=[], action='append', dest='match_regexps', help=_('Only links that match this regular expression will be followed. ' 'This option can be specified multiple times, in which case as long ' 'as a link matches any one regexp, it will be followed. By default all ' 'links are followed.')) parser.add_option('--filter-regexp', default=[], action='append', dest='filter_regexps', help=_('Any link that matches this regular expression will be ignored.' ' This option can be specified multiple times, in which case as' ' long as any regexp matches a link, it will be ignored. By' ' default, no links are ignored. If both filter regexp and match' ' regexp are specified, then filter regexp is applied first.')) parser.add_option('--dont-download-stylesheets', action='store_true', default=False, help=_('Do not download CSS stylesheets.'), dest='no_stylesheets') parser.add_option('--verbose', help=_('Show detailed output information. Useful for debugging'), default=False, action='store_true', dest='verbose') return parser def create_fetcher(options, image_map={}, log=None): if log is None: log = Log(level=Log.DEBUG) if options.verbose else Log() return RecursiveFetcher(options, log, image_map={}) def main(args=sys.argv): parser = option_parser() options, args = parser.parse_args(args) if len(args) != 2: parser.print_help() return 1 fetcher = create_fetcher(options) fetcher.start_fetch(args[1]) if __name__ == '__main__': sys.exit(main())
28,019
Python
.py
593
33.418212
146
0.543048
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,809
__init__.py
kovidgoyal_calibre/src/calibre/web/fetch/__init__.py
__license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>'
87
Python
.py
2
42.5
61
0.647059
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,810
templates.py
kovidgoyal_calibre/src/calibre/web/feeds/templates.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' import copy from lxml import etree, html from lxml.html.builder import BODY, BR, DIV, H2, H3, HEAD, HR, HTML, IMG, LI, SPAN, STRONG, STYLE, TABLE, TD, TITLE, TR, UL, A from lxml.html.builder import P as PT from calibre import isbytestring, strftime from calibre.utils.localization import _ def attrs(*args, **kw): rescale = kw.pop('rescale', None) if rescale is not None: kw['data-calibre-rescale'] = str(rescale) if args: kw['class'] = ' '.join(args) return kw # Regular templates class Template: IS_HTML = True def __init__(self, lang=None): self.html_lang = lang def generate(self, *args, **kwargs): if 'style' not in kwargs: kwargs['style'] = '' for key in kwargs.keys(): if isbytestring(kwargs[key]): kwargs[key] = kwargs[key].decode('utf-8', 'replace') if kwargs[key] is None: kwargs[key] = '' args = list(args) for i in range(len(args)): if isbytestring(args[i]): args[i] = args[i].decode('utf-8', 'replace') if args[i] is None: args[i] = '' self._generate(*args, **kwargs) return self def render(self, *args, **kwargs): if self.IS_HTML: return html.tostring(self.root, encoding='utf-8', include_meta_content_type=True, pretty_print=True) return etree.tostring(self.root, encoding='utf-8', xml_declaration=True, pretty_print=True) class EmbeddedContent(Template): def _generate(self, article, style=None, extra_css=None): content = article.content if article.content else '' summary = article.summary if article.summary else '' text = content if len(content) > len(summary) else summary head = HEAD(TITLE(article.title)) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) if isbytestring(text): text = text.decode('utf-8', 'replace') elements = html.fragments_fromstring(text) self.root = HTML(head, BODY(H2(article.title), DIV())) div = self.root.find('body').find('div') if elements and isinstance(elements[0], str): div.text = elements[0] elements = list(elements)[1:] for elem in elements: if hasattr(elem, 'getparent'): elem.getparent().remove(elem) else: elem = SPAN(elem) div.append(elem) class IndexTemplate(Template): def _generate(self, title, masthead, datefmt, feeds, extra_css=None, style=None): self.IS_HTML = False date = strftime(datefmt) head = HEAD(TITLE(title)) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) ul = UL(attrs('calibre_feed_list')) for i, feed in enumerate(feeds): if len(feed): li = LI(A(feed.title, attrs('feed', rescale=120, href='feed_%d/index.html'%i)), id='feed_%d'%i) ul.append(li) div = DIV( PT(IMG(src=masthead,alt="masthead"),style='text-align:center'), PT(date, style='text-align:right'), ul, attrs(rescale=100)) self.root = HTML(head, BODY(div)) if self.html_lang: self.root.set('lang', self.html_lang) class FeedTemplate(Template): def get_navbar(self, f, feeds, top=True): if len(feeds) < 2: return DIV() navbar = DIV('| ', attrs('calibre_navbar', rescale=70, style='text-align:center')) if not top: hr = HR() navbar.append(hr) navbar.text = None hr.tail = '| ' if f+1 < len(feeds): link = A(_('Next section'), href='../feed_%d/index.html'%(f+1)) link.tail = ' | ' navbar.append(link) link = A(_('Main menu'), href="../index.html") link.tail = ' | ' navbar.append(link) if f > 0: link = A(_('Previous section'), href='../feed_%d/index.html'%(f-1)) link.tail = ' |' navbar.append(link) if top: navbar.append(HR()) return navbar def _generate(self, f, feeds, cutoff, extra_css=None, style=None): from calibre.utils.cleantext import clean_xml_chars feed = feeds[f] head = HEAD(TITLE(feed.title)) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) body = BODY() body.append(self.get_navbar(f, feeds)) div = DIV( H2(feed.title, attrs('calibre_feed_title', rescale=160)), attrs(rescale=100) ) body.append(div) if getattr(feed, 'image', None): div.append(DIV(IMG( alt=feed.image_alt if feed.image_alt else '', src=feed.image_url ), attrs('calibre_feed_image'))) if getattr(feed, 'description', None): d = DIV(clean_xml_chars(feed.description), attrs('calibre_feed_description', rescale=80)) d.append(BR()) div.append(d) ul = UL(attrs('calibre_article_list')) for i, article in enumerate(feed.articles): if not getattr(article, 'downloaded', False): continue li = LI( A(article.title, attrs('article', rescale=120, href=article.url)), SPAN(article.formatted_date, attrs('article_date')), attrs(rescale=100, id='article_%d'%i, style='padding-bottom:0.5em') ) if article.summary: li.append(DIV(clean_xml_chars(cutoff(article.text_summary)), attrs('article_description', rescale=70))) ul.append(li) div.append(ul) div.append(self.get_navbar(f, feeds, top=False)) self.root = HTML(head, body) if self.html_lang: self.root.set('lang', self.html_lang) class NavBarTemplate(Template): def _generate(self, bottom, feed, art, number_of_articles_in_feed, two_levels, url, __appname__, prefix='', center=True, extra_css=None, style=None): head = HEAD(TITLE('navbar')) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) if prefix and not prefix.endswith('/'): prefix += '/' align = 'center' if center else 'left' navbar = DIV(attrs('calibre_navbar', rescale=70, style='text-align:'+align)) if bottom: if not url.startswith('file://'): navbar.append(HR()) text = 'This article was downloaded by ' p = PT(text, STRONG(__appname__), A(url, href=url, rel='calibre-downloaded-from'), style='text-align:left; max-width: 100%; overflow: hidden;') p[0].tail = ' from ' navbar.append(p) navbar.append(BR()) navbar.append(BR()) else: next_art = 'feed_%d'%(feed+1) if art == number_of_articles_in_feed - 1 \ else 'article_%d'%(art+1) up = '../..' if art == number_of_articles_in_feed - 1 else '..' href = '%s%s/%s/index.html'%(prefix, up, next_art) navbar.text = '| ' navbar.append(A(_('Next'), href=href)) href = '%s../index.html#article_%d'%(prefix, art) next(navbar.iterchildren(reversed=True)).tail = ' | ' navbar.append(A(_('Section menu'), href=href)) href = '%s../../index.html#feed_%d'%(prefix, feed) next(navbar.iterchildren(reversed=True)).tail = ' | ' navbar.append(A(_('Main menu'), href=href)) if art > 0 and not bottom: href = '%s../article_%d/index.html'%(prefix, art-1) next(navbar.iterchildren(reversed=True)).tail = ' | ' navbar.append(A(_('Previous'), href=href)) next(navbar.iterchildren(reversed=True)).tail = ' | ' if not bottom: navbar.append(HR()) self.root = HTML(head, BODY(navbar)) # Touchscreen templates class TouchscreenIndexTemplate(Template): def _generate(self, title, masthead, datefmt, feeds, extra_css=None, style=None): self.IS_HTML = False date = '{}, {} {}, {}'.format(strftime('%A'), strftime('%B'), strftime('%d').lstrip('0'), strftime('%Y')) masthead_p = etree.Element("p") masthead_p.set("style","text-align:center") masthead_img = etree.Element("img") masthead_img.set("src",masthead) masthead_img.set("alt","masthead") masthead_p.append(masthead_img) head = HEAD(TITLE(title)) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) toc = TABLE(attrs('toc'),width="100%",border="0",cellpadding="3px") for i, feed in enumerate(feeds): if len(feed): tr = TR() tr.append(TD(attrs(rescale=120), A(feed.title, href='feed_%d/index.html'%i))) tr.append(TD('%s' % len(feed.articles), style="text-align:right")) toc.append(tr) div = DIV( masthead_p, H3(attrs('publish_date'),date), DIV(attrs('divider')), toc) self.root = HTML(head, BODY(div)) if self.html_lang: self.root.set('lang', self.html_lang) class TouchscreenFeedTemplate(Template): def _generate(self, f, feeds, cutoff, extra_css=None, style=None): from calibre.utils.cleantext import clean_xml_chars def trim_title(title,clip=18): if len(title)>clip: tokens = title.split(' ') new_title_tokens = [] new_title_len = 0 if len(tokens[0]) > clip: return tokens[0][:clip] + '...' for token in tokens: if len(token) + new_title_len < clip: new_title_tokens.append(token) new_title_len += len(token) else: new_title_tokens.append('...') title = ' '.join(new_title_tokens) break return title self.IS_HTML = False feed = feeds[f] # Construct the navbar navbar_t = TABLE(attrs('touchscreen_navbar')) navbar_tr = TR() # Previous Section link = '' if f > 0: link = A(attrs('feed_link'), trim_title(feeds[f-1].title), href='../feed_%d/index.html' % int(f-1)) navbar_tr.append(TD(attrs('feed_prev'),link)) # Up to Sections link = A(_('Sections'), href="../index.html") navbar_tr.append(TD(attrs('feed_up'),link)) # Next Section link = '' if f < len(feeds)-1: link = A(attrs('feed_link'), trim_title(feeds[f+1].title), href='../feed_%d/index.html' % int(f+1)) navbar_tr.append(TD(attrs('feed_next'),link)) navbar_t.append(navbar_tr) top_navbar = navbar_t bottom_navbar = copy.copy(navbar_t) # print "\n%s\n" % etree.tostring(navbar_t, pretty_print=True) # Build the page head = HEAD(TITLE(feed.title)) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) body = BODY() div = DIV( top_navbar, H2(feed.title, attrs('feed_title')) ) body.append(div) if getattr(feed, 'image', None): div.append(DIV(IMG( alt=feed.image_alt if feed.image_alt else '', src=feed.image_url ), attrs('calibre_feed_image'))) if getattr(feed, 'description', None): d = DIV(clean_xml_chars(feed.description), attrs('calibre_feed_description', rescale=80)) d.append(BR()) div.append(d) for i, article in enumerate(feed.articles): if not getattr(article, 'downloaded', False): continue div_td = DIV(attrs('article_summary'), A(article.title, attrs('summary_headline',rescale=120, href=article.url))) if article.author: div_td.append(DIV(article.author, attrs('summary_byline', rescale=100))) if article.summary: div_td.append(DIV(cutoff(article.text_summary), attrs('summary_text', rescale=100))) div.append(div_td) div.append(bottom_navbar) self.root = HTML(head, body) if self.html_lang: self.root.set('lang', self.html_lang) class TouchscreenNavBarTemplate(Template): def _generate(self, bottom, feed, art, number_of_articles_in_feed, two_levels, url, __appname__, prefix='', center=True, extra_css=None, style=None): head = HEAD(TITLE('navbar')) if style: head.append(STYLE(style, type='text/css')) if extra_css: head.append(STYLE(extra_css, type='text/css')) navbar = DIV() navbar_t = TABLE(attrs('touchscreen_navbar')) navbar_tr = TR() if bottom and not url.startswith('file://'): navbar.append(HR()) text = 'This article was downloaded by ' p = PT(text, STRONG(__appname__), A(url, href=url, rel='calibre-downloaded-from'), style='text-align:left; max-width: 100%; overflow: hidden;') p[0].tail = ' from ' navbar.append(p) navbar.append(BR()) # | Previous if art > 0: link = A(attrs('article_link'),_('Previous'),href='%s../article_%d/index.html'%(prefix, art-1)) navbar_tr.append(TD(attrs('article_prev'),link)) else: navbar_tr.append(TD(attrs('article_prev'),'')) # | Articles | Sections | link = A(attrs('articles_link'),_('Articles'), href='%s../index.html#article_%d'%(prefix, art)) navbar_tr.append(TD(attrs('article_articles_list'),link)) link = A(attrs('sections_link'),_('Sections'), href='%s../../index.html#feed_%d'%(prefix, feed)) navbar_tr.append(TD(attrs('article_sections_list'),link)) # | Next next_art = 'feed_%d'%(feed+1) if art == number_of_articles_in_feed - 1 \ else 'article_%d'%(art+1) up = '../..' if art == number_of_articles_in_feed - 1 else '..' link = A(attrs('article_link'), _('Next'), href='%s%s/%s/index.html'%(prefix, up, next_art)) navbar_tr.append(TD(attrs('article_next'),link)) navbar_t.append(navbar_tr) navbar.append(navbar_t) # print "\n%s\n" % etree.tostring(navbar, pretty_print=True) self.root = HTML(head, BODY(navbar))
15,799
Python
.py
363
31.584022
126
0.534903
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,811
__init__.py
kovidgoyal_calibre/src/calibre/web/feeds/__init__.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' ''' Contains the logic for parsing feeds. ''' import copy import re import time import traceback from builtins import _ from calibre import force_unicode, replace_entities, strftime from calibre.utils.cleantext import clean_ascii_chars, clean_xml_chars from calibre.utils.date import dt_factory, local_tz, utcnow from calibre.utils.logging import default_log from polyglot.builtins import string_or_bytes class Article: def __init__(self, id, title, url, author, summary, published, content): from lxml import html self.downloaded = False self.id = id if not title or not isinstance(title, string_or_bytes): title = _('Unknown') title = force_unicode(title, 'utf-8') self._title = clean_xml_chars(title).strip() try: self._title = replace_entities(self._title) except Exception: pass self._title = clean_ascii_chars(self._title) self.url = url self.author = author self.toc_thumbnail = None self.internal_toc_entries = () if author and not isinstance(author, str): author = author.decode('utf-8', 'replace') if summary and not isinstance(summary, str): summary = summary.decode('utf-8', 'replace') summary = clean_xml_chars(summary) if summary else summary self.summary = summary if summary and '<' in summary: try: s = html.fragment_fromstring(summary, create_parent=True) summary = html.tostring(s, method='text', encoding='unicode') except: print('Failed to process article summary, deleting:') print(summary.encode('utf-8')) traceback.print_exc() summary = '' self.text_summary = clean_ascii_chars(summary) self.author = author self.content = content self.date = published self.utctime = dt_factory(self.date, assume_utc=True, as_utc=True) self.localtime = self.utctime.astimezone(local_tz) self._formatted_date = None @property def formatted_date(self): if self._formatted_date is None: self._formatted_date = strftime(" [%a, %d %b %H:%M]", t=self.localtime.timetuple()) return self._formatted_date @formatted_date.setter def formatted_date(self, val): if isinstance(val, str): self._formatted_date = val @property def title(self): t = self._title if not isinstance(t, str) and hasattr(t, 'decode'): t = t.decode('utf-8', 'replace') return t @title.setter def title(self, val): self._title = clean_ascii_chars(val) def __repr__(self): return \ ('''\ Title : %s URL : %s Author : %s Summary : %s Date : %s TOC thumb : %s Has content : %s '''%(self.title, self.url, self.author, self.summary[:20]+'...', self.localtime.strftime('%a, %d %b, %Y %H:%M'), self.toc_thumbnail, bool(self.content))) def __str__(self): return repr(self) def is_same_as(self, other_article): # if self.title != getattr(other_article, 'title', False): # return False if self.url: return self.url == getattr(other_article, 'url', False) return self.content == getattr(other_article, 'content', False) class Feed: def __init__(self, get_article_url=lambda item: item.get('link', None), log=default_log): ''' Parse a feed into articles. ''' self.logger = log self.get_article_url = get_article_url def populate_from_feed(self, feed, title=None, oldest_article=7, max_articles_per_feed=100): entries = feed.entries feed = feed.feed self.title = feed.get('title', _('Unknown section')) if not title else title self.description = feed.get('description', '') image = feed.get('image', {}) self.image_url = image.get('href', None) self.image_width = image.get('width', 88) self.image_height = image.get('height', 31) self.image_alt = image.get('title', '') self.articles = [] self.id_counter = 0 self.added_articles = [] self.oldest_article = oldest_article for item in entries: if len(self.articles) >= max_articles_per_feed: break self.parse_article(item) def populate_from_preparsed_feed(self, title, articles, oldest_article=7, max_articles_per_feed=100): self.title = str(title if title else _('Unknown feed')) self.description = '' self.image_url = None self.articles = [] self.added_articles = [] self.oldest_article = oldest_article self.id_counter = 0 for item in articles: if len(self.articles) >= max_articles_per_feed: break self.id_counter += 1 id = item.get('id', None) if not id: id = 'internal id#%s'%self.id_counter if id in self.added_articles: return self.added_articles.append(id) published = time.gmtime(item.get('timestamp', time.time())) title = item.get('title', _('Untitled article')) link = item.get('url', None) description = item.get('description', '') content = item.get('content', '') author = item.get('author', '') article = Article(id, title, link, author, description, published, content) delta = utcnow() - article.utctime if delta.days*24*3600 + delta.seconds <= 24*3600*self.oldest_article: self.articles.append(article) else: t = strftime('%a, %d %b, %Y %H:%M', article.localtime.timetuple()) self.logger.debug('Skipping article %s (%s) from feed %s as it is too old.'% (title, t, self.title)) d = item.get('date', '') article.formatted_date = d def parse_article(self, item): self.id_counter += 1 id = item.get('id', None) if not id: id = 'internal id#%s'%self.id_counter if id in self.added_articles: return published = None for date_field in ('date_parsed', 'published_parsed', 'updated_parsed'): published = item.get(date_field, None) if published is not None: break if not published: from dateutil.parser import parse for date_field in ('date', 'published', 'updated'): try: published = parse(item[date_field]).timetuple() except Exception: continue break if not published: published = time.gmtime() self.added_articles.append(id) title = item.get('title', _('Untitled article')) if title.startswith('<'): title = re.sub(r'<.+?>', '', title) try: link = self.get_article_url(item) except: self.logger.warning('Failed to get link for %s'%title) self.logger.debug(traceback.format_exc()) link = None description = item.get('summary', None) author = item.get('author', None) content = [i.value for i in item.get('content', []) if i.value] content = [i if isinstance(i, str) else i.decode('utf-8', 'replace') for i in content] content = '\n'.join(content) if not content.strip(): content = None if not link and not content: return article = Article(id, title, link, author, description, published, content) delta = utcnow() - article.utctime if delta.days*24*3600 + delta.seconds <= 24*3600*self.oldest_article: self.articles.append(article) else: try: self.logger.debug('Skipping article %s (%s) from feed %s as it is too old.'% (title, article.localtime.strftime('%a, %d %b, %Y %H:%M'), self.title)) except UnicodeDecodeError: if not isinstance(title, str): title = title.decode('utf-8', 'replace') self.logger.debug('Skipping article %s as it is too old'%title) def reverse(self): self.articles.reverse() def __iter__(self): return iter(self.articles) def __len__(self): return len(self.articles) def __repr__(self): res = [('%20s\n'%'').replace(' ', '_')+repr(art) for art in self] return '\n'+'\n'.join(res)+'\n' def __str__(self): return repr(self) def has_embedded_content(self): length = 0 for a in self: if a.content or a.summary: length += max(len(a.content if a.content else ''), len(a.summary if a.summary else '')) return length > 2000 * len(self) def has_article(self, article): for a in self: if a.is_same_as(article): return True return False def find(self, article): for i, a in enumerate(self): if a.is_same_as(article): return i return -1 def remove(self, article): i = self.index(article) if i > -1: self.articles[i:i+1] = [] def remove_article(self, article): try: self.articles.remove(article) except ValueError: pass class FeedCollection(list): def __init__(self, feeds): list.__init__(self, [f for f in feeds if len(f.articles) > 0]) found_articles = set() duplicates = set() def in_set(s, a): for x in s: if a.is_same_as(x): return x return None print('#feeds', len(self)) print(list(map(len, self))) for f in self: dups = [] for a in f: first = in_set(found_articles, a) if first is not None: dups.append(a) duplicates.add((first, f)) else: found_articles.add(a) for x in dups: f.articles.remove(x) self.duplicates = duplicates print(len(duplicates)) print(list(map(len, self))) # raise def find_article(self, article): for j, f in enumerate(self): for i, a in enumerate(f): if a is article: return (j, i) def restore_duplicates(self): temp = [] for article, feed in self.duplicates: art = copy.deepcopy(article) j, i = self.find_article(article) art.url = '../feed_%d/article_%d/index.html'%(j, i) temp.append((feed, art)) for feed, art in temp: feed.articles.append(art) def feed_from_xml(raw_xml, title=None, oldest_article=7, max_articles_per_feed=100, get_article_url=lambda item: item.get('link', None), log=default_log): from calibre.web.feeds.feedparser import parse # Handle unclosed escaped entities. They trip up feedparser and HBR for one # generates them raw_xml = re.sub(br'(&amp;#\d+)([^0-9;])', br'\1;\2', raw_xml) feed = parse(raw_xml) pfeed = Feed(get_article_url=get_article_url, log=log) pfeed.populate_from_feed(feed, title=title, oldest_article=oldest_article, max_articles_per_feed=max_articles_per_feed) return pfeed def feeds_from_index(index, oldest_article=7, max_articles_per_feed=100, log=default_log): ''' @param index: A parsed index as returned by L{BasicNewsRecipe.parse_index}. @return: A list of L{Feed} objects. @rtype: list ''' feeds = [] for title, articles in index: pfeed = Feed(log=log) pfeed.populate_from_preparsed_feed(title, articles, oldest_article=oldest_article, max_articles_per_feed=max_articles_per_feed) feeds.append(pfeed) return feeds
12,640
Python
.py
319
29.178683
105
0.555266
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,812
news.py
kovidgoyal_calibre/src/calibre/web/feeds/news.py
__license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' ''' Defines various abstract base classes that can be subclassed to create powerful news fetching recipes. ''' __docformat__ = "restructuredtext en" import io import os import re import sys import time import traceback from collections import defaultdict from contextlib import closing from urllib.parse import urlparse, urlsplit from calibre import __appname__, as_unicode, browser, force_unicode, iswindows, preferred_encoding, random_user_agent, strftime from calibre.ebooks.BeautifulSoup import BeautifulSoup, CData, NavigableString, Tag from calibre.ebooks.metadata import MetaInformation from calibre.ebooks.metadata.opf2 import OPFCreator from calibre.ebooks.metadata.toc import TOC from calibre.ptempfile import PersistentTemporaryFile from calibre.utils.date import now as nowf from calibre.utils.icu import numeric_sort_key from calibre.utils.img import add_borders_to_image, image_to_data, save_cover_data_to from calibre.utils.localization import _, canonicalize_lang, ngettext from calibre.utils.logging import ThreadSafeWrapper from calibre.utils.threadpool import NoResultsPending, ThreadPool, WorkRequest from calibre.web import Recipe from calibre.web.feeds import Feed, feed_from_xml, feeds_from_index, templates from calibre.web.fetch.simple import AbortArticle, RecursiveFetcher from calibre.web.fetch.simple import option_parser as web2disk_option_parser from calibre.web.fetch.utils import prepare_masthead_image from polyglot.builtins import string_or_bytes def classes(classes): q = frozenset(classes.split(' ')) return dict(attrs={ 'class': lambda x: x and frozenset(x.split()).intersection(q)}) def prefixed_classes(classes): q = frozenset(classes.split(' ')) def matcher(x): if x: for candidate in frozenset(x.split()): for x in q: if candidate.startswith(x): return True return False return {'attrs': {'class': matcher}} class LoginFailed(ValueError): pass class DownloadDenied(ValueError): pass class BasicNewsRecipe(Recipe): ''' Base class that contains logic needed in all recipes. By overriding progressively more of the functionality in this class, you can make progressively more customized/powerful recipes. For a tutorial introduction to creating recipes, see :doc:`news`. ''' #: The title to use for the e-book title = _('Unknown News Source') #: A couple of lines that describe the content this recipe downloads. #: This will be used primarily in a GUI that presents a list of recipes. description = '' #: The author of this recipe __author__ = __appname__ #: Minimum calibre version needed to use this recipe requires_version = (0, 6, 0) #: The language that the news is in. Must be an ISO-639 code either #: two or three characters long language = 'und' #: Maximum number of articles to download from each feed. This is primarily #: useful for feeds that don't have article dates. For most feeds, you should #: use :attr:`BasicNewsRecipe.oldest_article` max_articles_per_feed = 100 #: Oldest article to download from this news source. In days. oldest_article = 7.0 #: Number of levels of links to follow on article webpages recursions = 0 #: The default delay between consecutive downloads in seconds. The argument may be a #: floating point number to indicate a more precise time. See :meth:`get_url_specific_delay` #: to implement per URL delays. delay = 0 #: Publication type #: Set to newspaper, magazine or blog. If set to None, no publication type #: metadata will be written to the opf file. publication_type = 'unknown' #: Number of simultaneous downloads. Set to 1 if the server is picky. #: Automatically reduced to 1 if :attr:`BasicNewsRecipe.delay` > 0 simultaneous_downloads = 5 #: Timeout for fetching files from server in seconds timeout = 120.0 #: The format string for the date shown on the first page. #: By default: Day_Name, Day_Number Month_Name Year timefmt = ' [%a, %d %b %Y]' #: List of feeds to download. #: Can be either ``[url1, url2, ...]`` or ``[('title1', url1), ('title2', url2),...]`` feeds = None #: Max number of characters in the short description summary_length = 500 #: Convenient flag to disable loading of stylesheets for websites #: that have overly complex stylesheets unsuitable for conversion #: to e-book formats. #: If True stylesheets are not downloaded and processed no_stylesheets = False #: Convenient flag to strip all JavaScript tags from the downloaded HTML remove_javascript = True #: If True the GUI will ask the user for a username and password #: to use while downloading. #: If set to "optional" the use of a username and password becomes optional needs_subscription = False #: If True the navigation bar is center aligned, otherwise it is left aligned center_navbar = True #: Specify an override encoding for sites that have an incorrect #: charset specification. The most common being specifying ``latin1`` and #: using ``cp1252``. If None, try to detect the encoding. If it is a #: callable, the callable is called with two arguments: The recipe object #: and the source to be decoded. It must return the decoded source. encoding = None #: Normally we try to guess if a feed has full articles embedded in it #: based on the length of the embedded content. If `None`, then the #: default guessing is used. If `True` then the we always assume the feeds has #: embedded content and if `False` we always assume the feed does not have #: embedded content. use_embedded_content = None #: Set to True and implement :meth:`get_obfuscated_article` to handle #: websites that try to make it difficult to scrape content. articles_are_obfuscated = False #: Reverse the order of articles in each feed reverse_article_order = False #: Automatically extract all the text from downloaded article pages. Uses #: the algorithms from the readability project. Setting this to True, means #: that you do not have to worry about cleaning up the downloaded HTML #: manually (though manual cleanup will always be superior). auto_cleanup = False #: Specify elements that the auto cleanup algorithm should never remove. #: The syntax is a XPath expression. For example:: #: #: auto_cleanup_keep = '//div[@id="article-image"]' will keep all divs with #: id="article-image" #: auto_cleanup_keep = '//*[@class="important"]' will keep all elements #: with class="important" #: auto_cleanup_keep = '//div[@id="article-image"]|//span[@class="important"]' #: will keep all divs with id="article-image" and spans #: with class="important" #: auto_cleanup_keep = None #: Specify any extra :term:`CSS` that should be added to downloaded :term:`HTML` files. #: It will be inserted into `<style>` tags, just before the closing #: `</head>` tag thereby overriding all :term:`CSS` except that which is #: declared using the style attribute on individual :term:`HTML` tags. #: Note that if you want to programmatically generate the extra_css override #: the :meth:`get_extra_css()` method instead. #: For example:: #: #: extra_css = '.heading { font: serif x-large }' #: extra_css = None #: If True empty feeds are removed from the output. #: This option has no effect if parse_index is overridden in #: the sub class. It is meant only for recipes that return a list #: of feeds using `feeds` or :meth:`get_feeds`. It is also used if you use #: the ignore_duplicate_articles option. remove_empty_feeds = False #: List of regular expressions that determines which links to follow. #: If empty, it is ignored. Used only if is_link_wanted is #: not implemented. For example:: #: #: match_regexps = [r'page=[0-9]+'] #: #: will match all URLs that have `page=some number` in them. #: #: Only one of :attr:`BasicNewsRecipe.match_regexps` or #: :attr:`BasicNewsRecipe.filter_regexps` should be defined. match_regexps = [] #: List of regular expressions that determines which links to ignore. #: If empty it is ignored. Used only if is_link_wanted is not #: implemented. For example:: #: #: filter_regexps = [r'ads\.doubleclick\.net'] #: #: will remove all URLs that have `ads.doubleclick.net` in them. #: #: Only one of :attr:`BasicNewsRecipe.match_regexps` or #: :attr:`BasicNewsRecipe.filter_regexps` should be defined. filter_regexps = [] #: Recipe specific options to control the conversion of the downloaded #: content into an e-book. These will override any user or plugin specified #: values, so only use if absolutely necessary. For example:: #: #: conversion_options = { #: 'base_font_size' : 16, #: 'linearize_tables' : True, #: } #: conversion_options = {} #: List of tags to be removed. Specified tags are removed from downloaded HTML. #: A tag is specified as a dictionary of the form:: #: #: { #: name : 'tag name', #e.g. 'div' #: attrs : a dictionary, #e.g. {'class': 'advertisment'} #: } #: #: All keys are optional. For a full explanation of the search criteria, see #: `Beautiful Soup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/#searching-the-tree>`__ #: A common example:: #: #: remove_tags = [dict(name='div', class_='advert')] #: #: This will remove all `<div class="advert">` tags and all #: their children from the downloaded :term:`HTML`. remove_tags = [] #: Remove all tags that occur after the specified tag. #: For the format for specifying a tag see :attr:`BasicNewsRecipe.remove_tags`. #: For example:: #: #: remove_tags_after = [dict(id='content')] #: #: will remove all #: tags after the first element with `id="content"`. remove_tags_after = None #: Remove all tags that occur before the specified tag. #: For the format for specifying a tag see :attr:`BasicNewsRecipe.remove_tags`. #: For example:: #: #: remove_tags_before = dict(id='content') #: #: will remove all #: tags before the first element with `id="content"`. remove_tags_before = None #: List of attributes to remove from all tags. #: For example:: #: #: remove_attributes = ['style', 'font'] remove_attributes = [] #: Keep only the specified tags and their children. #: For the format for specifying a tag see :attr:`BasicNewsRecipe.remove_tags`. #: If this list is not empty, then the `<body>` tag will be emptied and re-filled with #: the tags that match the entries in this list. For example:: #: #: keep_only_tags = [dict(id=['content', 'heading'])] #: #: will keep only tags that have an `id` attribute of `"content"` or `"heading"`. keep_only_tags = [] #: List of :term:`regexp` substitution rules to run on the downloaded :term:`HTML`. #: Each element of the #: list should be a two element tuple. The first element of the tuple should #: be a compiled regular expression and the second a callable that takes #: a single match object and returns a string to replace the match. For example:: #: #: preprocess_regexps = [ #: (re.compile(r'<!--Article ends here-->.*</body>', re.DOTALL|re.IGNORECASE), #: lambda match: '</body>'), #: ] #: #: will remove everything from `<!--Article ends here-->` to `</body>`. preprocess_regexps = [] #: The CSS that is used to style the templates, i.e., the navigation bars and #: the Tables of Contents. Rather than overriding this variable, you should #: use `extra_css` in your recipe to customize look and feel. template_css = ''' .article_date { color: gray; font-family: monospace; } .article_description { text-indent: 0pt; } a.article { font-weight: bold; text-align:left; } a.feed { font-weight: bold; } .calibre_navbar { font-family:monospace; } ''' #: By default, calibre will use a default image for the masthead (Kindle only). #: Override this in your recipe to provide a URL to use as a masthead. masthead_url = None #: By default, the cover image returned by get_cover_url() will be used as #: the cover for the periodical. Overriding this in your recipe instructs #: calibre to render the downloaded cover into a frame whose width and height #: are expressed as a percentage of the downloaded cover. #: cover_margins = (10, 15, '#ffffff') pads the cover with a white margin #: 10px on the left and right, 15px on the top and bottom. #: Color names are defined `here <https://www.imagemagick.org/script/color.php>`_. #: Note that for some reason, white does not always work in Windows. Use #: #ffffff instead cover_margins = (0, 0, '#ffffff') #: Set to a non empty string to disable this recipe. #: The string will be used as the disabled message recipe_disabled = None #: Ignore duplicates of articles that are present in more than one section. #: A duplicate article is an article that has the same title and/or URL. #: To ignore articles with the same title, set this to:: #: #: ignore_duplicate_articles = {'title'} #: #: To use URLs instead, set it to:: #: #: ignore_duplicate_articles = {'url'} #: #: To match on title or URL, set it to:: #: #: ignore_duplicate_articles = {'title', 'url'} ignore_duplicate_articles = None # The following parameters control how the recipe attempts to minimize # JPEG image sizes #: Set this to False to ignore all scaling and compression parameters and #: pass images through unmodified. If True and the other compression #: parameters are left at their default values, JPEG images will be scaled to fit #: in the screen dimensions set by the output profile and compressed to size at #: most (w * h)/16 where w x h are the scaled image dimensions. compress_news_images = False #: The factor used when auto compressing JPEG images. If set to None, #: auto compression is disabled. Otherwise, the images will be reduced in size to #: (w * h)/compress_news_images_auto_size bytes if possible by reducing #: the quality level, where w x h are the image dimensions in pixels. #: The minimum JPEG quality will be 5/100 so it is possible this constraint #: will not be met. This parameter can be overridden by the parameter #: compress_news_images_max_size which provides a fixed maximum size for images. #: Note that if you enable scale_news_images_to_device then the image will #: first be scaled and then its quality lowered until its size is less than #: (w * h)/factor where w and h are now the *scaled* image dimensions. In #: other words, this compression happens after scaling. compress_news_images_auto_size = 16 #: Set JPEG quality so images do not exceed the size given (in KBytes). #: If set, this parameter overrides auto compression via compress_news_images_auto_size. #: The minimum JPEG quality will be 5/100 so it is possible this constraint #: will not be met. compress_news_images_max_size = None #: Rescale images to fit in the device screen dimensions set by the output profile. #: Ignored if no output profile is set. scale_news_images_to_device = True #: Maximum dimensions (w,h) to scale images to. If scale_news_images_to_device is True #: this is set to the device screen dimensions set by the output profile unless #: there is no profile set, in which case it is left at whatever value it has been #: assigned (default None). scale_news_images = None #: If set to True then links in downloaded articles that point to other downloaded articles are #: changed to point to the downloaded copy of the article rather than its original web URL. If you #: set this to True, you might also need to implement :meth:`canonicalize_internal_url` to work #: with the URL scheme of your particular website. resolve_internal_links = False #: Specify options specific to this recipe. These will be available for the user to customize #: in the Advanced tab of the Fetch News dialog or at the ebook-convert command line. The options #: are specified as a dictionary mapping option name to metadata about the option. For example:: #: #: recipe_specific_options = { #: 'edition_date': { #: 'short': 'The issue date to download', #: 'long': 'Specify a date in the format YYYY-mm-dd to download the issue corresponding to that date', #: 'default': 'current', #: } #: } #: #: When the recipe is run, self.recipe_specific_options will be a dict mapping option name to the option value #: specified by the user. When the option is unspecified by the user, it will have the value specified by 'default'. #: If no default is specified, the option will not be in the dict at all, when unspecified by the user. recipe_specific_options = None #: The simulated browser engine to use when downloading from servers. The default is to use the Python mechanize #: browser engine, which supports logging in. However, if you don't need logging in, consider changing this #: to either 'webengine' which uses an actual Chromium browser to do the network requests or 'qt' which #: uses the Qt Networking backend. Both 'webengine' and 'qt' support HTTP/2, which mechanize does not and #: are thus harder to fingerprint for bot protection services. browser_type = 'mechanize' #: Set to False if you do not want to use gzipped transfers with the mechanize browser. #: Note that some old servers flake out with gzip. handle_gzip = True # See the built-in recipes for examples of these settings. def short_title(self): return force_unicode(self.title, preferred_encoding) def is_link_wanted(self, url, tag): ''' Return True if the link should be followed or False otherwise. By default, raises NotImplementedError which causes the downloader to ignore it. :param url: The URL to be followed :param tag: The tag from which the URL was derived ''' raise NotImplementedError() def get_extra_css(self): ''' By default returns `self.extra_css`. Override if you want to programmatically generate the extra_css. ''' return self.extra_css def get_cover_url(self): ''' Return a :term:`URL` to the cover image for this issue or `None`. By default it returns the value of the member `self.cover_url` which is normally `None`. If you want your recipe to download a cover for the e-book override this method in your subclass, or set the member variable `self.cover_url` before this method is called. ''' return getattr(self, 'cover_url', None) def get_masthead_url(self): ''' Return a :term:`URL` to the masthead image for this issue or `None`. By default it returns the value of the member `self.masthead_url` which is normally `None`. If you want your recipe to download a masthead for the e-book override this method in your subclass, or set the member variable `self.masthead_url` before this method is called. Masthead images are used in Kindle MOBI files. ''' return getattr(self, 'masthead_url', None) def get_feeds(self): ''' Return a list of :term:`RSS` feeds to fetch for this profile. Each element of the list must be a 2-element tuple of the form (title, url). If title is None or an empty string, the title from the feed is used. This method is useful if your recipe needs to do some processing to figure out the list of feeds to download. If so, override in your subclass. ''' if not self.feeds: raise NotImplementedError() if self.test: return self.feeds[:self.test[0]] return self.feeds def get_url_specific_delay(self, url): ''' Return the delay in seconds before downloading this URL. If you want to programmatically determine the delay for the specified URL, override this method in your subclass, returning self.delay by default for URLs you do not want to affect. :return: A floating point number, the delay in seconds. ''' return self.delay @classmethod def print_version(cls, url): ''' Take a `url` pointing to the webpage with article content and return the :term:`URL` pointing to the print version of the article. By default does nothing. For example:: def print_version(self, url): return url + '?&pagewanted=print' ''' raise NotImplementedError() @classmethod def image_url_processor(cls, baseurl, url): ''' Perform some processing on image urls (perhaps removing size restrictions for dynamically generated images, etc.) and return the processed URL. Return None or an empty string to skip fetching the image. ''' return url def preprocess_image(self, img_data, image_url): ''' Perform some processing on downloaded image data. This is called on the raw data before any resizing is done. Must return the processed raw data. Return None to skip the image. ''' return img_data def get_browser(self, *args, **kwargs): ''' Return a browser instance used to fetch documents from the web. By default it returns a `mechanize <https://mechanize.readthedocs.io/en/latest/>`_ browser instance that supports cookies, ignores robots.txt, handles refreshes and has a random common user agent. To customize the browser override this method in your sub-class as:: def get_browser(self, *a, **kw): br = super().get_browser(*a, **kw) # Add some headers br.addheaders += [ ('My-Header', 'one'), ('My-Header2', 'two'), ] # Set some cookies br.set_cookie('name', 'value') br.set_cookie('name2', 'value2', domain='.mydomain.com') # Make a POST request with some data br.open('https://someurl.com', {'username': 'def', 'password': 'pwd'}).read() # Do a login via a simple web form (only supported with mechanize browsers) if self.username is not None and self.password is not None: br.open('https://www.nytimes.com/auth/login') br.select_form(name='login') br['USERID'] = self.username br['PASSWORD'] = self.password br.submit() return br ''' if 'user_agent' not in kwargs: # More and more news sites are serving JPEG XR images to IE ua = getattr(self, 'last_used_user_agent', None) or self.calibre_most_common_ua or random_user_agent(allow_ie=False) kwargs['user_agent'] = self.last_used_user_agent = ua self.log('Using user agent:', kwargs['user_agent']) if self.browser_type != 'mechanize': from calibre.scraper.qt import Browser, WebEngineBrowser return {'qt': Browser, 'webengine': WebEngineBrowser}[self.browser_type]( user_agent=kwargs['user_agent'], verify_ssl_certificates=kwargs.get('verify_ssl_certificates', False)) br = browser(*args, **kwargs) br.addheaders += [('Accept', '*/*')] if self.handle_gzip: br.set_handle_gzip(True) return br def clone_browser(self, br): ''' Clone the browser br. Cloned browsers are used for multi-threaded downloads, since mechanize is not thread safe. The default cloning routines should capture most browser customization, but if you do something exotic in your recipe, you should override this method in your recipe and clone manually. Cloned browser instances use the same, thread-safe CookieJar by default, unless you have customized cookie handling. ''' if callable(getattr(br, 'clone_browser', None)): return br.clone_browser() # Uh-oh recipe using something exotic, call get_browser return self.get_browser() @property def cloned_browser(self): if hasattr(self.get_browser, 'is_base_class_implementation') and self.browser_type == 'mechanize': # We are using the default get_browser, which means no need to # clone br = BasicNewsRecipe.get_browser(self) else: br = self.clone_browser(self.browser) return br def get_article_url(self, article): ''' Override in a subclass to customize extraction of the :term:`URL` that points to the content for each article. Return the article URL. It is called with `article`, an object representing a parsed article from a feed. See `feedparser <https://pythonhosted.org/feedparser/>`_. By default it looks for the original link (for feeds syndicated via a service like FeedBurner or Pheedo) and if found, returns that or else returns `article.link <https://pythonhosted.org/feedparser/reference-entry-link.html>`_. ''' for key in article.keys(): if key.endswith('_origlink'): url = article[key] if url and (url.startswith('http://') or url.startswith('https://')): return url ans = article.get('link', None) if not ans and getattr(article, 'links', None): for item in article.links: if item.get('rel', 'alternate') == 'alternate': ans = item['href'] break return ans def skip_ad_pages(self, soup): ''' This method is called with the source of each downloaded :term:`HTML` file, before any of the cleanup attributes like remove_tags, keep_only_tags are applied. Note that preprocess_regexps will have already been applied. It is meant to allow the recipe to skip ad pages. If the soup represents an ad page, return the HTML of the real page. Otherwise return None. `soup`: A `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/>`__ instance containing the downloaded :term:`HTML`. ''' return None def abort_article(self, msg=None): ''' Call this method inside any of the preprocess methods to abort the download for the current article. Useful to skip articles that contain inappropriate content, such as pure video articles. ''' raise AbortArticle(msg or _('Article download aborted')) def preprocess_raw_html(self, raw_html, url): ''' This method is called with the source of each downloaded :term:`HTML` file, before it is parsed into an object tree. raw_html is a unicode string representing the raw HTML downloaded from the web. url is the URL from which the HTML was downloaded. Note that this method acts *before* preprocess_regexps. This method must return the processed raw_html as a unicode object. ''' return raw_html def preprocess_raw_html_(self, raw_html, url): raw_html = self.preprocess_raw_html(raw_html, url) if self.auto_cleanup: try: raw_html = self.extract_readable_article(raw_html, url) except: self.log.exception('Auto cleanup of URL: %r failed'%url) return raw_html def preprocess_html(self, soup): ''' This method is called with the source of each downloaded :term:`HTML` file, before it is parsed for links and images. It is called after the cleanup as specified by remove_tags etc. It can be used to do arbitrarily powerful pre-processing on the :term:`HTML`. It should return `soup` after processing it. `soup`: A `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/>`__ instance containing the downloaded :term:`HTML`. ''' return soup def postprocess_html(self, soup, first_fetch): ''' This method is called with the source of each downloaded :term:`HTML` file, after it is parsed for links and images. It can be used to do arbitrarily powerful post-processing on the :term:`HTML`. It should return `soup` after processing it. :param soup: A `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/>`__ instance containing the downloaded :term:`HTML`. :param first_fetch: True if this is the first page of an article. ''' return soup def cleanup(self): ''' Called after all articles have been download. Use it to do any cleanup like logging out of subscription sites, etc. ''' pass def canonicalize_internal_url(self, url, is_link=True): ''' Return a set of canonical representations of ``url``. The default implementation uses just the server hostname and path of the URL, ignoring any query parameters, fragments, etc. The canonical representations must be unique across all URLs for this news source. If they are not, then internal links may be resolved incorrectly. :param is_link: Is True if the URL is coming from an internal link in an HTML file. False if the URL is the URL used to download an article. ''' try: parts = urlparse(url) except Exception: self.log.error('Failed to parse url: %r, ignoring' % url) return frozenset() nl = parts.netloc path = parts.path or '' if isinstance(nl, bytes): nl = nl.decode('utf-8', 'replace') if isinstance(path, bytes): path = path.decode('utf-8', 'replace') return frozenset({(nl, path.rstrip('/'))}) def index_to_soup(self, url_or_raw, raw=False, as_tree=False, save_raw=None): ''' Convenience method that takes an URL to the index page and returns a `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc>`__ of it. `url_or_raw`: Either a URL or the downloaded index page as a string ''' if re.match((br'\w+://' if isinstance(url_or_raw, bytes) else r'\w+://'), url_or_raw): # We may be called in a thread (in the skip_ad_pages method), so # clone the browser to be safe. We cannot use self.cloned_browser # as it may or may not actually clone the browser, depending on if # the recipe implements get_browser() or not br = self.clone_browser(self.browser) open_func = getattr(br, 'open_novisit', br.open) with closing(open_func(url_or_raw, timeout=self.timeout)) as f: _raw = f.read() if not _raw: raise RuntimeError('Could not fetch index from %s'%url_or_raw) else: _raw = url_or_raw if raw: return _raw if not isinstance(_raw, str) and self.encoding: if callable(self.encoding): _raw = self.encoding(_raw) else: _raw = _raw.decode(self.encoding, 'replace') from calibre.ebooks.chardet import strip_encoding_declarations, xml_to_unicode from calibre.utils.cleantext import clean_xml_chars if isinstance(_raw, str): _raw = strip_encoding_declarations(_raw) else: _raw = xml_to_unicode(_raw, strip_encoding_pats=True, resolve_entities=True)[0] _raw = clean_xml_chars(_raw) if save_raw: with open(save_raw, 'wb') as f: f.write(_raw.encode('utf-8')) if as_tree: from html5_parser import parse return parse(_raw) return BeautifulSoup(_raw) def extract_readable_article(self, html, url): ''' Extracts main article content from 'html', cleans up and returns as a (article_html, extracted_title) tuple. Based on the original readability algorithm by Arc90. ''' from lxml.html import document_fromstring, fragment_fromstring, tostring from calibre.ebooks.readability import readability doc = readability.Document(html, self.log, url=url, keep_elements=self.auto_cleanup_keep) article_html = doc.summary() extracted_title = doc.title() try: frag = fragment_fromstring(article_html) except: doc = document_fromstring(article_html) frag = doc.xpath('//body')[-1] if frag.tag == 'html': root = frag elif frag.tag == 'body': root = document_fromstring( '<html><head><title>%s</title></head></html>' % extracted_title) root.append(frag) else: root = document_fromstring( '<html><head><title>%s</title></head><body/></html>' % extracted_title) root.xpath('//body')[0].append(frag) body = root.xpath('//body')[0] has_title = False for x in body.iterdescendants(): if x.text == extracted_title: has_title = True inline_titles = body.xpath('//h1|//h2') if not has_title and not inline_titles: heading = body.makeelement('h2') heading.text = extracted_title body.insert(0, heading) raw_html = tostring(root, encoding='unicode') return raw_html def sort_index_by(self, index, weights): ''' Convenience method to sort the titles in `index` according to `weights`. `index` is sorted in place. Returns `index`. `index`: A list of titles. `weights`: A dictionary that maps weights to titles. If any titles in index are not in weights, they are assumed to have a weight of 0. ''' weights = defaultdict(int, weights) index.sort(key=lambda x: weights[x]) return index def parse_index(self): ''' This method should be implemented in recipes that parse a website instead of feeds to generate a list of articles. Typical uses are for news sources that have a "Print Edition" webpage that lists all the articles in the current print edition. If this function is implemented, it will be used in preference to :meth:`BasicNewsRecipe.parse_feeds`. It must return a list. Each element of the list must be a 2-element tuple of the form ``('feed title', list of articles)``. Each list of articles must contain dictionaries of the form:: { 'title' : article title, 'url' : URL of print version, 'date' : The publication date of the article as a string, 'description' : A summary of the article 'content' : The full article (can be an empty string). Obsolete do not use, instead save the content to a temporary file and pass a file:///path/to/temp/file.html as the URL. } For an example, see the recipe for downloading `The Atlantic`. In addition, you can add 'author' for the author of the article. If you want to abort processing for some reason and have calibre show the user a simple message instead of an error, call :meth:`abort_recipe_processing`. ''' raise NotImplementedError() def abort_recipe_processing(self, msg): ''' Causes the recipe download system to abort the download of this recipe, displaying a simple feedback message to the user. ''' from calibre.ebooks.conversion import ConversionUserFeedBack raise ConversionUserFeedBack(_('Failed to download %s')%self.title, msg) def get_obfuscated_article(self, url): ''' If you set `articles_are_obfuscated` this method is called with every article URL. It should return the path to a file on the filesystem that contains the article HTML. That file is processed by the recursive HTML fetching engine, so it can contain links to pages/images on the web. Alternately, you can return a dictionary of the form: {'data': <HTML data>, 'url': <the resolved URL of the article>}. This avoids needing to create temporary files. The `url` key in the dictionary is useful if the effective URL of the article is different from the URL passed into this method, for example, because of redirects. It can be omitted if the URL is unchanged. This method is typically useful for sites that try to make it difficult to access article content automatically. ''' raise NotImplementedError() def add_toc_thumbnail(self, article, src): ''' Call this from populate_article_metadata with the src attribute of an <img> tag from the article that is appropriate for use as the thumbnail representing the article in the Table of Contents. Whether the thumbnail is actually used is device dependent (currently only used by the Kindles). Note that the referenced image must be one that was successfully downloaded, otherwise it will be ignored. ''' if not src or not hasattr(article, 'toc_thumbnail'): return src = src.replace('\\', '/') if re.search(r'feed_\d+/article_\d+/images/img', src, flags=re.I) is None: self.log.warn('Ignoring invalid TOC thumbnail image: %r'%src) return article.toc_thumbnail = re.sub(r'^.*?feed', 'feed', src, flags=re.IGNORECASE) def populate_article_metadata(self, article, soup, first): ''' Called when each HTML page belonging to article is downloaded. Intended to be used to get article metadata like author/summary/etc. from the parsed HTML (soup). :param article: A object of class :class:`calibre.web.feeds.Article`. If you change the summary, remember to also change the text_summary :param soup: Parsed HTML belonging to this article :param first: True iff the parsed HTML is the first page of the article. ''' pass def postprocess_book(self, oeb, opts, log): ''' Run any needed post processing on the parsed downloaded e-book. :param oeb: An OEBBook object :param opts: Conversion options ''' pass def __init__(self, options, log, progress_reporter): ''' Initialize the recipe. :param options: Parsed commandline options :param log: Logging object :param progress_reporter: A Callable that takes two arguments: progress (a number between 0 and 1) and a string message. The message should be optional. ''' self.log = ThreadSafeWrapper(log) if not isinstance(self.title, str): self.title = str(self.title, 'utf-8', 'replace') self.debug = options.verbose > 1 self.output_dir = os.path.abspath(os.getcwd()) self.verbose = options.verbose self.test = options.test if self.test and not isinstance(self.test, tuple): self.test = (2, 2) self.username = options.username self.password = options.password self.lrf = options.lrf self.output_profile = options.output_profile self.touchscreen = getattr(self.output_profile, 'touchscreen', False) if self.touchscreen: self.template_css += self.output_profile.touchscreen_news_css if self.test: self.max_articles_per_feed = self.test[1] self.simultaneous_downloads = min(4, self.simultaneous_downloads) if self.debug: self.verbose = True self.report_progress = progress_reporter if self.needs_subscription and ( self.username is None or self.password is None or ( not self.username and not self.password)): if self.needs_subscription != 'optional': raise ValueError(_('The "%s" recipe needs a username and password.')%self.title) self.browser = self.get_browser() self.image_map, self.image_counter = {}, 1 self.css_map = {} web2disk_cmdline = ['web2disk', '--timeout', str(self.timeout), '--max-recursions', str(self.recursions), '--delay', str(self.delay), ] if self.verbose: web2disk_cmdline.append('--verbose') if self.no_stylesheets: web2disk_cmdline.append('--dont-download-stylesheets') for reg in self.match_regexps: web2disk_cmdline.extend(['--match-regexp', reg]) for reg in self.filter_regexps: web2disk_cmdline.extend(['--filter-regexp', reg]) if options.output_profile.short_name in ('default', 'tablet'): self.scale_news_images_to_device = False elif self.scale_news_images_to_device: self.scale_news_images = options.output_profile.screen_size self.web2disk_options = web2disk_option_parser().parse_args(web2disk_cmdline)[0] for extra in ('keep_only_tags', 'remove_tags', 'preprocess_regexps', 'skip_ad_pages', 'preprocess_html', 'remove_tags_after', 'remove_tags_before', 'is_link_wanted', 'compress_news_images', 'compress_news_images_max_size', 'compress_news_images_auto_size', 'scale_news_images'): setattr(self.web2disk_options, extra, getattr(self, extra)) self.web2disk_options.postprocess_html = self._postprocess_html self.web2disk_options.preprocess_image = self.preprocess_image self.web2disk_options.encoding = self.encoding self.web2disk_options.preprocess_raw_html = self.preprocess_raw_html_ self.web2disk_options.get_delay = self.get_url_specific_delay if self.delay > 0: self.simultaneous_downloads = 1 self.navbar = templates.TouchscreenNavBarTemplate() if self.touchscreen else \ templates.NavBarTemplate() self.failed_downloads = [] self.partial_failures = [] self.aborted_articles = [] self.recipe_specific_options_metadata = rso = self.recipe_specific_options or {} self.recipe_specific_options = {k: rso[k]['default'] for k in rso if 'default' in rso[k]} for x in (options.recipe_specific_option or ()): k, sep, v = x.partition(':') if not sep: raise ValueError(f'{x} is not a valid recipe specific option') if k not in rso: raise KeyError(f'{k} is not an option supported by: {self.title}') self.recipe_specific_options[k] = v if self.recipe_specific_options: log('Recipe specific options:') for k, v in self.recipe_specific_options.items(): log(' ', f'{k} = {v}') def _postprocess_html(self, soup, first_fetch, job_info): if self.no_stylesheets: for link in soup.findAll('link'): if (link.get('type') or 'text/css').lower() == 'text/css' and 'stylesheet' in (link.get('rel') or ('stylesheet',)): link.extract() for style in soup.findAll('style'): style.extract() head = soup.find('head') if not head: head = soup.find('body') if not head: head = soup.find(True) css = self.template_css + '\n\n' + (self.get_extra_css() or '') style = soup.new_tag('style', type='text/css', title='override_css') style.append(css) head.append(style) if first_fetch and job_info: url, f, a, feed_len = job_info body = soup.find('body') if body is not None: templ = self.navbar.generate(False, f, a, feed_len, not self.has_single_feed, url, __appname__, center=self.center_navbar, extra_css=self.get_extra_css() or '') elem = BeautifulSoup(templ.render(doctype='xhtml').decode('utf-8')).find('div') body.insert(0, elem) # This is needed because otherwise inserting elements into # the soup breaks find() soup = BeautifulSoup(soup.decode_contents()) if self.remove_javascript: for script in list(soup.findAll('script')): script.extract() for o in soup.findAll(onload=True): del o['onload'] for attr in self.remove_attributes: for x in soup.findAll(attrs={attr:True}): del x[attr] for bad_tag in list(soup.findAll(['base', 'iframe', 'canvas', 'embed', 'button', 'command', 'datalist', 'video', 'audio', 'noscript', 'link', 'meta'])): # link tags can be used for preloading causing network activity in # calibre viewer. meta tags can do all sorts of crazy things, # including http-equiv refresh, viewport shenanigans, etc. bad_tag.extract() # srcset causes some viewers, like calibre's to load images from the # web, and it also possible causes iBooks on iOS to barf, see # https://bugs.launchpad.net/bugs/1713986 for img in soup.findAll('img', srcset=True): del img['srcset'] ans = self.postprocess_html(soup, first_fetch) # Nuke HTML5 tags for x in ans.findAll(['article', 'aside', 'header', 'footer', 'nav', 'figcaption', 'figure', 'section']): x.get_attribute_list('class').append(f'calibre-nuked-tag-{x.name}') x.name = 'div' if job_info: url, f, a, feed_len = job_info try: article = self.feed_objects[f].articles[a] except: self.log.exception('Failed to get article object for postprocessing') pass else: self.populate_article_metadata(article, ans, first_fetch) return ans def download(self): ''' Download and pre-process all articles from the feeds in this recipe. This method should be called only once on a particular Recipe instance. Calling it more than once will lead to undefined behavior. :return: Path to index.html ''' try: res = self.build_index() self.report_progress(1, _('Download finished')) if self.failed_downloads: self.log.warning(_('Failed to download the following articles:')) for feed, article, debug in self.failed_downloads: self.log.warning(article.title, 'from', feed.title) self.log.debug(article.url) self.log.debug(debug) if self.partial_failures: self.log.warning(_('Failed to download parts of the following articles:')) for feed, atitle, aurl, debug in self.partial_failures: self.log.warning(atitle + _(' from ') + feed) self.log.debug(aurl) self.log.warning(_('\tFailed links:')) for l, tb in debug: self.log.warning(l) self.log.debug(tb) return res finally: self.cleanup() @property def lang_for_html(self): try: lang = self.language.replace('_', '-').partition('-')[0].lower() if lang == 'und': lang = None except: lang = None return lang def feeds2index(self, feeds): templ = (templates.TouchscreenIndexTemplate if self.touchscreen else templates.IndexTemplate) templ = templ(lang=self.lang_for_html) css = self.template_css + '\n\n' +(self.get_extra_css() or '') timefmt = self.timefmt return templ.generate(self.title, "mastheadImage.jpg", timefmt, feeds, extra_css=css).render(doctype='xhtml') @classmethod def description_limiter(cls, src): if not src: return '' src = force_unicode(src, 'utf-8') pos = cls.summary_length fuzz = 50 si = src.find(';', pos) if si > 0 and si-pos > fuzz: si = -1 gi = src.find('>', pos) if gi > 0 and gi-pos > fuzz: gi = -1 npos = max(si, gi) if npos < 0: npos = pos ans = src[:npos+1] if len(ans) < len(src): from calibre.utils.cleantext import clean_xml_chars # Truncating the string could cause a dangling UTF-16 half-surrogate, which will cause lxml to barf, clean it ans = clean_xml_chars(ans) + '\u2026' return ans def feed2index(self, f, feeds): feed = feeds[f] if feed.image_url is not None: # Download feed image imgdir = os.path.join(self.output_dir, 'images') if not os.path.isdir(imgdir): os.makedirs(imgdir) if feed.image_url in self.image_map: feed.image_url = self.image_map[feed.image_url] else: bn = urlsplit(feed.image_url).path if bn: bn = bn.rpartition('/')[-1] if bn: img = os.path.join(imgdir, 'feed_image_%d%s'%(self.image_counter, os.path.splitext(bn)[-1])) try: with open(img, 'wb') as fi, closing(self.browser.open(feed.image_url, timeout=self.timeout)) as r: fi.write(r.read()) self.image_counter += 1 feed.image_url = img self.image_map[feed.image_url] = img except: pass if isinstance(feed.image_url, bytes): feed.image_url = feed.image_url.decode(sys.getfilesystemencoding(), 'strict') templ = (templates.TouchscreenFeedTemplate if self.touchscreen else templates.FeedTemplate) templ = templ(lang=self.lang_for_html) css = self.template_css + '\n\n' +(self.get_extra_css() or '') return templ.generate(f, feeds, self.description_limiter, extra_css=css).render(doctype='xhtml') def _fetch_article(self, url, dir_, f, a, num_of_feeds, preloaded=None): br = self.browser if hasattr(self.get_browser, 'is_base_class_implementation'): # We are using the default get_browser, which means no need to # clone br = BasicNewsRecipe.get_browser(self) else: br = self.clone_browser(self.browser) self.web2disk_options.browser = br fetcher = RecursiveFetcher(self.web2disk_options, self.log, self.image_map, self.css_map, (url, f, a, num_of_feeds)) fetcher.browser = br fetcher.base_dir = dir_ fetcher.current_dir = dir_ fetcher.show_progress = False fetcher.image_url_processor = self.image_url_processor if preloaded is not None: fetcher.preloaded_urls[url] = preloaded res, path, failures = fetcher.start_fetch(url), fetcher.downloaded_paths, fetcher.failed_links if not res or not os.path.exists(res): msg = _('Could not fetch article.') + ' ' if self.debug: msg += _('The debug traceback is available earlier in this log') else: msg += _('Run with -vv to see the reason') raise Exception(msg) return res, path, failures def fetch_article(self, url, dir, f, a, num_of_feeds): return self._fetch_article(url, dir, f, a, num_of_feeds) def fetch_obfuscated_article(self, url, dir, f, a, num_of_feeds): x = self.get_obfuscated_article(url) if isinstance(x, dict): data = x['data'] if isinstance(data, str): data = data.encode(self.encoding or 'utf-8') url = x.get('url', url) else: with open(x, 'rb') as of: data = of.read() os.remove(x) return self._fetch_article(url, dir, f, a, num_of_feeds, preloaded=data) def fetch_embedded_article(self, article, dir, f, a, num_of_feeds): templ = templates.EmbeddedContent() raw = templ.generate(article).render('html') with PersistentTemporaryFile('_feeds2disk.html') as pt: pt.write(raw) url = ('file:'+pt.name) if iswindows else ('file://'+pt.name) return self._fetch_article(url, dir, f, a, num_of_feeds) def remove_duplicate_articles(self, feeds): seen_keys = defaultdict(set) remove = [] for f in feeds: for article in f: for key in self.ignore_duplicate_articles: val = getattr(article, key) seen = seen_keys[key] if val: if val in seen: remove.append((f, article)) else: seen.add(val) for feed, article in remove: self.log.debug('Removing duplicate article: %s from section: %s'%( article.title, feed.title)) feed.remove_article(article) if self.remove_empty_feeds: feeds = [f for f in feeds if len(f) > 0] return feeds def build_index(self): self.report_progress(0, _('Fetching feeds...')) feeds = None try: feeds = feeds_from_index(self.parse_index(), oldest_article=self.oldest_article, max_articles_per_feed=self.max_articles_per_feed, log=self.log) self.report_progress(0, _('Got feeds from index page')) except NotImplementedError: pass if feeds is None: feeds = self.parse_feeds() if not feeds: raise ValueError('No articles found, aborting') if self.ignore_duplicate_articles is not None: feeds = self.remove_duplicate_articles(feeds) self.report_progress(0, _('Trying to download cover...')) self.download_cover() self.report_progress(0, _('Generating masthead...')) self.resolve_masthead() if self.test: feeds = feeds[:self.test[0]] self.has_single_feed = len(feeds) == 1 index = os.path.join(self.output_dir, 'index.html') html = self.feeds2index(feeds) with open(index, 'wb') as fi: fi.write(html) self.jobs = [] if self.reverse_article_order: for feed in feeds: if hasattr(feed, 'reverse'): feed.reverse() self.feed_objects = feeds for f, feed in enumerate(feeds): feed_dir = os.path.join(self.output_dir, 'feed_%d'%f) if not os.path.isdir(feed_dir): os.makedirs(feed_dir) for a, article in enumerate(feed): if a >= self.max_articles_per_feed: break art_dir = os.path.join(feed_dir, 'article_%d'%a) if not os.path.isdir(art_dir): os.makedirs(art_dir) try: url = self.print_version(article.url) except NotImplementedError: url = article.url except: self.log.exception('Failed to find print version for: '+article.url) url = None if not url: continue func, arg = (self.fetch_embedded_article, article) \ if self.use_embedded_content or (self.use_embedded_content is None and feed.has_embedded_content()) \ else \ ((self.fetch_obfuscated_article if self.articles_are_obfuscated else self.fetch_article), url) req = WorkRequest(func, (arg, art_dir, f, a, len(feed)), {}, (f, a), self.article_downloaded, self.error_in_article_download) req.feed = feed req.article = article req.feed_dir = feed_dir self.jobs.append(req) self.jobs_done = 0 tp = ThreadPool(self.simultaneous_downloads) for req in self.jobs: tp.putRequest(req, block=True, timeout=0) self.report_progress(0, ngettext( 'Starting download in a single thread...', 'Starting download [{} threads]...', self.simultaneous_downloads).format(self.simultaneous_downloads)) while True: try: tp.poll() time.sleep(0.1) except NoResultsPending: break for f, feed in enumerate(feeds): html = self.feed2index(f,feeds) feed_dir = os.path.join(self.output_dir, 'feed_%d'%f) with open(os.path.join(feed_dir, 'index.html'), 'wb') as fi: fi.write(html) self.create_opf(feeds) self.report_progress(1, _('Feeds downloaded to %s')%index) return index def _download_cover(self): self.cover_path = None try: cu = self.get_cover_url() except Exception as err: self.log.error(_('Could not download cover: %s')%as_unicode(err)) self.log.debug(traceback.format_exc()) else: if not cu: return cdata = None if hasattr(cu, 'read'): cdata = cu.read() cu = getattr(cu, 'name', 'cover.jpg') elif os.access(cu, os.R_OK): with open(cu, 'rb') as f: cdata = f.read() else: self.report_progress(1, _('Downloading cover from %s')%cu) with closing(self.browser.open(cu, timeout=self.timeout)) as r: cdata = r.read() if not cdata: return ext = cu.split('/')[-1].rpartition('.')[-1].lower().strip() if ext == 'pdf': from calibre.ebooks.metadata.pdf import get_metadata stream = io.BytesIO(cdata) cdata = None mi = get_metadata(stream) if mi.cover_data and mi.cover_data[1]: cdata = mi.cover_data[1] if not cdata: return if self.cover_margins[0] or self.cover_margins[1]: cdata = image_to_data(add_borders_to_image(cdata, left=self.cover_margins[0],right=self.cover_margins[0], top=self.cover_margins[1],bottom=self.cover_margins[1], border_color=self.cover_margins[2])) cpath = os.path.join(self.output_dir, 'cover.jpg') save_cover_data_to(cdata, cpath) self.cover_path = cpath def download_cover(self): self.cover_path = None try: self._download_cover() except: self.log.exception('Failed to download cover') self.cover_path = None def _download_masthead(self, mu): if hasattr(mu, 'rpartition'): ext = mu.rpartition('.')[-1] if '?' in ext: ext = '' else: ext = mu.name.rpartition('.')[-1] ext = ext.lower() if ext else 'jpg' mpath = os.path.join(self.output_dir, 'masthead_source.'+ext) outfile = os.path.join(self.output_dir, 'mastheadImage.jpg') if hasattr(mu, 'read'): with open(mpath, 'wb') as mfile: mfile.write(mu.read()) elif os.access(mu, os.R_OK): with open(mpath, 'wb') as mfile: mfile.write(open(mu, 'rb').read()) else: with open(mpath, 'wb') as mfile, closing(self.browser.open(mu, timeout=self.timeout)) as r: mfile.write(r.read()) self.report_progress(1, _('Masthead image downloaded')) self.prepare_masthead_image(mpath, outfile) self.masthead_path = outfile if os.path.exists(mpath): os.remove(mpath) def download_masthead(self, url): try: self._download_masthead(url) except: self.log.exception("Failed to download supplied masthead_url") def resolve_masthead(self): self.masthead_path = None try: murl = self.get_masthead_url() except: self.log.exception('Failed to get masthead url') murl = None if murl is not None: # Try downloading the user-supplied masthead_url # Failure sets self.masthead_path to None self.download_masthead(murl) if self.masthead_path is None: self.log.info("Synthesizing mastheadImage") self.masthead_path = os.path.join(self.output_dir, 'mastheadImage.jpg') try: self.default_masthead_image(self.masthead_path) except: self.log.exception('Failed to generate default masthead image') self.masthead_path = None def default_cover(self, cover_file): ''' Create a generic cover for recipes that don't have a cover ''' try: from calibre.ebooks.covers import create_cover title = self.title if isinstance(self.title, str) else \ self.title.decode(preferred_encoding, 'replace') date = strftime(self.timefmt).replace('[', '').replace(']', '') img_data = create_cover(title, [date]) cover_file.write(img_data) cover_file.flush() except: self.log.exception('Failed to generate default cover') return False return True def get_masthead_title(self): 'Override in subclass to use something other than the recipe title' return self.title MI_WIDTH = 600 MI_HEIGHT = 60 def default_masthead_image(self, out_path): from calibre.ebooks import generate_masthead generate_masthead(self.get_masthead_title(), output_path=out_path, width=self.MI_WIDTH, height=self.MI_HEIGHT) def prepare_masthead_image(self, path_to_image, out_path): prepare_masthead_image(path_to_image, out_path, self.MI_WIDTH, self.MI_HEIGHT) def publication_date(self): ''' Use this method to set the date when this issue was published. Defaults to the moment of download. Must return a :class:`datetime.datetime` object. ''' return nowf() def create_opf(self, feeds, dir=None): if dir is None: dir = self.output_dir title = self.short_title() pdate = self.publication_date() if self.output_profile.periodical_date_in_title: title += strftime(self.timefmt, pdate) mi = MetaInformation(title, [__appname__]) mi.publisher = __appname__ mi.author_sort = __appname__ if self.publication_type: mi.publication_type = 'periodical:'+self.publication_type+':'+self.short_title() mi.timestamp = nowf() article_titles, aseen = [], set() for (af, aa) in self.aborted_articles: aseen.add(aa.title) for (ff, fa, tb) in self.failed_downloads: aseen.add(fa.title) for f in feeds: for a in f: if a.title and a.title not in aseen: aseen.add(a.title) article_titles.append(force_unicode(a.title, 'utf-8')) desc = self.description if not isinstance(desc, str): desc = desc.decode('utf-8', 'replace') mi.comments = (_('Articles in this issue:' ) + '\n\n' + '\n\n'.join(article_titles)) + '\n\n' + desc language = canonicalize_lang(self.language) if language is not None: mi.language = language mi.pubdate = pdate opf_path = os.path.join(dir, 'index.opf') ncx_path = os.path.join(dir, 'index.ncx') opf = OPFCreator(dir, mi) # Add mastheadImage entry to <guide> section mp = getattr(self, 'masthead_path', None) if mp is not None and os.access(mp, os.R_OK): from calibre.ebooks.metadata.opf2 import Guide ref = Guide.Reference(os.path.basename(self.masthead_path), os.getcwd()) ref.type = 'masthead' ref.title = 'Masthead Image' opf.guide.append(ref) manifest = [os.path.join(dir, 'feed_%d'%i) for i in range(len(feeds))] manifest.append(os.path.join(dir, 'index.html')) manifest.append(os.path.join(dir, 'index.ncx')) # Get cover cpath = getattr(self, 'cover_path', None) if cpath is None: pf = open(os.path.join(dir, 'cover.jpg'), 'wb') if self.default_cover(pf): cpath = pf.name if cpath is not None and os.access(cpath, os.R_OK): opf.cover = cpath manifest.append(cpath) # Get masthead mpath = getattr(self, 'masthead_path', None) if mpath is not None and os.access(mpath, os.R_OK): manifest.append(mpath) opf.create_manifest_from_files_in(manifest) for mani in opf.manifest: if mani.path.endswith('.ncx'): mani.id = 'ncx' if mani.path.endswith('mastheadImage.jpg'): mani.id = 'masthead-image' entries = ['index.html'] toc = TOC(base_path=dir) self.play_order_counter = 0 self.play_order_map = {} self.article_url_map = aumap = defaultdict(set) def feed_index(num, parent): f = feeds[num] for j, a in enumerate(f): if getattr(a, 'downloaded', False): adir = 'feed_%d/article_%d/'%(num, j) auth = a.author if not auth: auth = None desc = a.text_summary if not desc: desc = None else: desc = self.description_limiter(desc) tt = a.toc_thumbnail if a.toc_thumbnail else None entries.append('%sindex.html'%adir) po = self.play_order_map.get(entries[-1], None) if po is None: self.play_order_counter += 1 po = self.play_order_counter arelpath = '%sindex.html'%adir for curl in self.canonicalize_internal_url(a.orig_url, is_link=False): aumap[curl].add(arelpath) article_toc_entry = parent.add_item(arelpath, None, a.title if a.title else _('Untitled article'), play_order=po, author=auth, description=desc, toc_thumbnail=tt) for entry in a.internal_toc_entries: anchor = entry.get('anchor') if anchor: self.play_order_counter += 1 po += 1 article_toc_entry.add_item( arelpath, entry['anchor'], entry['title'] or _('Unknown section'), play_order=po ) last = os.path.join(self.output_dir, ('%sindex.html'%adir).replace('/', os.sep)) for sp in a.sub_pages: prefix = os.path.commonprefix([opf_path, sp]) relp = sp[len(prefix):] entries.append(relp.replace(os.sep, '/')) last = sp if os.path.exists(last): with open(last, 'rb') as fi: src = fi.read().decode('utf-8') soup = BeautifulSoup(src) body = soup.find('body') if body is not None: prefix = '/'.join('..'for i in range(2*len(re.findall(r'link\d+', last)))) templ = self.navbar.generate(True, num, j, len(f), not self.has_single_feed, a.orig_url, __appname__, prefix=prefix, center=self.center_navbar) elem = BeautifulSoup(templ.render(doctype='xhtml').decode('utf-8')).find('div') body.insert(len(body.contents), elem) with open(last, 'wb') as fi: fi.write(str(soup).encode('utf-8')) if len(feeds) == 0: raise Exception('All feeds are empty, aborting.') if len(feeds) > 1: for i, f in enumerate(feeds): entries.append('feed_%d/index.html'%i) po = self.play_order_map.get(entries[-1], None) if po is None: self.play_order_counter += 1 po = self.play_order_counter auth = getattr(f, 'author', None) if not auth: auth = None desc = getattr(f, 'description', None) if not desc: desc = None feed_index(i, toc.add_item('feed_%d/index.html'%i, None, f.title, play_order=po, description=desc, author=auth)) else: entries.append('feed_%d/index.html'%0) feed_index(0, toc) for i, p in enumerate(entries): entries[i] = os.path.join(dir, p.replace('/', os.sep)) opf.create_spine(entries) opf.set_toc(toc) with open(opf_path, 'wb') as opf_file, open(ncx_path, 'wb') as ncx_file: opf.render(opf_file, ncx_file) def article_downloaded(self, request, result): index = os.path.join(os.path.dirname(result[0]), 'index.html') if index != result[0]: if os.path.exists(index): os.remove(index) os.rename(result[0], index) a = request.requestID[1] article = request.article self.log.debug('Downloaded article:', article.title, 'from', article.url) article.orig_url = article.url article.url = 'article_%d/index.html'%a article.downloaded = True article.sub_pages = result[1][1:] self.jobs_done += 1 self.report_progress(float(self.jobs_done)/len(self.jobs), _('Article downloaded: %s')%force_unicode(article.title)) if result[2]: self.partial_failures.append((request.feed.title, article.title, article.url, result[2])) def error_in_article_download(self, request, traceback): self.jobs_done += 1 if traceback and re.search('^AbortArticle:', traceback, flags=re.M) is not None: self.log.warn('Aborted download of article:', request.article.title, 'from', request.article.url) self.report_progress(float(self.jobs_done)/len(self.jobs), _('Article download aborted: %s')%force_unicode(request.article.title)) self.aborted_articles.append((request.feed, request.article)) else: self.log.error('Failed to download article:', request.article.title, 'from', request.article.url) self.log.debug(traceback) self.log.debug('\n') self.report_progress(float(self.jobs_done)/len(self.jobs), _('Article download failed: %s')%force_unicode(request.article.title)) self.failed_downloads.append((request.feed, request.article, traceback)) def parse_feeds(self): ''' Create a list of articles from the list of feeds returned by :meth:`BasicNewsRecipe.get_feeds`. Return a list of :class:`Feed` objects. ''' feeds = self.get_feeds() parsed_feeds = [] br = self.browser for obj in feeds: if isinstance(obj, string_or_bytes): title, url = None, obj else: title, url = obj if isinstance(title, bytes): title = title.decode('utf-8') if isinstance(url, bytes): url = url.decode('utf-8') if url.startswith('feed://'): url = 'http'+url[4:] self.report_progress(0, _('Fetching feed')+' %s...'%(title if title else url)) try: purl = urlparse(url, allow_fragments=False) if purl.username or purl.password: hostname = purl.hostname if purl.port: hostname += f':{purl.port}' url = purl._replace(netloc=hostname).geturl() if purl.username and purl.password: br.add_password(url, purl.username, purl.password) with closing(br.open_novisit(url, timeout=self.timeout)) as f: raw = f.read() parsed_feeds.append(feed_from_xml( raw, title=title, log=self.log, oldest_article=self.oldest_article, max_articles_per_feed=self.max_articles_per_feed, get_article_url=self.get_article_url )) except Exception as err: feed = Feed() msg = 'Failed feed: %s'%(title if title else url) feed.populate_from_preparsed_feed(msg, []) feed.description = as_unicode(err) parsed_feeds.append(feed) self.log.exception(msg) delay = self.get_url_specific_delay(url) if delay > 0: time.sleep(delay) remove = [fl for fl in parsed_feeds if len(fl) == 0 and self.remove_empty_feeds] for f in remove: parsed_feeds.remove(f) return parsed_feeds @classmethod def tag_to_string(self, tag, use_alt=True, normalize_whitespace=True): ''' Convenience method to take a `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/>`_ :code:`Tag` and extract the text from it recursively, including any CDATA sections and alt tag attributes. Return a possibly empty Unicode string. `use_alt`: If `True` try to use the alt attribute for tags that don't have any textual content `tag`: `BeautifulSoup <https://www.crummy.com/software/BeautifulSoup/bs4/doc/>`_ :code:`Tag` ''' if tag is None: return '' if isinstance(tag, string_or_bytes): return tag if callable(getattr(tag, 'xpath', None)) and not hasattr(tag, 'contents'): # a lxml tag from lxml.etree import tostring ans = tostring(tag, method='text', encoding='unicode', with_tail=False) else: strings = [] for item in tag.contents: if isinstance(item, (NavigableString, CData)): strings.append(item.string) elif isinstance(item, Tag): res = self.tag_to_string(item) if res: strings.append(res) elif use_alt: try: strings.append(item['alt']) except KeyError: pass ans = ''.join(strings) if normalize_whitespace: ans = re.sub(r'\s+', ' ', ans) return ans @classmethod def soup(cls, raw): return BeautifulSoup(raw) @classmethod def adeify_images(cls, soup): ''' If your recipe when converted to EPUB has problems with images when viewed in Adobe Digital Editions, call this method from within :meth:`postprocess_html`. ''' for item in soup.findAll('img'): for attrib in ['height','width','border','align','style']: try: del item[attrib] except KeyError: pass oldParent = item.parent myIndex = oldParent.contents.index(item) item.extract() divtag = soup.new_tag('div') brtag = soup.new_tag('br') oldParent.insert(myIndex,divtag) divtag.append(item) divtag.append(brtag) return soup def internal_postprocess_book(self, oeb, opts, log): if self.resolve_internal_links and self.article_url_map: seen = set() for item in oeb.spine: for a in item.data.xpath('//*[local-name()="a" and @href]'): if a.get('rel') == 'calibre-downloaded-from': continue url = a.get('href') for curl in self.canonicalize_internal_url(url): articles = self.article_url_map.get(curl) if articles: arelpath = sorted(articles, key=numeric_sort_key)[0] a.set('href', item.relhref(arelpath)) if a.text and len(a) == 0: a.text = a.text + '·' # mark as local link if url not in seen: log.debug(f'Resolved internal URL: {url} -> {arelpath}') seen.add(url) class CustomIndexRecipe(BasicNewsRecipe): def custom_index(self): ''' Return the filesystem path to a custom HTML document that will serve as the index for this recipe. The index document will typically contain many `<a href="...">` tags that point to resources on the internet that should be downloaded. ''' raise NotImplementedError def create_opf(self): mi = MetaInformation(self.title + strftime(self.timefmt), [__appname__]) mi.publisher = __appname__ mi.author_sort = __appname__ mi = OPFCreator(self.output_dir, mi) mi.create_manifest_from_files_in([self.output_dir]) mi.create_spine([os.path.join(self.output_dir, 'index.html')]) with open(os.path.join(self.output_dir, 'index.opf'), 'wb') as opf_file: mi.render(opf_file) def download(self): index = os.path.abspath(self.custom_index()) url = 'file:'+index if iswindows else 'file://'+index self.web2disk_options.browser = self.clone_browser(self.browser) fetcher = RecursiveFetcher(self.web2disk_options, self.log) fetcher.base_dir = self.output_dir fetcher.current_dir = self.output_dir fetcher.show_progress = False res = fetcher.start_fetch(url) self.create_opf() return res class AutomaticNewsRecipe(BasicNewsRecipe): auto_cleanup = True class CalibrePeriodical(BasicNewsRecipe): #: Set this to the slug for the calibre periodical calibre_periodicals_slug = None LOG_IN = 'https://news.calibre-ebook.com/accounts/login' needs_subscription = True __author__ = 'calibre Periodicals' def get_browser(self): br = BasicNewsRecipe.get_browser(self) br.open(self.LOG_IN) br.select_form(name='login') br['username'] = self.username br['password'] = self.password raw = br.submit().read() if 'href="/my-account"' not in raw: raise LoginFailed( _('Failed to log in, check your username and password for' ' the calibre Periodicals service.')) return br get_browser.is_base_class_implementation = True def download(self): self.log('Fetching downloaded recipe') try: raw = self.browser.open_novisit( 'https://news.calibre-ebook.com/subscribed_files/%s/0/temp.downloaded_recipe' % self.calibre_periodicals_slug ).read() except Exception as e: if hasattr(e, 'getcode') and e.getcode() == 403: raise DownloadDenied( _('You do not have permission to download this issue.' ' Either your subscription has expired or you have' ' exceeded the maximum allowed downloads for today.')) raise f = io.BytesIO(raw) from calibre.utils.zipfile import ZipFile zf = ZipFile(f) zf.extractall() zf.close() from glob import glob from calibre.web.feeds.recipes import compile_recipe try: recipe = compile_recipe(open(glob('*.recipe')[0], 'rb').read()) self.conversion_options = recipe.conversion_options except: self.log.exception('Failed to compile downloaded recipe') return os.path.abspath('index.html')
83,763
Python
.py
1,729
37.218623
160
0.593016
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,813
collection.py
kovidgoyal_calibre/src/calibre/web/feeds/recipes/collection.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2009, Kovid Goyal <kovid@kovidgoyal.net>' __docformat__ = 'restructuredtext en' import calendar import json import os import zipfile from datetime import timedelta from threading import RLock from typing import Dict, NamedTuple, Optional, Sequence from lxml import etree from lxml.builder import ElementMaker from calibre import force_unicode from calibre.constants import numeric_version from calibre.utils.date import EPOCH, UNDEFINED_DATE, isoformat, local_tz, utcnow from calibre.utils.date import now as nowf from calibre.utils.iso8601 import parse_iso8601 from calibre.utils.localization import _ from calibre.utils.recycle_bin import delete_file from calibre.utils.resources import get_path as P from calibre.utils.xml_parse import safe_xml_fromstring from polyglot.builtins import iteritems NS = 'http://calibre-ebook.com/recipe_collection' E = ElementMaker(namespace=NS, nsmap={None:NS}) def iterate_over_builtin_recipe_files(): exclude = ['craigslist', 'toronto_sun'] d = os.path.dirname base = os.path.join(d(d(d(d(d(d(os.path.abspath(__file__))))))), 'recipes') for f in os.listdir(base): fbase, ext = os.path.splitext(f) if ext != '.recipe' or fbase in exclude: continue f = os.path.join(base, f) rid = os.path.splitext(os.path.relpath(f, base).replace(os.sep, '/'))[0] yield rid, f def normalize_language(x: str) -> str: lang, sep, country = x.replace('-', '_').partition('_') if sep == '_': x = f'{lang.lower()}{sep}{country.upper()}' else: x = lang.lower() return x def serialize_recipe(urn, recipe_class): from xml.sax.saxutils import quoteattr def attr(n, d, normalize=lambda x: x): ans = getattr(recipe_class, n, d) if isinstance(ans, bytes): ans = ans.decode('utf-8', 'replace') return quoteattr(normalize(ans)) default_author = _('You') if urn.startswith('custom:') else _('Unknown') ns = getattr(recipe_class, 'needs_subscription', False) if not ns: ns = 'no' if ns is True: ns = 'yes' options = '' rso = getattr(recipe_class, 'recipe_specific_options', None) if rso: options = f' options={quoteattr(json.dumps(rso))}' return (' <recipe id={id} title={title} author={author} language={language}' ' needs_subscription={needs_subscription} description={description}{options}/>').format(**{ 'id' : quoteattr(str(urn)), 'title' : attr('title', _('Unknown')), 'author' : attr('__author__', default_author), 'language' : attr('language', 'und', normalize_language), 'needs_subscription' : quoteattr(ns), 'description' : attr('description', ''), 'options' : options, }) def serialize_collection(mapping_of_recipe_classes): collection = [] for urn in sorted(mapping_of_recipe_classes.keys(), key=lambda key: force_unicode( getattr(mapping_of_recipe_classes[key], 'title', 'zzz'), 'utf-8')): try: recipe = serialize_recipe(urn, mapping_of_recipe_classes[urn]) except: import traceback traceback.print_exc() continue collection.append(recipe) items = '\n'.join(collection) return f'''<?xml version='1.0' encoding='utf-8'?> <recipe_collection xmlns="http://calibre-ebook.com/recipe_collection" count="{len(collection)}"> {items} </recipe_collection>'''.encode() def serialize_builtin_recipes(): from calibre.web.feeds.recipes import compile_recipe recipe_mapping = {} for rid, f in iterate_over_builtin_recipe_files(): with open(f, 'rb') as stream: try: recipe_class = compile_recipe(stream.read()) except: print('Failed to compile: %s'%f) raise if recipe_class is not None: recipe_mapping['builtin:'+rid] = recipe_class return serialize_collection(recipe_mapping) def get_builtin_recipe_collection(): return etree.parse(P('builtin_recipes.xml', allow_user_override=False)).getroot() def get_custom_recipe_collection(*args): from calibre.web.feeds.recipes import compile_recipe, custom_recipes bdir = os.path.dirname(custom_recipes.file_path) rmap = {} for id_, x in iteritems(custom_recipes): title, fname = x recipe = os.path.join(bdir, fname) try: with open(recipe, 'rb') as f: recipe = f.read().decode('utf-8') recipe_class = compile_recipe(recipe) if recipe_class is not None: rmap['custom:%s'%id_] = recipe_class except: print('Failed to load recipe from: %r'%fname) import traceback traceback.print_exc() continue return safe_xml_fromstring(serialize_collection(rmap), recover=False) def update_custom_recipe(id_, title, script): update_custom_recipes([(id_, title, script)]) def update_custom_recipes(script_ids): from calibre.web.feeds.recipes import custom_recipe_filename, custom_recipes bdir = os.path.dirname(custom_recipes.file_path) for id_, title, script in script_ids: id_ = str(int(id_)) existing = custom_recipes.get(id_, None) if existing is None: fname = custom_recipe_filename(id_, title) else: fname = existing[1] if isinstance(script, str): script = script.encode('utf-8') custom_recipes[id_] = (title, fname) if not os.path.exists(bdir): os.makedirs(bdir) with open(os.path.join(bdir, fname), 'wb') as f: f.write(script) def add_custom_recipe(title, script): add_custom_recipes({title:script}) def add_custom_recipes(script_map): from calibre.web.feeds.recipes import custom_recipe_filename, custom_recipes id_ = 1000 keys = tuple(map(int, custom_recipes)) if keys: id_ = max(keys)+1 bdir = os.path.dirname(custom_recipes.file_path) with custom_recipes: for title, script in iteritems(script_map): fid = str(id_) fname = custom_recipe_filename(fid, title) if isinstance(script, str): script = script.encode('utf-8') custom_recipes[fid] = (title, fname) if not os.path.exists(bdir): os.makedirs(bdir) with open(os.path.join(bdir, fname), 'wb') as f: f.write(script) id_ += 1 def remove_custom_recipe(id_): from calibre.web.feeds.recipes import custom_recipes id_ = str(int(id_)) existing = custom_recipes.get(id_, None) if existing is not None: bdir = os.path.dirname(custom_recipes.file_path) fname = existing[1] del custom_recipes[id_] try: delete_file(os.path.join(bdir, fname)) except: pass def get_custom_recipe(id_): from calibre.web.feeds.recipes import custom_recipes id_ = str(int(id_)) existing = custom_recipes.get(id_, None) if existing is not None: bdir = os.path.dirname(custom_recipes.file_path) fname = existing[1] with open(os.path.join(bdir, fname), 'rb') as f: return f.read().decode('utf-8') def get_builtin_recipe_titles(): return [r.get('title') for r in get_builtin_recipe_collection()] def download_builtin_recipe(urn): import bz2 from calibre.utils.config_base import prefs from calibre.utils.https import get_https_resource_securely recipe_source = bz2.decompress(get_https_resource_securely( 'https://code.calibre-ebook.com/recipe-compressed/'+urn, headers={'CALIBRE-INSTALL-UUID':prefs['installation_uuid']})) recipe_source = recipe_source.decode('utf-8') from calibre.web.feeds.recipes import compile_recipe recipe = compile_recipe(recipe_source) # ensure the downloaded recipe is at least compile-able if recipe is None: raise ValueError('Failed to find recipe object in downloaded recipe: ' + urn) if recipe.requires_version > numeric_version: raise ValueError(f'Downloaded recipe for {urn} requires calibre >= {recipe.requires_version}') return recipe_source def get_builtin_recipe(urn): with zipfile.ZipFile(P('builtin_recipes.zip', allow_user_override=False), 'r') as zf: return zf.read(urn+'.recipe').decode('utf-8') def get_builtin_recipe_by_title(title, log=None, download_recipe=False): for x in get_builtin_recipe_collection(): if x.get('title') == title: urn = x.get('id')[8:] if download_recipe: try: if log is not None: log('Trying to get latest version of recipe:', urn) return download_builtin_recipe(urn) except: if log is None: import traceback traceback.print_exc() else: log.exception( 'Failed to download recipe, using builtin version') return get_builtin_recipe(urn) def get_builtin_recipe_by_id(id_, log=None, download_recipe=False): for x in get_builtin_recipe_collection(): if x.get('id') == id_: urn = x.get('id')[8:] if download_recipe: try: if log is not None: log('Trying to get latest version of recipe:', urn) return download_builtin_recipe(urn) except: if log is None: import traceback traceback.print_exc() else: log.exception( 'Failed to download recipe, using builtin version') return get_builtin_recipe(urn) class RecipeCustomization(NamedTuple): add_title_tag: bool = False custom_tags: Sequence[str] = () keep_issues: int = 0 recipe_specific_options: Optional[Dict[str, str]] = None class SchedulerConfig: def __init__(self): from calibre.utils.config import config_dir from calibre.utils.lock import ExclusiveFile self.conf_path = os.path.join(config_dir, 'scheduler.xml') old_conf_path = os.path.join(config_dir, 'scheduler.pickle') self.root = E.recipe_collection() self.lock = RLock() if os.access(self.conf_path, os.R_OK): with ExclusiveFile(self.conf_path) as f: try: self.root = safe_xml_fromstring(f.read(), recover=False) except: print('Failed to read recipe scheduler config') import traceback traceback.print_exc() elif os.path.exists(old_conf_path): self.migrate_old_conf(old_conf_path) def iter_recipes(self): for x in self.root: if x.tag == '{%s}scheduled_recipe'%NS: yield x def iter_accounts(self): for x in self.root: if x.tag == '{%s}account_info'%NS: yield x def iter_customization(self): for x in self.root: if x.tag == '{%s}recipe_customization'%NS: yield x def schedule_recipe(self, recipe, schedule_type, schedule, last_downloaded=None): with self.lock: for x in list(self.iter_recipes()): if x.get('id', False) == recipe.get('id'): ld = x.get('last_downloaded', None) if ld and last_downloaded is None: try: last_downloaded = parse_iso8601(ld) except Exception: pass self.root.remove(x) break if last_downloaded is None: last_downloaded = EPOCH sr = E.scheduled_recipe({ 'id' : recipe.get('id'), 'title': recipe.get('title'), 'last_downloaded':isoformat(last_downloaded), }, self.serialize_schedule(schedule_type, schedule)) self.root.append(sr) self.write_scheduler_file() # 'keep_issues' argument for recipe-specific number of copies to keep def customize_recipe(self, urn, val: RecipeCustomization): with self.lock: for x in list(self.iter_customization()): if x.get('id') == urn: self.root.remove(x) cs = E.recipe_customization({ 'keep_issues' : str(val.keep_issues), 'id' : urn, 'add_title_tag' : 'yes' if val.add_title_tag else 'no', 'custom_tags' : ','.join(val.custom_tags), 'recipe_specific_options': json.dumps(val.recipe_specific_options or {}), }) self.root.append(cs) self.write_scheduler_file() def un_schedule_recipe(self, recipe_id): with self.lock: for x in list(self.iter_recipes()): if x.get('id', False) == recipe_id: self.root.remove(x) break self.write_scheduler_file() def update_last_downloaded(self, recipe_id): with self.lock: now = utcnow() for x in self.iter_recipes(): if x.get('id', False) == recipe_id: typ, sch, last_downloaded = self.un_serialize_schedule(x) if typ == 'interval': # Prevent downloads more frequent than once an hour actual_interval = now - last_downloaded nominal_interval = timedelta(days=sch) if abs(actual_interval - nominal_interval) < \ timedelta(hours=1): now = last_downloaded + nominal_interval x.set('last_downloaded', isoformat(now)) break self.write_scheduler_file() def get_to_be_downloaded_recipes(self): ans = [] with self.lock: for recipe in self.iter_recipes(): if self.recipe_needs_to_be_downloaded(recipe): ans.append(recipe.get('id')) return ans def write_scheduler_file(self): from calibre.utils.lock import ExclusiveFile self.root.text = '\n\n\t' for x in self.root: x.tail = '\n\n\t' if len(self.root) > 0: self.root[-1].tail = '\n\n' with ExclusiveFile(self.conf_path) as f: f.seek(0) f.truncate() f.write(etree.tostring(self.root, encoding='utf-8', xml_declaration=True, pretty_print=False)) def serialize_schedule(self, typ, schedule): s = E.schedule({'type':typ}) if typ == 'interval': if schedule < 0.04: schedule = 0.04 text = '%f'%schedule elif typ == 'day/time': text = '%d:%d:%d'%schedule elif typ in ('days_of_week', 'days_of_month'): dw = ','.join(map(str, map(int, schedule[0]))) text = '%s:%d:%d'%(dw, schedule[1], schedule[2]) else: raise ValueError('Unknown schedule type: %r'%typ) s.text = text return s def un_serialize_schedule(self, recipe): for x in recipe.iterdescendants(): if 'schedule' in x.tag: sch, typ = x.text, x.get('type') if typ == 'interval': sch = float(sch) elif typ == 'day/time': sch = list(map(int, sch.split(':'))) elif typ in ('days_of_week', 'days_of_month'): parts = sch.split(':') days = list(map(int, [x.strip() for x in parts[0].split(',')])) sch = [days, int(parts[1]), int(parts[2])] try: ld = parse_iso8601(recipe.get('last_downloaded')) except Exception: ld = UNDEFINED_DATE return typ, sch, ld def recipe_needs_to_be_downloaded(self, recipe): try: typ, sch, ld = self.un_serialize_schedule(recipe) except: return False def is_time(now, hour, minute): return now.hour > hour or \ (now.hour == hour and now.minute >= minute) def is_weekday(day, now): return day < 0 or day > 6 or \ day == calendar.weekday(now.year, now.month, now.day) def was_downloaded_already_today(ld_local, now): return ld_local.date() == now.date() if typ == 'interval': return utcnow() - ld > timedelta(sch) elif typ == 'day/time': now = nowf() try: ld_local = ld.astimezone(local_tz) except Exception: return False day, hour, minute = sch return is_weekday(day, now) and \ not was_downloaded_already_today(ld_local, now) and \ is_time(now, hour, minute) elif typ == 'days_of_week': now = nowf() try: ld_local = ld.astimezone(local_tz) except Exception: return False days, hour, minute = sch have_day = False for day in days: if is_weekday(day, now): have_day = True break return have_day and \ not was_downloaded_already_today(ld_local, now) and \ is_time(now, hour, minute) elif typ == 'days_of_month': now = nowf() try: ld_local = ld.astimezone(local_tz) except Exception: return False days, hour, minute = sch have_day = now.day in days return have_day and \ not was_downloaded_already_today(ld_local, now) and \ is_time(now, hour, minute) return False def set_account_info(self, urn, un, pw): with self.lock: for x in list(self.iter_accounts()): if x.get('id', False) == urn: self.root.remove(x) break ac = E.account_info({'id':urn, 'username':un, 'password':pw}) self.root.append(ac) self.write_scheduler_file() def get_account_info(self, urn): with self.lock: for x in self.iter_accounts(): if x.get('id', False) == urn: return x.get('username', ''), x.get('password', '') def clear_account_info(self, urn): with self.lock: for x in self.iter_accounts(): if x.get('id', False) == urn: x.getparent().remove(x) self.write_scheduler_file() break def get_customize_info(self, urn): keep_issues = 0 add_title_tag = True custom_tags = () recipe_specific_options = {} with self.lock: for x in self.iter_customization(): if x.get('id', False) == urn: keep_issues = int(x.get('keep_issues', '0')) add_title_tag = x.get('add_title_tag', 'yes') == 'yes' custom_tags = tuple(i.strip() for i in x.get('custom_tags', '').split(',')) recipe_specific_options = json.loads(x.get('recipe_specific_options', '{}')) break return RecipeCustomization(add_title_tag, custom_tags, keep_issues, recipe_specific_options) def get_schedule_info(self, urn): with self.lock: for x in self.iter_recipes(): if x.get('id', False) == urn: ans = list(self.un_serialize_schedule(x)) return ans def migrate_old_conf(self, old_conf_path): from calibre.utils.config import DynamicConfig c = DynamicConfig('scheduler') for r in c.get('scheduled_recipes', []): try: self.add_old_recipe(r) except: continue for k in c.keys(): if k.startswith('recipe_account_info'): try: urn = k.replace('recipe_account_info_', '') if urn.startswith('recipe_'): urn = 'builtin:'+urn[7:] else: urn = 'custom:%d'%int(urn) try: username, password = c[k] except: username = password = '' self.set_account_info(urn, str(username), str(password)) except: continue del c self.write_scheduler_file() try: os.remove(old_conf_path) except: pass def add_old_recipe(self, r): urn = None if r['builtin'] and r['id'].startswith('recipe_'): urn = 'builtin:'+r['id'][7:] elif not r['builtin']: try: urn = 'custom:%d'%int(r['id']) except: return schedule = r['schedule'] typ = 'interval' if schedule > 1e5: typ = 'day/time' raw = '%d'%int(schedule) day = int(raw[0]) - 1 hour = int(raw[2:4]) - 1 minute = int(raw[-2:]) - 1 if day >= 7: day = -1 schedule = [day, hour, minute] recipe = {'id':urn, 'title':r['title']} self.schedule_recipe(recipe, typ, schedule, last_downloaded=r['last_downloaded'])
22,209
Python
.py
529
30.15879
126
0.547884
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,814
__init__.py
kovidgoyal_calibre/src/calibre/web/feeds/recipes/__init__.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2008, Kovid Goyal <kovid at kovidgoyal.net>' ''' Builtin recipes. ''' import io import re import time from calibre.ebooks.BeautifulSoup import BeautifulSoup from calibre.utils.config import JSONConfig from calibre.web.feeds.news import AutomaticNewsRecipe, BasicNewsRecipe, CalibrePeriodical, CustomIndexRecipe from polyglot.builtins import codepoint_to_chr, itervalues basic_recipes = (BasicNewsRecipe, AutomaticNewsRecipe, CustomIndexRecipe, CalibrePeriodical) custom_recipes = JSONConfig('custom_recipes/index.json') def custom_recipe_filename(id_, title): from calibre.utils.filenames import ascii_filename return ascii_filename(title[:50]) + \ ('_%s.recipe'%id_) def compile_recipe(src): ''' Compile the code in src and return a recipe object, if found. :param src: Python source code as bytestring or unicode object :return: Recipe class or None, if no such class was found in src ''' if not isinstance(src, str): match = re.search(br'coding[:=]\s*([-\w.]+)', src[:200]) enc = match.group(1).decode('utf-8') if match else 'utf-8' src = src.decode(enc) # Python complains if there is a coding declaration in a unicode string src = re.sub(r'^#.*coding\s*[:=]\s*([-\w.]+)', '#', src.lstrip('\ufeff'), flags=re.MULTILINE) # Translate newlines to \n src = io.StringIO(src, newline=None).getvalue() namespace = { 'BasicNewsRecipe':BasicNewsRecipe, 'AutomaticNewsRecipe':AutomaticNewsRecipe, 'time':time, 're':re, 'BeautifulSoup':BeautifulSoup, 'unicode': str, 'unichr': codepoint_to_chr, 'xrange': range, } exec(src, namespace) ua = namespace.get('calibre_most_common_ua') for x in itervalues(namespace): if (isinstance(x, type) and issubclass(x, BasicNewsRecipe) and x not in basic_recipes): x.calibre_most_common_ua = ua return x return None
2,081
Python
.py
51
34.45098
109
0.667328
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,815
model.py
kovidgoyal_calibre/src/calibre/web/feeds/recipes/model.py
#!/usr/bin/env python # License: GPLv3 Copyright: 2009, Kovid Goyal <kovid at kovidgoyal.net> import copy import zipfile from functools import total_ordering from qt.core import QAbstractItemModel, QApplication, QFont, QIcon, QModelIndex, QPalette, QPixmap, Qt, pyqtSignal from calibre import force_unicode from calibre.utils.icu import primary_sort_key from calibre.utils.localization import _, countrycode_to_name, get_language from calibre.utils.resources import get_path as P from calibre.utils.search_query_parser import ParseException, SearchQueryParser from calibre.web.feeds.recipes.collection import ( SchedulerConfig, add_custom_recipe, add_custom_recipes, download_builtin_recipe, get_builtin_recipe, get_builtin_recipe_collection, get_custom_recipe, get_custom_recipe_collection, remove_custom_recipe, update_custom_recipe, update_custom_recipes, ) from polyglot.builtins import iteritems class NewsTreeItem: def __init__(self, builtin, custom, scheduler_config, parent=None): self.builtin, self.custom = builtin, custom self.scheduler_config = scheduler_config self.parent = parent if self.parent is not None: self.parent.append(self) self.children = [] def row(self): if self.parent is not None: return self.parent.children.index(self) return 0 def append(self, child): child.parent = self self.children.append(child) def data(self, role): return None def flags(self): return Qt.ItemFlag.ItemIsEnabled|Qt.ItemFlag.ItemIsSelectable def sort(self): self.children.sort() for child in self.children: child.sort() def prune(self): for child in list(self.children): if len(child.children) == 0: self.children.remove(child) child.parent = None def parse_lang_code(x: str) -> str: lang, sep, country = x.partition('_') country = country.upper() ans = get_language(lang) if country: ans = _('{language} ({country})').format(language=ans, country=countrycode_to_name(country)) return ans @total_ordering class NewsCategory(NewsTreeItem): def __init__(self, category, builtin, custom, scheduler_config, parent): NewsTreeItem.__init__(self, builtin, custom, scheduler_config, parent) self.category = self.cdata = category self.cdata = self.category if self.category == _('Scheduled'): self.sortq = 0, '' elif self.category == _('Custom'): self.sortq = 1, '' else: self.cdata = parse_lang_code(self.cdata) self.sortq = 2, self.cdata self.bold_font = QFont() self.bold_font.setBold(True) self.bold_font = (self.bold_font) def data(self, role): if role == Qt.ItemDataRole.DisplayRole: return (self.cdata + ' [%d]'%len(self.children)) elif role == Qt.ItemDataRole.FontRole: return self.bold_font elif role == Qt.ItemDataRole.ForegroundRole and self.category == _('Scheduled'): return QApplication.instance().palette().color(QPalette.ColorRole.Link) elif role == Qt.ItemDataRole.UserRole: return f'::category::{self.sortq[0]}' return None def flags(self): return Qt.ItemFlag.ItemIsEnabled def __eq__(self, other): return self.cdata == other.cdata def __lt__(self, other): return self.sortq < getattr(other, 'sortq', (3, '')) @total_ordering class NewsItem(NewsTreeItem): def __init__(self, urn, title, default_icon, custom_icon, favicons, zf, builtin, custom, scheduler_config, parent): NewsTreeItem.__init__(self, builtin, custom, scheduler_config, parent) self.urn, self.title = urn, title if isinstance(self.title, bytes): self.title = force_unicode(self.title) self.sortq = primary_sort_key(self.title) self.icon = self.default_icon = None self.default_icon = default_icon self.favicons, self.zf = favicons, zf if 'custom:' in self.urn: self.icon = custom_icon def data(self, role): if role == Qt.ItemDataRole.DisplayRole: return (self.title) if role == Qt.ItemDataRole.DecorationRole: if self.icon is None: icon = '%s.png'%self.urn[8:] p = QPixmap() if icon in self.favicons: try: with zipfile.ZipFile(self.zf, 'r') as zf: p.loadFromData(zf.read(self.favicons[icon])) except Exception: pass if not p.isNull(): self.icon = (QIcon(p)) else: self.icon = self.default_icon return self.icon if role == Qt.ItemDataRole.UserRole: return self.urn def __eq__(self, other): return self.urn == other.urn def __lt__(self, other): return self.sortq < other.sortq class AdaptSQP(SearchQueryParser): def __init__(self, *args, **kwargs): pass class RecipeModel(QAbstractItemModel, AdaptSQP): LOCATIONS = ['all'] searched = pyqtSignal(object) def __init__(self, *args): QAbstractItemModel.__init__(self, *args) SearchQueryParser.__init__(self, locations=['all']) self.default_icon = (QIcon.ic('news.png')) self.custom_icon = (QIcon.ic('user_profile.png')) self.builtin_recipe_collection = get_builtin_recipe_collection() self.scheduler_config = SchedulerConfig() try: with zipfile.ZipFile(P('builtin_recipes.zip', allow_user_override=False), 'r') as zf: self.favicons = {x.filename: x for x in zf.infolist() if x.filename.endswith('.png')} except: self.favicons = {} self.do_refresh() def get_builtin_recipe(self, urn, download=True): if download: try: return download_builtin_recipe(urn) except: import traceback traceback.print_exc() return get_builtin_recipe(urn) def get_recipe(self, urn, download=True): coll = self.custom_recipe_collection if urn.startswith('custom:') \ else self.builtin_recipe_collection for recipe in coll: if recipe.get('id', False) == urn: if coll is self.builtin_recipe_collection: return self.get_builtin_recipe(urn[8:], download=download) return get_custom_recipe(int(urn[len('custom:'):])) def update_custom_recipe(self, urn, title, script): id_ = int(urn[len('custom:'):]) update_custom_recipe(id_, title, script) self.custom_recipe_collection = get_custom_recipe_collection() def update_custom_recipes(self, script_urn_map): script_ids = [] for urn, title_script in iteritems(script_urn_map): id_ = int(urn[len('custom:'):]) (title, script) = title_script script_ids.append((id_, title, script)) update_custom_recipes(script_ids) self.custom_recipe_collection = get_custom_recipe_collection() def add_custom_recipe(self, title, script): add_custom_recipe(title, script) self.custom_recipe_collection = get_custom_recipe_collection() def add_custom_recipes(self, scriptmap): add_custom_recipes(scriptmap) self.custom_recipe_collection = get_custom_recipe_collection() def remove_custom_recipes(self, urns): ids = [int(x[len('custom:'):]) for x in urns] for id_ in ids: remove_custom_recipe(id_) self.custom_recipe_collection = get_custom_recipe_collection() def do_refresh(self, restrict_to_urns=frozenset()): self.custom_recipe_collection = get_custom_recipe_collection() zf = P('builtin_recipes.zip', allow_user_override=False) def factory(cls, parent, *args): args = list(args) if cls is NewsItem: args.extend([self.default_icon, self.custom_icon, self.favicons, zf]) args += [self.builtin_recipe_collection, self.custom_recipe_collection, self.scheduler_config, parent] return cls(*args) def ok(urn): if restrict_to_urns is None: return False return not restrict_to_urns or urn in restrict_to_urns new_root = factory(NewsTreeItem, None) scheduled = factory(NewsCategory, new_root, _('Scheduled')) custom = factory(NewsCategory, new_root, _('Custom')) lang_map = {} self.all_urns = set() self.showing_count = 0 self.builtin_count = 0 for x in self.custom_recipe_collection: urn = x.get('id') self.all_urns.add(urn) if ok(urn): factory(NewsItem, custom, urn, x.get('title')) self.showing_count += 1 for x in self.builtin_recipe_collection: urn = x.get('id') self.all_urns.add(urn) if ok(urn): lang = x.get('language', 'und') if lang: lang = lang.replace('-', '_') if lang not in lang_map: lang_map[lang] = factory(NewsCategory, new_root, lang) factory(NewsItem, lang_map[lang], urn, x.get('title')) self.showing_count += 1 self.builtin_count += 1 for x in self.scheduler_config.iter_recipes(): urn = x.get('id') if urn not in self.all_urns: self.scheduler_config.un_schedule_recipe(urn) continue if ok(urn): factory(NewsItem, scheduled, urn, x.get('title')) new_root.prune() new_root.sort() self.root = new_root self.reset() def reset(self): self.beginResetModel(), self.endResetModel() def recipe_from_urn(self, urn): coll = self.custom_recipe_collection if 'custom:' in urn else \ self.builtin_recipe_collection for x in coll: if x.get('id', None) == urn: return copy.deepcopy(x) def schedule_info_from_urn(self, urn): return self.scheduler_config.get_schedule_info(urn) def account_info_from_urn(self, urn): return self.scheduler_config.get_account_info(urn) def universal_set(self): return self.all_urns def get_customize_info(self, urn): return self.scheduler_config.get_customize_info(urn) def get_recipe_specific_option_metadata(self, urn): return self.scheduler_config.get_recipe_specific_option_metadata(urn) def get_matches(self, location, query): query = query.strip().lower() if not query: return self.universal_set() results = set() for urn in self.universal_set(): recipe = self.recipe_from_urn(urn) if query in recipe.get('title', '').lower() or \ query in recipe.get('description', '').lower(): results.add(urn) return results def search(self, query): results = [] try: query = str(query).strip() if query: results = self.parse(query) if not results: results = None except ParseException: results = [] self.do_refresh(restrict_to_urns=results) self.searched.emit(True) def columnCount(self, parent): return 1 def data(self, index, role): if not index.isValid(): return None item = index.internalPointer() return item.data(role) def headerData(self, *args): return None def flags(self, index): if not index.isValid(): return Qt.ItemFlag.ItemIsEnabled|Qt.ItemFlag.ItemIsSelectable item = index.internalPointer() return item.flags() def resort(self): self.do_refresh() def index(self, row, column, parent): if not self.hasIndex(row, column, parent): return QModelIndex() if not parent.isValid(): parent_item = self.root else: parent_item = parent.internalPointer() try: child_item = parent_item.children[row] except IndexError: return QModelIndex() ans = self.createIndex(row, column, child_item) return ans def parent(self, index): if not index.isValid(): return QModelIndex() child_item = index.internalPointer() parent_item = child_item.parent if parent_item is self.root or parent_item is None: return QModelIndex() ans = self.createIndex(parent_item.row(), 0, parent_item) return ans def rowCount(self, parent): if parent.column() > 0: return 0 if not parent.isValid(): parent_item = self.root else: parent_item = parent.internalPointer() return len(parent_item.children) def update_recipe_schedule(self, urn, schedule_type, schedule, add_title_tag=True, custom_tags=[]): recipe = self.recipe_from_urn(urn) self.scheduler_config.schedule_recipe(recipe, schedule_type, schedule, add_title_tag=add_title_tag, custom_tags=custom_tags) def update_last_downloaded(self, urn): self.scheduler_config.update_last_downloaded(urn) def set_account_info(self, urn, un, pw): self.scheduler_config.set_account_info(urn, un, pw) def clear_account_info(self, urn): self.scheduler_config.clear_account_info(urn) def get_account_info(self, urn): return self.scheduler_config.get_account_info(urn) def get_schedule_info(self, urn): return self.scheduler_config.get_schedule_info(urn) def un_schedule_recipe(self, urn): self.scheduler_config.un_schedule_recipe(urn) def schedule_recipe(self, urn, sched_type, schedule): self.scheduler_config.schedule_recipe(self.recipe_from_urn(urn), sched_type, schedule) def customize_recipe(self, urn, val): self.scheduler_config.customize_recipe(urn, val) def get_to_be_downloaded_recipes(self): ans = self.scheduler_config.get_to_be_downloaded_recipes() ans2 = [x for x in ans if self.get_recipe(x, download=False) is not None] for x in set(ans) - set(ans2): self.un_schedule_recipe(x) return ans2 def scheduled_urns(self): ans = [] with self.scheduler_config.lock: for recipe in self.scheduler_config.iter_recipes(): ans.append(recipe.get('id')) return ans
15,099
Python
.py
363
31.548209
114
0.605227
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,816
break_iterator.py
kovidgoyal_calibre/src/calibre/spell/break_iterator.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from threading import Lock from calibre.utils.icu import _icu from calibre.utils.localization import lang_as_iso639_1 _iterators = {} _sentence_iterators = {} _lock = Lock() def get_iterator(lang): it = _iterators.get(lang) if it is None: it = _iterators[lang] = _icu.BreakIterator(_icu.UBRK_WORD, lang_as_iso639_1(lang) or lang) return it def get_sentence_iterator(lang): it = _sentence_iterators.get(lang) if it is None: it = _sentence_iterators[lang] = _icu.BreakIterator(_icu.UBRK_SENTENCE, lang_as_iso639_1(lang) or lang) return it def split_into_words(text, lang='en'): with _lock: it = get_iterator(lang) it.set_text(text) return [text[p:p+s] for p, s in it.split2()] def split_into_words_and_positions(text, lang='en'): with _lock: it = get_iterator(lang) it.set_text(text) return it.split2() def sentence_positions(text, lang='en'): with _lock: it = get_sentence_iterator(lang) it.set_text(text) return it.split2() def split_into_sentences(text, lang='en'): with _lock: it = get_sentence_iterator(lang) it.set_text(text) return tuple(text[p:p+s] for p, s in it.split2()) def index_of(needle, haystack, lang='en'): with _lock: it = get_iterator(lang) it.set_text(haystack) return it.index(needle) def count_words(text, lang='en'): with _lock: it = get_iterator(lang) it.set_text(text) return it.count_words() def split_long_sentences(sentence: str, offset: int, lang: str = 'en', limit: int = 2048): if len(sentence) <= limit: yield offset, sentence return buf, total, start_at = [], 0, 0 def a(s, e): nonlocal total, start_at t = sentence[s:e] if not buf: start_at = s buf.append(t) total += len(t) for start, length in split_into_words_and_positions(sentence, lang): a(start, start + length) if total >= limit: yield offset + start_at, ' '.join(buf) buf, total = [], 0 if buf: yield offset + start_at, ' '.join(buf) PARAGRAPH_SEPARATOR = '\u2029' def split_into_sentences_for_tts_embed( text: str, lang: str = 'en', ): import re def sub(m): return PARAGRAPH_SEPARATOR + ' ' * (len(m.group()) - 1) text = re.sub(r'\n{2,}', sub, text.replace('\r', ' ')).replace('\n', ' ') yield from sentence_positions(text, lang) def split_into_sentences_for_tts( text: str, lang: str = 'en', min_sentence_length: int = 32, max_sentence_length: int = 1024, PARAGRAPH_SEPARATOR: str = PARAGRAPH_SEPARATOR): import re def sub(m): return PARAGRAPH_SEPARATOR + ' ' * (len(m.group()) - 1) text = re.sub(r'\n{2,}', sub, text.replace('\r', ' ')).replace('\n', ' ') pending_start, pending_sentence = 0, '' for start, length in sentence_positions(text, lang): end = start + length sentence = text[start:end].rstrip().replace('\n', ' ').strip() if not sentence: continue if len(sentence) < min_sentence_length and text[end-1] != PARAGRAPH_SEPARATOR: if pending_sentence: pending_sentence += ' ' + sentence if len(pending_sentence) >= min_sentence_length: yield pending_start, pending_sentence pending_start, pending_sentence = 0, '' else: pending_start, pending_sentence = start, sentence continue for start, sentence in split_long_sentences(sentence, start, lang, limit=max_sentence_length): if pending_sentence: sentence = pending_sentence + ' ' + sentence start = pending_start pending_start, pending_sentence = 0, '' yield start, sentence if pending_sentence: yield pending_start, pending_sentence
4,087
Python
.py
106
31
145
0.605063
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,817
__init__.py
kovidgoyal_calibre/src/calibre/spell/__init__.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from collections import namedtuple from calibre.utils.localization import canonicalize_lang, load_iso3166 DictionaryLocale = namedtuple('DictionaryLocale', 'langcode countrycode') def get_codes(): data = load_iso3166() return data['codes'], data['three_map'] def parse_lang_code(raw): raw = raw or '' parts = raw.replace('_', '-').split('-') lc = canonicalize_lang(parts[0]) if lc is None: raise ValueError('Invalid language code: %r' % raw) cc = None for sc in ['Cyrl', 'Latn']: if sc in parts: parts.remove(sc) if len(parts) > 1: ccodes, ccodemap = get_codes()[:2] q = parts[1].upper() if q in ccodes: cc = q else: cc = ccodemap.get(q, None) return DictionaryLocale(lc, cc)
916
Python
.py
27
28
73
0.622727
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,818
import_from.py
kovidgoyal_calibre/src/calibre/spell/import_from.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' import codecs import glob import os import re import sys import tempfile from lxml import etree from calibre import browser from calibre.constants import config_dir from calibre.utils.resources import get_path as P from calibre.utils.xml_parse import safe_xml_fromstring from calibre.utils.zipfile import ZipFile from polyglot.builtins import iteritems NS_MAP = { 'oor': "http://openoffice.org/2001/registry", 'xs': "http://www.w3.org/2001/XMLSchema", 'manifest': 'http://openoffice.org/2001/manifest', } def XPath(x): return etree.XPath(x, namespaces=NS_MAP) BUILTIN_LOCALES = {'en-US', 'en-GB', 'es-ES'} ONLINE_DICTIONARY_BASE_URL = 'https://raw.githubusercontent.com/LibreOffice/dictionaries/master/' def parse_xcu(raw, origin='%origin%'): ' Get the dictionary and affix file names as well as supported locales for each dictionary ' ans = {} root = safe_xml_fromstring(raw) for node in XPath('//prop[@oor:name="Format"]/value[text()="DICT_SPELL"]/../..')(root): value = XPath('descendant::prop[@oor:name="Locations"]/value')(node) if len(value[0]) == 0: # The value node has no children, use its text paths = ''.join(XPath('descendant::prop[@oor:name="Locations"]/value/text()')(node)).replace('%origin%', origin).split() else: # Use the text of the value nodes children paths = [c.text.replace('%origin%', origin) for v in value for c in v.iterchildren('*') if c.text] aff, dic = paths if paths[0].endswith('.aff') else reversed(paths) locales = ''.join(XPath('descendant::prop[@oor:name="Locales"]/value/text()')(node)).split() if not locales: locales = [str(item) for item in XPath('descendant::prop[@oor:name="Locales"]/value/it/text()')(node)] ans[(dic, aff)] = locales return ans def convert_to_utf8(dic_data, aff_data, errors='strict'): m = re.search(br'^SET\s+(\S+)$', aff_data[:2048], flags=re.MULTILINE) if m is not None: enc = m.group(1).decode('ascii', 'replace') if enc.upper() not in ('UTF-8', 'UTF8'): try: codecs.lookup(enc) except LookupError: pass else: aff_data = aff_data[:m.start()] + b'SET UTF-8' + aff_data[m.end():] aff_data = aff_data.decode(enc, errors).encode('utf-8') dic_data = dic_data.decode(enc, errors).encode('utf-8') return dic_data, aff_data def import_from_libreoffice_source_tree(source_path): dictionaries = {} for x in glob.glob(os.path.join(source_path, '*', 'dictionaries.xcu')): origin = os.path.dirname(x) with open(x, 'rb') as f: dictionaries.update(parse_xcu(f.read(), origin)) base = P('dictionaries', allow_user_override=False) want_locales = set(BUILTIN_LOCALES) for (dic, aff), locales in iteritems(dictionaries): c = set(locales) & want_locales if c: locale = tuple(c)[0] want_locales.discard(locale) dest = os.path.join(base, locale) if not os.path.exists(dest): os.makedirs(dest) with open(dic, 'rb') as df, open(aff, 'rb') as af: dd, ad = convert_to_utf8(df.read(), af.read()) for src, raw in ((dic, dd), (aff, ad)): with open(os.path.join(dest, locale + os.path.splitext(src)[1]), 'wb') as df: df.write(raw) with open(os.path.join(dest, 'locales'), 'wb') as f: locales.sort(key=lambda x: (0, x) if x == locale else (1, x)) f.write(('\n'.join(locales)).encode('utf-8')) if want_locales: raise Exception('Failed to find dictionaries for some wanted locales: %s' % want_locales) def fill_country_code(x): return {'lt':'lt_LT'}.get(x, x) def uniq(vals, kmap=lambda x:x): ''' Remove all duplicates from vals, while preserving order. kmap must be a callable that returns a hashable value for every item in vals ''' vals = vals or () lvals = (kmap(x) for x in vals) seen = set() seen_add = seen.add return tuple(x for x, k in zip(vals, lvals) if k not in seen and not seen_add(k)) def _import_from_virtual_directory(read_file_func, name, dest_dir=None, prefix='dic-'): from calibre.spell.dictionary import parse_lang_code dest_dir = dest_dir or os.path.join(config_dir, 'dictionaries') if not os.path.exists(dest_dir): os.makedirs(dest_dir) num = 0 root = safe_xml_fromstring(read_file_func('META-INF/manifest.xml')) xcu = XPath('//manifest:file-entry[@manifest:media-type="application/vnd.sun.star.configuration-data"]')(root)[0].get( '{%s}full-path' % NS_MAP['manifest']) for (dic, aff), locales in iteritems(parse_xcu(read_file_func(xcu), origin='')): dic, aff = dic.lstrip('/'), aff.lstrip('/') d = tempfile.mkdtemp(prefix=prefix, dir=dest_dir) locales = uniq([x for x in map(fill_country_code, locales) if parse_lang_code(x).countrycode]) if not locales: continue metadata = [name] + list(locales) with open(os.path.join(d, 'locales'), 'wb') as f: f.write(('\n'.join(metadata)).encode('utf-8')) dd, ad = convert_to_utf8(read_file_func(dic), read_file_func(aff)) with open(os.path.join(d, '%s.dic' % locales[0]), 'wb') as f: f.write(dd) with open(os.path.join(d, '%s.aff' % locales[0]), 'wb') as f: f.write(ad) num += 1 return num def import_from_oxt(source_path, name, dest_dir=None, prefix='dic-'): with ZipFile(source_path) as zf: def read_file(key): try: return zf.open(key).read() except KeyError: # Some dictionaries apparently put the xcu in a sub-directory # and incorrectly make paths relative to that directory instead # of the root, for example: # http://extensions.libreoffice.org/extension-center/italian-dictionary-thesaurus-hyphenation-patterns/releases/4.1/dict-it.oxt while key.startswith('../'): key = key[3:] return zf.open(key.lstrip('/')).read() return _import_from_virtual_directory(read_file, name, dest_dir=dest_dir, prefix=prefix) def import_from_online(directory, name, dest_dir=None, prefix='dic-'): br = browser(timeout=30) def read_file(key): try: rp = br.open('/'.join((ONLINE_DICTIONARY_BASE_URL, directory, key))) return rp.read() except Exception as err: if getattr(err, 'code', -1) != 404: raise # Some dictionaries apparently put the dic and aff file in a # sub-directory dictionaries and incorrectly make paths relative # to that directory instead of the root, for example: # https://github.com/LibreOffice/dictionaries/tree/master/ca rp = br.open('/'.join((ONLINE_DICTIONARY_BASE_URL, directory, 'dictionaries', key))) return rp.read() return _import_from_virtual_directory(read_file, name, dest_dir=dest_dir, prefix=prefix) if __name__ == '__main__': import_from_libreoffice_source_tree(sys.argv[-1])
7,432
Python
.py
150
41.006667
143
0.617497
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,819
dictionary.py
kovidgoyal_calibre/src/calibre/spell/dictionary.py
#!/usr/bin/env python # License: GPLv3 Copyright: 2014, Kovid Goyal <kovid at kovidgoyal.net> import glob import json import os import re import shutil import sys from collections import defaultdict, namedtuple from functools import partial from itertools import chain from calibre import prints from calibre.constants import config_dir, filesystem_encoding, iswindows from calibre.spell import parse_lang_code from calibre.utils.config import JSONConfig from calibre.utils.icu import capitalize from calibre.utils.localization import _, get_lang, get_system_locale from calibre.utils.resources import get_path as P from polyglot.builtins import iteritems, itervalues Dictionary = namedtuple('Dictionary', 'primary_locale locales dicpath affpath builtin name id') LoadedDictionary = namedtuple('Dictionary', 'primary_locale locales obj builtin name id') dprefs = JSONConfig('dictionaries/prefs.json') dprefs.defaults['preferred_dictionaries'] = {} dprefs.defaults['preferred_locales'] = {} dprefs.defaults['user_dictionaries'] = [{'name':_('Default'), 'is_active':True, 'words':[]}] not_present = object() class UserDictionary: __slots__ = ('name', 'is_active', 'words') def __init__(self, **kwargs): self.name = kwargs['name'] self.is_active = kwargs['is_active'] self.words = {(w, langcode) for w, langcode in kwargs['words']} def serialize(self): return {'name':self.name, 'is_active': self.is_active, 'words':[ (w, l) for w, l in self.words]} _builtins = _custom = None def builtin_dictionaries(): global _builtins if _builtins is None: dics = [] for lc in glob.glob(os.path.join(P('dictionaries', allow_user_override=False), '*/locales')): with open(lc, 'rb') as lcf: locales = list(filter(None, lcf.read().decode('utf-8').splitlines())) locale = locales[0] base = os.path.dirname(lc) dics.append(Dictionary( parse_lang_code(locale), frozenset(map(parse_lang_code, locales)), os.path.join(base, '%s.dic' % locale), os.path.join(base, '%s.aff' % locale), True, None, None)) _builtins = frozenset(dics) return _builtins def catalog_online_dictionaries(): loaded = json.loads(P('dictionaries/online-catalog.json', allow_user_override=False, data=True)) try: loaded.update(json.loads(P('dictionaries/online-catalog.json', data=True))) except: pass rslt = [] for lang, directory in loaded.items(): rslt.append({'primary_locale':parse_lang_code(lang), 'name':lang,'directory':directory}) return rslt def custom_dictionaries(reread=False): global _custom if _custom is None or reread: dics = [] for lc in glob.glob(os.path.join(config_dir, 'dictionaries', '*/locales')): with open(lc, 'rb') as cdf: locales = list(filter(None, cdf.read().decode('utf-8').splitlines())) try: name, locale, locales = locales[0], locales[1], locales[1:] except IndexError: continue base = os.path.dirname(lc) ploc = parse_lang_code(locale) if ploc.countrycode is None: continue dics.append(Dictionary( ploc, frozenset(filter(lambda x:x.countrycode is not None, map(parse_lang_code, locales))), os.path.join(base, '%s.dic' % locale), os.path.join(base, '%s.aff' % locale), False, name, os.path.basename(base))) _custom = frozenset(dics) return _custom default_en_locale = 'en-US' try: ul = parse_lang_code(get_system_locale() or 'en-US') except ValueError: ul = None if ul is not None and ul.langcode == 'eng' and ul.countrycode in 'GB BS BZ GH IE IN JM NZ TT'.split(): default_en_locale = 'en-' + ul.countrycode default_preferred_locales = {'eng':default_en_locale, 'deu':'de-DE', 'spa':'es-ES', 'fra':'fr-FR'} def best_locale_for_language(langcode): best_locale = dprefs['preferred_locales'].get(langcode, default_preferred_locales.get(langcode, None)) if best_locale is not None: return parse_lang_code(best_locale) def preferred_dictionary(locale): return {parse_lang_code(k):v for k, v in iteritems(dprefs['preferred_dictionaries'])}.get(locale, None) def remove_dictionary(dictionary): if dictionary.builtin: raise ValueError('Cannot remove builtin dictionaries') base = os.path.dirname(dictionary.dicpath) shutil.rmtree(base) dprefs['preferred_dictionaries'] = {k:v for k, v in iteritems(dprefs['preferred_dictionaries']) if v != dictionary.id} def rename_dictionary(dictionary, name): lf = os.path.join(os.path.dirname(dictionary.dicpath), 'locales') with open(lf, 'r+b') as f: lines = f.read().splitlines() lines[:1] = [name.encode('utf-8')] f.seek(0), f.truncate(), f.write(b'\n'.join(lines)) custom_dictionaries(reread=True) def get_dictionary(locale, exact_match=False): preferred = preferred_dictionary(locale) # First find all dictionaries that match locale exactly exact_matches = {} for collection in (custom_dictionaries(), builtin_dictionaries()): for d in collection: if d.primary_locale == locale: exact_matches[d.id] = d for d in collection: for q in d.locales: if q == locale and d.id not in exact_matches: exact_matches[d.id] = d # If the user has specified a preferred dictionary for this locale, use it, # otherwise, if a builtin dictionary exists, use that if preferred in exact_matches: return exact_matches[preferred] # Return one of the exactly matching dictionaries, preferring user # installed to builtin ones for k in sorted(exact_matches, key=lambda x: (1, None) if x is None else (0, x)): return exact_matches[k] if exact_match: return # No dictionary matched the locale exactly, we will now fallback to # matching only on language. First see if a dictionary matching the # preferred locale for the language exists. best_locale = best_locale_for_language(locale.langcode) if best_locale is not None: ans = get_dictionary(best_locale, exact_match=True) if ans is not None: return ans # Now just return any dictionary that matches the language, preferring user # installed ones to builtin ones for collection in (custom_dictionaries(), builtin_dictionaries()): for d in sorted(collection, key=lambda d: d.name or ''): if d.primary_locale.langcode == locale.langcode: return d def load_dictionary(dictionary): from calibre_extensions import hunspell def fix_path(path): if isinstance(path, bytes): path = path.decode(filesystem_encoding) path = os.path.abspath(path) if iswindows: path = fr'\\?\{path}' return path obj = hunspell.Dictionary(fix_path(dictionary.dicpath), fix_path(dictionary.affpath)) return LoadedDictionary(dictionary.primary_locale, dictionary.locales, obj, dictionary.builtin, dictionary.name, dictionary.id) class Dictionaries: def __init__(self): self.remove_hyphenation = re.compile('[\u2010-]+') self.negative_pat = re.compile(r'-[.\d+]') self.fix_punctuation_pat = re.compile(r'''[:.]''') self.dictionaries = {} self.word_cache = {} self.ignored_words = set() self.added_user_words = {} try: self.default_locale = parse_lang_code(get_lang()) except ValueError: self.default_locale = parse_lang_code('en-US') self.ui_locale = self.default_locale def initialize(self, force=False): if force or not hasattr(self, 'active_user_dictionaries'): self.read_user_dictionaries() def clear_caches(self): self.dictionaries.clear(), self.word_cache.clear() def clear_ignored(self): self.ignored_words.clear() def dictionary_for_locale(self, locale): ans = self.dictionaries.get(locale, not_present) if ans is not_present: ans = get_dictionary(locale) if ans is not None: ans = load_dictionary(ans) for ud in self.active_user_dictionaries: for word, langcode in ud.words: if langcode == locale.langcode: try: ans.obj.add(word) except Exception: # not critical since all it means is that the word won't show up in suggestions prints(f'Failed to add the word {word!r} to the dictionary for {locale}', file=sys.stderr) self.dictionaries[locale] = ans return ans def ignore_word(self, word, locale): self.ignored_words.add((word, locale.langcode)) self.word_cache[(word, locale)] = True def unignore_word(self, word, locale): self.ignored_words.discard((word, locale.langcode)) self.word_cache.pop((word, locale), None) def is_word_ignored(self, word, locale): return (word, locale.langcode) in self.ignored_words @property def all_user_dictionaries(self): return chain(self.active_user_dictionaries, self.inactive_user_dictionaries) def user_dictionary(self, name): for ud in self.all_user_dictionaries: if ud.name == name: return ud def read_user_dictionaries(self): self.active_user_dictionaries = [] self.inactive_user_dictionaries = [] for d in dprefs['user_dictionaries'] or dprefs.defaults['user_dictionaries']: d = UserDictionary(**d) (self.active_user_dictionaries if d.is_active else self.inactive_user_dictionaries).append(d) def mark_user_dictionary_as_active(self, name, is_active=True): d = self.user_dictionary(name) if d is not None: d.is_active = is_active self.save_user_dictionaries() return True return False def save_user_dictionaries(self): dprefs['user_dictionaries'] = [d.serialize() for d in self.all_user_dictionaries] def add_user_words(self, words, langcode): for d in itervalues(self.dictionaries): if d and getattr(d.primary_locale, 'langcode', None) == langcode: for word in words: d.obj.add(word) def remove_user_words(self, words, langcode): for d in itervalues(self.dictionaries): if d and d.primary_locale.langcode == langcode: for word in words: d.obj.remove(word) def add_to_user_dictionary(self, name, word, locale): ud = self.user_dictionary(name) if ud is None: raise ValueError('Cannot add to the dictionary named: %s as no such dictionary exists' % name) wl = len(ud.words) if isinstance(word, (set, frozenset)): ud.words |= word self.add_user_words({x[0] for x in word}, locale.langcode) else: ud.words.add((word, locale.langcode)) self.add_user_words((word,), locale.langcode) if len(ud.words) > wl: self.save_user_dictionaries() try: self.word_cache.pop((word, locale), None) except TypeError: pass # word is a set, ignore return True return False def remove_from_user_dictionaries(self, word, locale): key = (word, locale.langcode) changed = False for ud in self.active_user_dictionaries: if key in ud.words: changed = True ud.words.discard(key) if changed: self.word_cache.pop((word, locale), None) self.save_user_dictionaries() self.remove_user_words((word,), locale.langcode) return changed def remove_from_user_dictionary(self, name, words): changed = False removals = defaultdict(set) keys = [(w, l.langcode) for w, l in words] for d in self.all_user_dictionaries: if d.name == name: for key in keys: if key in d.words: d.words.discard(key) removals[key[1]].add(key[0]) changed = True if changed: for key in words: self.word_cache.pop(key, None) for langcode, words in iteritems(removals): self.remove_user_words(words, langcode) self.save_user_dictionaries() return changed def word_in_user_dictionary(self, word, locale): key = (word, locale.langcode) for ud in self.active_user_dictionaries: if key in ud.words: return ud.name def create_user_dictionary(self, name): if name in {d.name for d in self.all_user_dictionaries}: raise ValueError('A dictionary named %s already exists' % name) d = UserDictionary(name=name, is_active=True, words=()) self.active_user_dictionaries.append(d) self.save_user_dictionaries() def remove_user_dictionary(self, name): changed = False for x in (self.active_user_dictionaries, self.inactive_user_dictionaries): for d in tuple(x): if d.name == name: x.remove(d) changed = True if changed: self.save_user_dictionaries() self.clear_caches() return changed def rename_user_dictionary(self, name, new_name): changed = False for d in self.all_user_dictionaries: if d.name == name: d.name = new_name changed = True if changed: self.save_user_dictionaries() return changed def recognized(self, word, locale=None): locale = locale or self.default_locale key = (word, locale) ans = self.word_cache.get(key, None) if ans is None: lkey = (word, locale.langcode) ans = False if lkey in self.ignored_words: ans = True else: for ud in self.active_user_dictionaries: if lkey in ud.words: ans = True break else: d = self.dictionary_for_locale(locale) if d is not None: try: ans = d.obj.recognized(word.replace('\u2010', '-')) except ValueError: pass else: ans = True if ans is False and self.negative_pat.match(word) is not None: ans = True self.word_cache[key] = ans return ans def suggestions(self, word, locale=None): locale = locale or self.default_locale d = self.dictionary_for_locale(locale) has_unicode_hyphen = '\u2010' in word ans = () def add_suggestion(w, ans): return (w,) + tuple(x for x in ans if x != w) if d is not None: try: ans = d.obj.suggest(str(word).replace('\u2010', '-')) except ValueError: pass else: dehyphenated_word = self.remove_hyphenation.sub('', word) if len(dehyphenated_word) != len(word) and self.recognized(dehyphenated_word, locale): # Ensure the de-hyphenated word is present and is the first suggestion ans = add_suggestion(dehyphenated_word, ans) else: m = self.fix_punctuation_pat.search(word) if m is not None: w1, w2 = word[:m.start()], word[m.end():] if self.recognized(w1) and self.recognized(w2): fw = w1 + m.group() + ' ' + w2 ans = add_suggestion(fw, ans) if capitalize(w2) != w2: fw = w1 + m.group() + ' ' + capitalize(w2) ans = add_suggestion(fw, ans) if has_unicode_hyphen: ans = tuple(w.replace('-', '\u2010') for w in ans) return ans def build_test(): dictionaries = Dictionaries() dictionaries.initialize() eng = parse_lang_code('en') if not dictionaries.recognized('recognized', locale=eng): raise AssertionError('The word recognized was not recognized') def find_tests(): import unittest class TestDictionaries(unittest.TestCase): def setUp(self): dictionaries = Dictionaries() dictionaries.initialize() eng = parse_lang_code('en-GB') self.recognized = partial(dictionaries.recognized, locale=eng) self.suggestions = partial(dictionaries.suggestions, locale=eng) def ar(self, w): if not self.recognized(w): raise AssertionError('The word %r was not recognized' % w) def test_dictionaries(self): for w in 'recognized one-half one\u2010half'.split(): self.ar(w) d = load_dictionary(get_dictionary(parse_lang_code('es-ES'))).obj self.assertTrue(d.recognized('Ahí')) self.assertIn('one\u2010half', self.suggestions('oone\u2010half')) d = load_dictionary(get_dictionary(parse_lang_code('es'))).obj self.assertIn('adequately', self.suggestions('ade-quately')) self.assertIn('magic. Wand', self.suggestions('magic.wand')) self.assertIn('List', self.suggestions('Lis𝑘t')) return unittest.TestLoader().loadTestsFromTestCase(TestDictionaries) def test(): from calibre.utils.run_tests import run_cli run_cli(find_tests()) if __name__ == '__main__': test()
18,220
Python
.py
402
34.803483
146
0.604318
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,820
errors.py
kovidgoyal_calibre/src/css_selectors/errors.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>' class SelectorError(ValueError): """Common parent for SelectorSyntaxError and ExpressionError""" class SelectorSyntaxError(SelectorError): """Parsing a selector that does not match the grammar.""" class ExpressionError(SelectorError): """Unknown or unsupported selector (eg. pseudo-class)."""
445
Python
.py
10
41.5
67
0.754098
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,821
select.py
kovidgoyal_calibre/src/css_selectors/select.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>' import itertools import re from collections import OrderedDict, defaultdict from functools import wraps from itertools import chain from lxml import etree from css_selectors.errors import ExpressionError from css_selectors.ordered_set import OrderedSet from css_selectors.parser import Element, FunctionalPseudoElement, ascii_lower, parse from polyglot.builtins import iteritems, itervalues PARSE_CACHE_SIZE = 200 parse_cache = OrderedDict() XPATH_CACHE_SIZE = 30 xpath_cache = OrderedDict() # Test that the string is not empty and does not contain whitespace is_non_whitespace = re.compile(r'^[^ \t\r\n\f]+$').match def get_parsed_selector(raw): try: return parse_cache[raw] except KeyError: parse_cache[raw] = ans = parse(raw) if len(parse_cache) > PARSE_CACHE_SIZE: parse_cache.pop(next(iter(parse_cache))) return ans def get_compiled_xpath(expr): try: return xpath_cache[expr] except KeyError: xpath_cache[expr] = ans = etree.XPath(expr) if len(xpath_cache) > XPATH_CACHE_SIZE: xpath_cache.pop(next(iter(xpath_cache))) return ans class AlwaysIn: def __contains__(self, x): return True always_in = AlwaysIn() def trace_wrapper(func): @wraps(func) def trace(*args, **kwargs): targs = args[1:] if args and isinstance(args[0], Select) else args print('Called:', func.__name__, 'with args:', targs, kwargs or '') return func(*args, **kwargs) return trace def normalize_language_tag(tag): """Return a list of normalized combinations for a `BCP 47` language tag. Example: >>> normalize_language_tag('de_AT-1901') ['de-at-1901', 'de-at', 'de-1901', 'de'] """ # normalize: tag = ascii_lower(tag).replace('_','-') # split (except singletons, which mark the following tag as non-standard): tag = re.sub(r'-([a-zA-Z0-9])-', r'-\1_', tag) subtags = [subtag.replace('_', '-') for subtag in tag.split('-')] base_tag = (subtags.pop(0),) taglist = {base_tag[0]} # find all combinations of subtags for n in range(len(subtags), 0, -1): for tags in itertools.combinations(subtags, n): taglist.add('-'.join(base_tag + tags)) return taglist INAPPROPRIATE_PSEUDO_CLASSES = frozenset(( 'active', 'after', 'disabled', 'visited', 'link', 'before', 'focus', 'first-letter', 'enabled', 'first-line', 'hover', 'checked', 'target')) class Select: ''' This class implements CSS Level 3 selectors (http://www.w3.org/TR/css3-selectors) on an lxml tree, with caching for performance. To use: >>> from css_selectors import Select >>> select = Select(root) # Where root is an lxml document >>> print(tuple(select('p.myclass'))) Tags are returned in document order. Note that attribute and tag names are matched case-insensitively. Class and id values are also matched case-insensitively. Also namespaces are ignored (this is for performance of the common case). The UI related selectors are not implemented, such as :enabled, :disabled, :checked, :hover, etc. Similarly, the non-element related selectors such as ::first-line, ::first-letter, ::before, etc. are not implemented. WARNING: This class uses internal caches. You *must not* make any changes to the lxml tree. If you do make some changes, either create a new Select object or call :meth:`invalidate_caches`. This class can be easily sub-classed to work with tree implementations other than lxml. Simply override the methods in the ``Tree Integration`` block below. The caching works by maintaining internal maps from classes/ids/tag names/etc. to node sets. These caches are populated as needed, and used for all subsequent selections. Thus, for best performance you should use the same selector object for finding the matching nodes for multiple queries. Of course, remember not to change the tree in between queries. ''' combinator_mapping = { ' ': 'descendant', '>': 'child', '+': 'direct_adjacent', '~': 'indirect_adjacent', } attribute_operator_mapping = { 'exists': 'exists', '=': 'equals', '~=': 'includes', '|=': 'dashmatch', '^=': 'prefixmatch', '$=': 'suffixmatch', '*=': 'substringmatch', } def __init__(self, root, default_lang=None, ignore_inappropriate_pseudo_classes=False, dispatch_map=None, trace=False): if hasattr(root, 'getroot'): root = root.getroot() self.root = root self.dispatch_map = dispatch_map or default_dispatch_map self.invalidate_caches() self.default_lang = default_lang if trace: self.dispatch_map = {k:trace_wrapper(v) for k, v in iteritems(self.dispatch_map)} if ignore_inappropriate_pseudo_classes: self.ignore_inappropriate_pseudo_classes = INAPPROPRIATE_PSEUDO_CLASSES else: self.ignore_inappropriate_pseudo_classes = frozenset() # External API {{{ def invalidate_caches(self): 'Invalidate all caches. You must call this before using this object if you have made changes to the HTML tree' self._element_map = None self._id_map = None self._class_map = None self._attrib_map = None self._attrib_space_map = None self._lang_map = None self.map_tag_name = ascii_lower if '{' in self.root.tag: def map_tag_name(x): return ascii_lower(x.rpartition('}')[2]) self.map_tag_name = map_tag_name def __call__(self, selector, root=None): ''' Return an iterator over all matching tags, in document order. Normally, all matching tags in the document are returned, is you specify root, then only tags that are root or descendants of root are returned. Note that this can be very expensive if root has a lot of descendants. ''' seen = set() if root is not None: root = frozenset(self.itertag(root)) for parsed_selector in get_parsed_selector(selector): for item in self.iterparsedselector(parsed_selector): if item not in seen and (root is None or item in root): yield item seen.add(item) def has_matches(self, selector, root=None): 'Return True iff selector matches at least one item in the tree' for elem in self(selector, root=root): return True return False # }}} def iterparsedselector(self, parsed_selector): type_name = type(parsed_selector).__name__ try: func = self.dispatch_map[ascii_lower(type_name)] except KeyError: raise ExpressionError('%s is not supported' % type_name) for item in func(self, parsed_selector): yield item @property def element_map(self): if self._element_map is None: self._element_map = em = defaultdict(OrderedSet) for tag in self.itertag(): em[self.map_tag_name(tag.tag)].add(tag) return self._element_map @property def id_map(self): if self._id_map is None: self._id_map = im = defaultdict(OrderedSet) lower = ascii_lower for elem in self.iteridtags(): im[lower(elem.get('id'))].add(elem) return self._id_map @property def class_map(self): if self._class_map is None: self._class_map = cm = defaultdict(OrderedSet) lower = ascii_lower for elem in self.iterclasstags(): for cls in elem.get('class').split(): cm[lower(cls)].add(elem) return self._class_map @property def attrib_map(self): if self._attrib_map is None: self._attrib_map = am = defaultdict(lambda : defaultdict(OrderedSet)) map_attrib_name = ascii_lower if '{' in self.root.tag: def map_attrib_name(x): return ascii_lower(x.rpartition('}')[2]) for tag in self.itertag(): for attr, val in iteritems(tag.attrib): am[map_attrib_name(attr)][val].add(tag) return self._attrib_map @property def attrib_space_map(self): if self._attrib_space_map is None: self._attrib_space_map = am = defaultdict(lambda : defaultdict(OrderedSet)) map_attrib_name = ascii_lower if '{' in self.root.tag: def map_attrib_name(x): return ascii_lower(x.rpartition('}')[2]) for tag in self.itertag(): for attr, val in iteritems(tag.attrib): for v in val.split(): am[map_attrib_name(attr)][v].add(tag) return self._attrib_space_map @property def lang_map(self): if self._lang_map is None: self._lang_map = lm = defaultdict(OrderedSet) dl = normalize_language_tag(self.default_lang) if self.default_lang else None lmap = {tag:dl for tag in self.itertag()} if dl else {} for tag in self.itertag(): lang = None for attr in ('{http://www.w3.org/XML/1998/namespace}lang', 'lang'): lang = tag.get(attr) if lang: lang = normalize_language_tag(lang) for dtag in self.itertag(tag): lmap[dtag] = lang for tag, langs in iteritems(lmap): for lang in langs: lm[lang].add(tag) return self._lang_map # Tree Integration {{{ def itertag(self, tag=None): return (self.root if tag is None else tag).iter('*') def iterdescendants(self, tag=None): return (self.root if tag is None else tag).iterdescendants('*') def iterchildren(self, tag=None): return (self.root if tag is None else tag).iterchildren('*') def itersiblings(self, tag=None, preceding=False): return (self.root if tag is None else tag).itersiblings('*', preceding=preceding) def iteridtags(self): return get_compiled_xpath('//*[@id]')(self.root) def iterclasstags(self): return get_compiled_xpath('//*[@class]')(self.root) def sibling_count(self, child, before=True, same_type=False): ' Return the number of siblings before or after child or raise ValueError if child has no parent. ' parent = child.getparent() if parent is None: raise ValueError('Child has no parent') if same_type: siblings = OrderedSet(child.itersiblings(preceding=before)) return len(self.element_map[self.map_tag_name(child.tag)] & siblings) else: if before: return parent.index(child) return len(parent) - parent.index(child) - 1 def all_sibling_count(self, child, same_type=False): ' Return the number of siblings of child or raise ValueError if child has no parent ' parent = child.getparent() if parent is None: raise ValueError('Child has no parent') if same_type: siblings = OrderedSet(chain(child.itersiblings(preceding=False), child.itersiblings(preceding=True))) return len(self.element_map[self.map_tag_name(child.tag)] & siblings) else: return len(parent) - 1 def is_empty(self, elem): ' Return True iff elem has no child tags and no text content ' for child in elem: # Check for comment/PI nodes with tail text if child.tail: return False return len(tuple(elem.iterchildren('*'))) == 0 and not elem.text # }}} # Combinators {{{ def select_combinedselector(cache, combined): """Translate a combined selector.""" combinator = cache.combinator_mapping[combined.combinator] # Fast path for when the sub-selector is all elements right = None if isinstance(combined.subselector, Element) and ( combined.subselector.element or '*') == '*' else cache.iterparsedselector(combined.subselector) for item in cache.dispatch_map[combinator](cache, cache.iterparsedselector(combined.selector), right): yield item def select_descendant(cache, left, right): """right is a child, grand-child or further descendant of left""" right = always_in if right is None else frozenset(right) for ancestor in left: for descendant in cache.iterdescendants(ancestor): if descendant in right: yield descendant def select_child(cache, left, right): """right is an immediate child of left""" right = always_in if right is None else frozenset(right) for parent in left: for child in cache.iterchildren(parent): if child in right: yield child def select_direct_adjacent(cache, left, right): """right is a sibling immediately after left""" right = always_in if right is None else frozenset(right) for parent in left: for sibling in cache.itersiblings(parent): if sibling in right: yield sibling break def select_indirect_adjacent(cache, left, right): """right is a sibling after left, immediately or not""" right = always_in if right is None else frozenset(right) for parent in left: for sibling in cache.itersiblings(parent): if sibling in right: yield sibling # }}} def select_element(cache, selector): """A type or universal selector.""" element = selector.element if not element or element == '*': for elem in cache.itertag(): yield elem else: for elem in cache.element_map[ascii_lower(element)]: yield elem def select_hash(cache, selector): 'An id selector' items = cache.id_map[ascii_lower(selector.id)] if len(items) > 0: for elem in cache.iterparsedselector(selector.selector): if elem in items: yield elem def select_class(cache, selector): 'A class selector' items = cache.class_map[ascii_lower(selector.class_name)] if items: for elem in cache.iterparsedselector(selector.selector): if elem in items: yield elem def select_negation(cache, selector): 'Implement :not()' exclude = frozenset(cache.iterparsedselector(selector.subselector)) for item in cache.iterparsedselector(selector.selector): if item not in exclude: yield item # Attribute selectors {{{ def select_attrib(cache, selector): operator = cache.attribute_operator_mapping[selector.operator] items = frozenset(cache.dispatch_map[operator](cache, ascii_lower(selector.attrib), selector.value)) for item in cache.iterparsedselector(selector.selector): if item in items: yield item def select_exists(cache, attrib, value=None): for elem_set in itervalues(cache.attrib_map[attrib]): for elem in elem_set: yield elem def select_equals(cache, attrib, value): for elem in cache.attrib_map[attrib][value]: yield elem def select_includes(cache, attrib, value): if is_non_whitespace(value): for elem in cache.attrib_space_map[attrib][value]: yield elem def select_dashmatch(cache, attrib, value): if value: for val, elem_set in iteritems(cache.attrib_map[attrib]): if val == value or val.startswith(value + '-'): for elem in elem_set: yield elem def select_prefixmatch(cache, attrib, value): if value: for val, elem_set in iteritems(cache.attrib_map[attrib]): if val.startswith(value): for elem in elem_set: yield elem def select_suffixmatch(cache, attrib, value): if value: for val, elem_set in iteritems(cache.attrib_map[attrib]): if val.endswith(value): for elem in elem_set: yield elem def select_substringmatch(cache, attrib, value): if value: for val, elem_set in iteritems(cache.attrib_map[attrib]): if value in val: for elem in elem_set: yield elem # }}} # Function selectors {{{ def select_function(cache, function): """Select with a functional pseudo-class.""" fname = function.name.replace('-', '_') try: func = cache.dispatch_map[fname] except KeyError: raise ExpressionError( "The pseudo-class :%s() is unknown" % function.name) if fname == 'lang': items = frozenset(func(cache, function)) for item in cache.iterparsedselector(function.selector): if item in items: yield item else: for item in cache.iterparsedselector(function.selector): if func(cache, function, item): yield item def select_lang(cache, function): ' Implement :lang() ' if function.argument_types() not in (['STRING'], ['IDENT']): raise ExpressionError("Expected a single string or ident for :lang(), got %r" % function.arguments) lang = function.arguments[0].value if lang: lang = ascii_lower(lang) lp = lang + '-' for tlang, elem_set in iteritems(cache.lang_map): if tlang == lang or (tlang is not None and tlang.startswith(lp)): for elem in elem_set: yield elem def select_nth_child(cache, function, elem): ' Implement :nth-child() ' a, b = function.parsed_arguments try: num = cache.sibling_count(elem) + 1 except ValueError: return False if a == 0: return num == b n = (num - b) / a return n.is_integer() and n > -1 def select_nth_last_child(cache, function, elem): ' Implement :nth-last-child() ' a, b = function.parsed_arguments try: num = cache.sibling_count(elem, before=False) + 1 except ValueError: return False if a == 0: return num == b n = (num - b) / a return n.is_integer() and n > -1 def select_nth_of_type(cache, function, elem): ' Implement :nth-of-type() ' a, b = function.parsed_arguments try: num = cache.sibling_count(elem, same_type=True) + 1 except ValueError: return False if a == 0: return num == b n = (num - b) / a return n.is_integer() and n > -1 def select_nth_last_of_type(cache, function, elem): ' Implement :nth-last-of-type() ' a, b = function.parsed_arguments try: num = cache.sibling_count(elem, before=False, same_type=True) + 1 except ValueError: return False if a == 0: return num == b n = (num - b) / a return n.is_integer() and n > -1 # }}} # Pseudo elements {{{ def pseudo_func(f): f.is_pseudo = True return f @pseudo_func def allow_all(cache, item): return True def get_func_for_pseudo(cache, ident): try: func = cache.dispatch_map[ident.replace('-', '_')] except KeyError: if ident in cache.ignore_inappropriate_pseudo_classes: func = allow_all else: raise ExpressionError( "The pseudo-class :%s is not supported" % ident) try: func.is_pseudo except AttributeError: raise ExpressionError( "The pseudo-class :%s is invalid" % ident) return func def select_selector(cache, selector): if selector.pseudo_element is None: for item in cache.iterparsedselector(selector.parsed_tree): yield item return if isinstance(selector.pseudo_element, FunctionalPseudoElement): raise ExpressionError( "The pseudo-element ::%s is not supported" % selector.pseudo_element.name) func = get_func_for_pseudo(cache, selector.pseudo_element) for item in cache.iterparsedselector(selector.parsed_tree): if func(cache, item): yield item def select_pseudo(cache, pseudo): func = get_func_for_pseudo(cache, pseudo.ident) if func is select_root: yield cache.root return for item in cache.iterparsedselector(pseudo.selector): if func(cache, item): yield item @pseudo_func def select_root(cache, elem): return elem is cache.root @pseudo_func def select_first_child(cache, elem): try: return cache.sibling_count(elem) == 0 except ValueError: return False @pseudo_func def select_last_child(cache, elem): try: return cache.sibling_count(elem, before=False) == 0 except ValueError: return False @pseudo_func def select_only_child(cache, elem): try: return cache.all_sibling_count(elem) == 0 except ValueError: return False @pseudo_func def select_first_of_type(cache, elem): try: return cache.sibling_count(elem, same_type=True) == 0 except ValueError: return False @pseudo_func def select_last_of_type(cache, elem): try: return cache.sibling_count(elem, before=False, same_type=True) == 0 except ValueError: return False @pseudo_func def select_only_of_type(cache, elem): try: return cache.all_sibling_count(elem, same_type=True) == 0 except ValueError: return False @pseudo_func def select_empty(cache, elem): return cache.is_empty(elem) # }}} default_dispatch_map = {name.partition('_')[2]:obj for name, obj in globals().items() if name.startswith('select_') and callable(obj)} if __name__ == '__main__': from pprint import pprint root = etree.fromstring( '<body xmlns="xxx" xml:lang="en"><p id="p" class="one two" lang="fr"><a id="a"/><b/><c/><d/></p></body>', parser=etree.XMLParser(recover=True, no_network=True, resolve_entities=False)) select = Select(root, ignore_inappropriate_pseudo_classes=True, trace=True) pprint(list(select('p:disabled')))
22,345
Python
.py
553
32.381555
144
0.632488
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,822
ordered_set.py
kovidgoyal_calibre/src/css_selectors/ordered_set.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>' import collections.abc from polyglot.builtins import string_or_bytes SLICE_ALL = slice(None) def is_iterable(obj): """ Are we being asked to look up a list of things, instead of a single thing? We check for the `__iter__` attribute so that this can cover types that don't have to be known by this module, such as NumPy arrays. Strings, however, should be considered as atomic values to look up, not iterables. """ return hasattr(obj, '__iter__') and not isinstance(obj, string_or_bytes) class OrderedSet(collections.abc.MutableSet): """ An OrderedSet is a custom MutableSet that remembers its order, so that every entry has an index that can be looked up. """ def __init__(self, iterable=None): self.items = [] self.map = {} if iterable is not None: for item in iterable: idx = self.map.get(item) if idx is None: self.map[item] = len(self.items) self.items.append(item) def __len__(self): return len(self.items) def __getitem__(self, index): """ Get the item at a given index. If `index` is a slice, you will get back that slice of items. If it's the slice [:], exactly the same object is returned. (If you want an independent copy of an OrderedSet, use `OrderedSet.copy()`.) If `index` is an iterable, you'll get the OrderedSet of items corresponding to those indices. This is similar to NumPy's "fancy indexing". """ if index == SLICE_ALL: return self elif hasattr(index, '__index__') or isinstance(index, slice): result = self.items[index] if isinstance(result, list): return OrderedSet(result) else: return result elif is_iterable(index): return OrderedSet([self.items[i] for i in index]) else: raise TypeError("Don't know how to index an OrderedSet by %r" % index) def copy(self): return OrderedSet(self) def __getstate__(self): return tuple(self) def __setstate__(self, state): self.__init__(state) def __contains__(self, key): return key in self.map def add(self, key): """ Add `key` as an item to this OrderedSet, then return its index. If `key` is already in the OrderedSet, return the index it already had. """ index = self.map.get(key) if index is None: self.map[key] = index = len(self.items) self.items.append(key) return index def index(self, key): """ Get the index of a given entry, raising an IndexError if it's not present. `key` can be an iterable of entries that is not a string, in which case this returns a list of indices. """ if is_iterable(key): return [self.index(subkey) for subkey in key] return self.map[key] def discard(self, key): index = self.map.get(key) if index is not None: self.items.pop(index) for item in self.items[index:]: self.map[item] -= 1 return True return False def __iter__(self): return iter(self.items) def __reversed__(self): return reversed(self.items) def __repr__(self): if not self: return '%s()' % (self.__class__.__name__,) return '%s(%r)' % (self.__class__.__name__, list(self)) def __eq__(self, other): if isinstance(other, OrderedSet): return len(self) == len(other) and self.items == other.items try: return type(other)(self.map) == other except TypeError: return False
4,014
Python
.py
107
28.560748
79
0.584794
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,823
__init__.py
kovidgoyal_calibre/src/css_selectors/__init__.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>' from css_selectors.errors import ExpressionError, SelectorError, SelectorSyntaxError from css_selectors.parser import parse from css_selectors.select import INAPPROPRIATE_PSEUDO_CLASSES, Select __all__ = ['parse', 'Select', 'INAPPROPRIATE_PSEUDO_CLASSES', 'SelectorError', 'SelectorSyntaxError', 'ExpressionError']
451
Python
.py
8
54.875
120
0.783599
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,824
tests.py
kovidgoyal_calibre/src/css_selectors/tests.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2015, Kovid Goyal <kovid at kovidgoyal.net>' import argparse import sys import unittest from lxml import etree, html from css_selectors.errors import ExpressionError, SelectorSyntaxError from css_selectors.parser import parse, tokenize from css_selectors.select import Select class TestCSSSelectors(unittest.TestCase): # Test data {{{ HTML_IDS = ''' <html id="html"><head> <link id="link-href" href="foo" /> <link id="link-nohref" /> </head><body> <div id="outer-div"> <a id="name-anchor" name="foo"></a> <a id="tag-anchor" rel="tag" href="http://localhost/foo">link</a> <a id="nofollow-anchor" rel="nofollow" href="https://example.org"> link</a> <ol id="first-ol" class="a b c"> <li id="first-li">content</li> <li id="second-li" lang="En-us"> <div id="li-div"> </div> </li> <li id="third-li" class="ab c"></li> <li id="fourth-li" class="ab c"></li> <li id="fifth-li"></li> <li id="sixth-li"></li> <li id="seventh-li"> </li> </ol> <p id="paragraph"> <b id="p-b">hi</b> <em id="p-em">there</em> <b id="p-b2">guy</b> <input type="checkbox" id="checkbox-unchecked" /> <input type="checkbox" id="checkbox-disabled" disabled="" /> <input type="text" id="text-checked" checked="checked" /> <input type="hidden" /> <input type="hidden" disabled="disabled" /> <input type="checkbox" id="checkbox-checked" checked="checked" /> <input type="checkbox" id="checkbox-disabled-checked" disabled="disabled" checked="checked" /> <fieldset id="fieldset" disabled="disabled"> <input type="checkbox" id="checkbox-fieldset-disabled" /> <input type="hidden" /> </fieldset> </p> <ol id="second-ol"> </ol> <map name="dummymap"> <area shape="circle" coords="200,250,25" href="foo.html" id="area-href" /> <area shape="default" id="area-nohref" /> </map> </div> <div id="foobar-div" foobar="ab bc cde"><span id="foobar-span"></span></div> </body></html> ''' HTML_SHAKESPEARE = ''' <!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> <html xmlns="http://www.w3.org/1999/xhtml" xml:lang="en" lang="en" debug="true"> <head> <meta http-equiv="Content-Type" content="text/html; charset=utf-8"/> </head> <body> <div id="test"> <div class="dialog"> <h2>As You Like It</h2> <div id="playwright"> by William Shakespeare </div> <div class="dialog scene thirdClass" id="scene1"> <h3>ACT I, SCENE III. A room in the palace.</h3> <div class="dialog"> <div class="direction">Enter CELIA and ROSALIND</div> </div> <div id="speech1" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.1">Why, cousin! why, Rosalind! Cupid have mercy! not a word?</div> </div> <div id="speech2" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.2">Not one to throw at a dog.</div> </div> <div id="speech3" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.3">No, thy words are too precious to be cast away upon</div> <div id="scene1.3.4">curs; throw some of them at me; come, lame me with reasons.</div> </div> <div id="speech4" class="character">ROSALIND</div> <div id="speech5" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.8">But is all this for your father?</div> </div> <div class="dialog"> <div id="scene1.3.5">Then there were two cousins laid up; when the one</div> <div id="scene1.3.6">should be lamed with reasons and the other mad</div> <div id="scene1.3.7">without any.</div> </div> <div id="speech6" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.9">No, some of it is for my child's father. O, how</div> <div id="scene1.3.10">full of briers is this working-day world!</div> </div> <div id="speech7" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.11">They are but burs, cousin, thrown upon thee in</div> <div id="scene1.3.12">holiday foolery: if we walk not in the trodden</div> <div id="scene1.3.13">paths our very petticoats will catch them.</div> </div> <div id="speech8" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.14">I could shake them off my coat: these burs are in my heart.</div> </div> <div id="speech9" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.15">Hem them away.</div> </div> <div id="speech10" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.16">I would try, if I could cry 'hem' and have him.</div> </div> <div id="speech11" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.17">Come, come, wrestle with thy affections.</div> </div> <div id="speech12" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.18">O, they take the part of a better wrestler than myself!</div> </div> <div id="speech13" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.19">O, a good wish upon you! you will try in time, in</div> <div id="scene1.3.20">despite of a fall. But, turning these jests out of</div> <div id="scene1.3.21">service, let us talk in good earnest: is it</div> <div id="scene1.3.22">possible, on such a sudden, you should fall into so</div> <div id="scene1.3.23">strong a liking with old Sir Rowland's youngest son?</div> </div> <div id="speech14" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.24">The duke my father loved his father dearly.</div> </div> <div id="speech15" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.25">Doth it therefore ensue that you should love his son</div> <div id="scene1.3.26">dearly? By this kind of chase, I should hate him,</div> <div id="scene1.3.27">for my father hated his father dearly; yet I hate</div> <div id="scene1.3.28">not Orlando.</div> </div> <div id="speech16" class="character">ROSALIND</div> <div title="wtf" class="dialog"> <div id="scene1.3.29">No, faith, hate him not, for my sake.</div> </div> <div id="speech17" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.30">Why should I not? doth he not deserve well?</div> </div> <div id="speech18" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.31">Let me love him for that, and do you love him</div> <div id="scene1.3.32">because I do. Look, here comes the duke.</div> </div> <div id="speech19" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.33">With his eyes full of anger.</div> <div class="direction">Enter DUKE FREDERICK, with Lords</div> </div> <div id="speech20" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.34">Mistress, dispatch you with your safest haste</div> <div id="scene1.3.35">And get you from our court.</div> </div> <div id="speech21" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.36">Me, uncle?</div> </div> <div id="speech22" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.37">You, cousin</div> <div id="scene1.3.38">Within these ten days if that thou be'st found</div> <div id="scene1.3.39">So near our public court as twenty miles,</div> <div id="scene1.3.40">Thou diest for it.</div> </div> <div id="speech23" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.41"> I do beseech your grace,</div> <div id="scene1.3.42">Let me the knowledge of my fault bear with me:</div> <div id="scene1.3.43">If with myself I hold intelligence</div> <div id="scene1.3.44">Or have acquaintance with mine own desires,</div> <div id="scene1.3.45">If that I do not dream or be not frantic,--</div> <div id="scene1.3.46">As I do trust I am not--then, dear uncle,</div> <div id="scene1.3.47">Never so much as in a thought unborn</div> <div id="scene1.3.48">Did I offend your highness.</div> </div> <div id="speech24" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.49">Thus do all traitors:</div> <div id="scene1.3.50">If their purgation did consist in words,</div> <div id="scene1.3.51">They are as innocent as grace itself:</div> <div id="scene1.3.52">Let it suffice thee that I trust thee not.</div> </div> <div id="speech25" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.53">Yet your mistrust cannot make me a traitor:</div> <div id="scene1.3.54">Tell me whereon the likelihood depends.</div> </div> <div id="speech26" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.55">Thou art thy father's daughter; there's enough.</div> </div> <div id="speech27" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.56">So was I when your highness took his dukedom;</div> <div id="scene1.3.57">So was I when your highness banish'd him:</div> <div id="scene1.3.58">Treason is not inherited, my lord;</div> <div id="scene1.3.59">Or, if we did derive it from our friends,</div> <div id="scene1.3.60">What's that to me? my father was no traitor:</div> <div id="scene1.3.61">Then, good my liege, mistake me not so much</div> <div id="scene1.3.62">To think my poverty is treacherous.</div> </div> <div id="speech28" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.63">Dear sovereign, hear me speak.</div> </div> <div id="speech29" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.64">Ay, Celia; we stay'd her for your sake,</div> <div id="scene1.3.65">Else had she with her father ranged along.</div> </div> <div id="speech30" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.66">I did not then entreat to have her stay;</div> <div id="scene1.3.67">It was your pleasure and your own remorse:</div> <div id="scene1.3.68">I was too young that time to value her;</div> <div id="scene1.3.69">But now I know her: if she be a traitor,</div> <div id="scene1.3.70">Why so am I; we still have slept together,</div> <div id="scene1.3.71">Rose at an instant, learn'd, play'd, eat together,</div> <div id="scene1.3.72">And wheresoever we went, like Juno's swans,</div> <div id="scene1.3.73">Still we went coupled and inseparable.</div> </div> <div id="speech31" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.74">She is too subtle for thee; and her smoothness,</div> <div id="scene1.3.75">Her very silence and her patience</div> <div id="scene1.3.76">Speak to the people, and they pity her.</div> <div id="scene1.3.77">Thou art a fool: she robs thee of thy name;</div> <div id="scene1.3.78">And thou wilt show more bright and seem more virtuous</div> <div id="scene1.3.79">When she is gone. Then open not thy lips:</div> <div id="scene1.3.80">Firm and irrevocable is my doom</div> <div id="scene1.3.81">Which I have pass'd upon her; she is banish'd.</div> </div> <div id="speech32" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.82">Pronounce that sentence then on me, my liege:</div> <div id="scene1.3.83">I cannot live out of her company.</div> </div> <div id="speech33" class="character">DUKE FREDERICK</div> <div class="dialog"> <div id="scene1.3.84">You are a fool. You, niece, provide yourself:</div> <div id="scene1.3.85">If you outstay the time, upon mine honour,</div> <div id="scene1.3.86">And in the greatness of my word, you die.</div> <div class="direction">Exeunt DUKE FREDERICK and Lords</div> </div> <div id="speech34" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.87">O my poor Rosalind, whither wilt thou go?</div> <div id="scene1.3.88">Wilt thou change fathers? I will give thee mine.</div> <div id="scene1.3.89">I charge thee, be not thou more grieved than I am.</div> </div> <div id="speech35" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.90">I have more cause.</div> </div> <div id="speech36" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.91"> Thou hast not, cousin;</div> <div id="scene1.3.92">Prithee be cheerful: know'st thou not, the duke</div> <div id="scene1.3.93">Hath banish'd me, his daughter?</div> </div> <div id="speech37" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.94">That he hath not.</div> </div> <div id="speech38" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.95">No, hath not? Rosalind lacks then the love</div> <div id="scene1.3.96">Which teacheth thee that thou and I am one:</div> <div id="scene1.3.97">Shall we be sunder'd? shall we part, sweet girl?</div> <div id="scene1.3.98">No: let my father seek another heir.</div> <div id="scene1.3.99">Therefore devise with me how we may fly,</div> <div id="scene1.3.100">Whither to go and what to bear with us;</div> <div id="scene1.3.101">And do not seek to take your change upon you,</div> <div id="scene1.3.102">To bear your griefs yourself and leave me out;</div> <div id="scene1.3.103">For, by this heaven, now at our sorrows pale,</div> <div id="scene1.3.104">Say what thou canst, I'll go along with thee.</div> </div> <div id="speech39" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.105">Why, whither shall we go?</div> </div> <div id="speech40" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.106">To seek my uncle in the forest of Arden.</div> </div> <div id="speech41" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.107">Alas, what danger will it be to us,</div> <div id="scene1.3.108">Maids as we are, to travel forth so far!</div> <div id="scene1.3.109">Beauty provoketh thieves sooner than gold.</div> </div> <div id="speech42" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.110">I'll put myself in poor and mean attire</div> <div id="scene1.3.111">And with a kind of umber smirch my face;</div> <div id="scene1.3.112">The like do you: so shall we pass along</div> <div id="scene1.3.113">And never stir assailants.</div> </div> <div id="speech43" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.114">Were it not better,</div> <div id="scene1.3.115">Because that I am more than common tall,</div> <div id="scene1.3.116">That I did suit me all points like a man?</div> <div id="scene1.3.117">A gallant curtle-axe upon my thigh,</div> <div id="scene1.3.118">A boar-spear in my hand; and--in my heart</div> <div id="scene1.3.119">Lie there what hidden woman's fear there will--</div> <div id="scene1.3.120">We'll have a swashing and a martial outside,</div> <div id="scene1.3.121">As many other mannish cowards have</div> <div id="scene1.3.122">That do outface it with their semblances.</div> </div> <div id="speech44" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.123">What shall I call thee when thou art a man?</div> </div> <div id="speech45" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.124">I'll have no worse a name than Jove's own page;</div> <div id="scene1.3.125">And therefore look you call me Ganymede.</div> <div id="scene1.3.126">But what will you be call'd?</div> </div> <div id="speech46" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.127">Something that hath a reference to my state</div> <div id="scene1.3.128">No longer Celia, but Aliena.</div> </div> <div id="speech47" class="character">ROSALIND</div> <div class="dialog"> <div id="scene1.3.129">But, cousin, what if we assay'd to steal</div> <div id="scene1.3.130">The clownish fool out of your father's court?</div> <div id="scene1.3.131">Would he not be a comfort to our travel?</div> </div> <div id="speech48" class="character">CELIA</div> <div class="dialog"> <div id="scene1.3.132">He'll go along o'er the wide world with me;</div> <div id="scene1.3.133">Leave me alone to woo him. Let's away,</div> <div id="scene1.3.134">And get our jewels and our wealth together,</div> <div id="scene1.3.135">Devise the fittest time and safest way</div> <div id="scene1.3.136">To hide us from pursuit that will be made</div> <div id="scene1.3.137">After my flight. Now go we in content</div> <div id="scene1.3.138">To liberty and not to banishment.</div> <div class="direction">Exeunt</div> </div> </div> </div> </div> </body> </html> ''' # }}} ae = unittest.TestCase.assertEqual def test_tokenizer(self): # {{{ tokens = [ type('')(item) for item in tokenize( r'E\ é > f [a~="y\"x"]:nth(/* fu /]* */-3.7)')] self.ae(tokens, [ "<IDENT 'E é' at 0>", "<S ' ' at 4>", "<DELIM '>' at 5>", "<S ' ' at 6>", # the no-break space is not whitespace in CSS "<IDENT 'f ' at 7>", # f\xa0 "<DELIM '[' at 9>", "<IDENT 'a' at 10>", "<DELIM '~' at 11>", "<DELIM '=' at 12>", "<STRING 'y\"x' at 13>", "<DELIM ']' at 19>", "<DELIM ':' at 20>", "<IDENT 'nth' at 21>", "<DELIM '(' at 24>", "<NUMBER '-3.7' at 37>", "<DELIM ')' at 41>", "<EOF at 42>", ]) # }}} def test_parser(self): # {{{ def repr_parse(css): selectors = parse(css) for selector in selectors: assert selector.pseudo_element is None return [repr(selector.parsed_tree).replace("(u'", "('") for selector in selectors] def parse_many(first, *others): result = repr_parse(first) for other in others: assert repr_parse(other) == result return result assert parse_many('*') == ['Element[*]'] assert parse_many('*|*') == ['Element[*]'] assert parse_many('*|foo') == ['Element[foo]'] assert parse_many('foo|*') == ['Element[foo|*]'] assert parse_many('foo|bar') == ['Element[foo|bar]'] # This will never match, but it is valid: assert parse_many('#foo#bar') == ['Hash[Hash[Element[*]#foo]#bar]'] assert parse_many( 'div>.foo', 'div> .foo', 'div >.foo', 'div > .foo', 'div \n> \t \t .foo', 'div\r>\n\n\n.foo', 'div\f>\f.foo' ) == ['CombinedSelector[Element[div] > Class[Element[*].foo]]'] assert parse_many('td.foo,.bar', 'td.foo, .bar', 'td.foo\t\r\n\f ,\t\r\n\f .bar' ) == [ 'Class[Element[td].foo]', 'Class[Element[*].bar]' ] assert parse_many('div, td.foo, div.bar span') == [ 'Element[div]', 'Class[Element[td].foo]', 'CombinedSelector[Class[Element[div].bar] ' '<followed> Element[span]]'] assert parse_many('div > p') == [ 'CombinedSelector[Element[div] > Element[p]]'] assert parse_many('td:first') == [ 'Pseudo[Element[td]:first]'] assert parse_many('td:first') == [ 'Pseudo[Element[td]:first]'] assert parse_many('td :first') == [ 'CombinedSelector[Element[td] ' '<followed> Pseudo[Element[*]:first]]'] assert parse_many('td :first') == [ 'CombinedSelector[Element[td] ' '<followed> Pseudo[Element[*]:first]]'] assert parse_many('a[name]', 'a[ name\t]') == [ 'Attrib[Element[a][name]]'] assert parse_many('a [name]') == [ 'CombinedSelector[Element[a] <followed> Attrib[Element[*][name]]]'] self.ae(parse_many('a[rel="include"]', 'a[rel = include]'), [ "Attrib[Element[a][rel = 'include']]"]) assert parse_many("a[hreflang |= 'en']", "a[hreflang|=en]") == [ "Attrib[Element[a][hreflang |= 'en']]"] self.ae(parse_many('div:nth-child(10)'), [ "Function[Element[div]:nth-child(['10'])]"]) assert parse_many(':nth-child(2n+2)') == [ "Function[Element[*]:nth-child(['2', 'n', '+2'])]"] assert parse_many('div:nth-of-type(10)') == [ "Function[Element[div]:nth-of-type(['10'])]"] assert parse_many('div div:nth-of-type(10) .aclass') == [ 'CombinedSelector[CombinedSelector[Element[div] <followed> ' "Function[Element[div]:nth-of-type(['10'])]] " '<followed> Class[Element[*].aclass]]'] assert parse_many('label:only') == [ 'Pseudo[Element[label]:only]'] assert parse_many('a:lang(fr)') == [ "Function[Element[a]:lang(['fr'])]"] assert parse_many('div:contains("foo")') == [ "Function[Element[div]:contains(['foo'])]"] assert parse_many('div#foobar') == [ 'Hash[Element[div]#foobar]'] assert parse_many('div:not(div.foo)') == [ 'Negation[Element[div]:not(Class[Element[div].foo])]'] assert parse_many('td ~ th') == [ 'CombinedSelector[Element[td] ~ Element[th]]'] # }}} def test_pseudo_elements(self): # {{{ def parse_pseudo(css): result = [] for selector in parse(css): pseudo = selector.pseudo_element pseudo = type('')(pseudo) if pseudo else pseudo # No Symbol here assert pseudo is None or isinstance(pseudo, type('')) selector = repr(selector.parsed_tree).replace("(u'", "('") result.append((selector, pseudo)) return result def parse_one(css): result = parse_pseudo(css) assert len(result) == 1 return result[0] self.ae(parse_one('foo'), ('Element[foo]', None)) self.ae(parse_one('*'), ('Element[*]', None)) self.ae(parse_one(':empty'), ('Pseudo[Element[*]:empty]', None)) # Special cases for CSS 2.1 pseudo-elements self.ae(parse_one(':BEfore'), ('Element[*]', 'before')) self.ae(parse_one(':aftER'), ('Element[*]', 'after')) self.ae(parse_one(':First-Line'), ('Element[*]', 'first-line')) self.ae(parse_one(':First-Letter'), ('Element[*]', 'first-letter')) self.ae(parse_one('::befoRE'), ('Element[*]', 'before')) self.ae(parse_one('::AFter'), ('Element[*]', 'after')) self.ae(parse_one('::firsT-linE'), ('Element[*]', 'first-line')) self.ae(parse_one('::firsT-letteR'), ('Element[*]', 'first-letter')) self.ae(parse_one('::text-content'), ('Element[*]', 'text-content')) self.ae(parse_one('::attr(name)'), ( "Element[*]", "FunctionalPseudoElement[::attr(['name'])]")) self.ae(parse_one('::Selection'), ('Element[*]', 'selection')) self.ae(parse_one('foo:after'), ('Element[foo]', 'after')) self.ae(parse_one('foo::selection'), ('Element[foo]', 'selection')) self.ae(parse_one('lorem#ipsum ~ a#b.c[href]:empty::selection'), ( 'CombinedSelector[Hash[Element[lorem]#ipsum] ~ ' 'Pseudo[Attrib[Class[Hash[Element[a]#b].c][href]]:empty]]', 'selection')) parse_pseudo('foo:before, bar, baz:after') == [ ('Element[foo]', 'before'), ('Element[bar]', None), ('Element[baz]', 'after')] # }}} def test_specificity(self): # {{{ def specificity(css): selectors = parse(css) assert len(selectors) == 1 return selectors[0].specificity() assert specificity('*') == (0, 0, 0) assert specificity(' foo') == (0, 0, 1) assert specificity(':empty ') == (0, 1, 0) assert specificity(':before') == (0, 0, 1) assert specificity('*:before') == (0, 0, 1) assert specificity(':nth-child(2)') == (0, 1, 0) assert specificity('.bar') == (0, 1, 0) assert specificity('[baz]') == (0, 1, 0) assert specificity('[baz="4"]') == (0, 1, 0) assert specificity('[baz^="4"]') == (0, 1, 0) assert specificity('#lipsum') == (1, 0, 0) assert specificity(':not(*)') == (0, 0, 0) assert specificity(':not(foo)') == (0, 0, 1) assert specificity(':not(.foo)') == (0, 1, 0) assert specificity(':not([foo])') == (0, 1, 0) assert specificity(':not(:empty)') == (0, 1, 0) assert specificity(':not(#foo)') == (1, 0, 0) assert specificity('foo:empty') == (0, 1, 1) assert specificity('foo:before') == (0, 0, 2) assert specificity('foo::before') == (0, 0, 2) assert specificity('foo:empty::before') == (0, 1, 2) assert specificity('#lorem + foo#ipsum:first-child > bar:first-line' ) == (2, 1, 3) # }}} def test_parse_errors(self): # {{{ def get_error(css): try: parse(css) except SelectorSyntaxError: # Py2, Py3, ... return str(sys.exc_info()[1]).replace("(u'", "('") self.ae(get_error('attributes(href)/html/body/a'), ( "Expected selector, got <DELIM '(' at 10>")) assert get_error('attributes(href)') == ( "Expected selector, got <DELIM '(' at 10>") assert get_error('html/body/a') == ( "Expected selector, got <DELIM '/' at 4>") assert get_error(' ') == ( "Expected selector, got <EOF at 1>") assert get_error('div, ') == ( "Expected selector, got <EOF at 5>") assert get_error(' , div') == ( "Expected selector, got <DELIM ',' at 1>") assert get_error('p, , div') == ( "Expected selector, got <DELIM ',' at 3>") assert get_error('div > ') == ( "Expected selector, got <EOF at 6>") assert get_error(' > div') == ( "Expected selector, got <DELIM '>' at 2>") assert get_error('foo|#bar') == ( "Expected ident or '*', got <HASH 'bar' at 4>") assert get_error('#.foo') == ( "Expected selector, got <DELIM '#' at 0>") assert get_error('.#foo') == ( "Expected ident, got <HASH 'foo' at 1>") assert get_error(':#foo') == ( "Expected ident, got <HASH 'foo' at 1>") assert get_error('[*]') == ( "Expected '|', got <DELIM ']' at 2>") assert get_error('[foo|]') == ( "Expected ident, got <DELIM ']' at 5>") assert get_error('[#]') == ( "Expected ident or '*', got <DELIM '#' at 1>") assert get_error('[foo=#]') == ( "Expected string or ident, got <DELIM '#' at 5>") assert get_error('[href]a') == ( "Expected selector, got <IDENT 'a' at 6>") assert get_error('[rel=stylesheet]') is None assert get_error('[rel:stylesheet]') == ( "Operator expected, got <DELIM ':' at 4>") assert get_error('[rel=stylesheet') == ( "Expected ']', got <EOF at 15>") assert get_error(':lang(fr)') is None assert get_error(':lang(fr') == ( "Expected an argument, got <EOF at 8>") assert get_error(':contains("foo') == ( "Unclosed string at 10") assert get_error('foo!') == ( "Expected selector, got <DELIM '!' at 3>") # Mis-placed pseudo-elements assert get_error('a:before:empty') == ( "Got pseudo-element ::before not at the end of a selector") assert get_error('li:before a') == ( "Got pseudo-element ::before not at the end of a selector") assert get_error(':not(:before)') == ( "Got pseudo-element ::before inside :not() at 12") assert get_error(':not(:not(a))') == ( "Got nested :not()") # }}} def test_select(self): # {{{ document = etree.fromstring(self.HTML_IDS, parser=etree.XMLParser(recover=True, no_network=True, resolve_entities=False)) select = Select(document) def select_ids(selector): for elem in select(selector): yield elem.get('id') def pcss(main, *selectors, **kwargs): result = list(select_ids(main)) for selector in selectors: self.ae(list(select_ids(selector)), result) return result all_ids = pcss('*') self.ae(all_ids[:6], [ 'html', None, 'link-href', 'link-nohref', None, 'outer-div']) self.ae(all_ids[-1:], ['foobar-span']) self.ae(pcss('div'), ['outer-div', 'li-div', 'foobar-div']) self.ae(pcss('DIV'), [ 'outer-div', 'li-div', 'foobar-div']) # case-insensitive in HTML self.ae(pcss('div div'), ['li-div']) self.ae(pcss('div, div div'), ['outer-div', 'li-div', 'foobar-div']) self.ae(pcss('a[name]'), ['name-anchor']) self.ae(pcss('a[NAme]'), ['name-anchor']) # case-insensitive in HTML: self.ae(pcss('a[rel]'), ['tag-anchor', 'nofollow-anchor']) self.ae(pcss('a[rel="tag"]'), ['tag-anchor']) self.ae(pcss('a[href*="localhost"]'), ['tag-anchor']) self.ae(pcss('a[href*=""]'), []) self.ae(pcss('a[href^="http"]'), ['tag-anchor', 'nofollow-anchor']) self.ae(pcss('a[href^="http:"]'), ['tag-anchor']) self.ae(pcss('a[href^=""]'), []) self.ae(pcss('a[href$="org"]'), ['nofollow-anchor']) self.ae(pcss('a[href$=""]'), []) self.ae(pcss('div[foobar~="bc"]', 'div[foobar~="cde"]', skip_webkit=True), ['foobar-div']) self.ae(pcss('[foobar~="ab bc"]', '[foobar~=""]', '[foobar~=" \t"]'), []) self.ae(pcss('div[foobar~="cd"]'), []) self.ae(pcss('*[lang|="En"]', '[lang|="En-us"]'), ['second-li']) # Attribute values are case sensitive self.ae(pcss('*[lang|="en"]', '[lang|="en-US"]', skip_webkit=True), []) self.ae(pcss('*[lang|="e"]'), []) self.ae(pcss(':lang("EN")', '*:lang(en-US)', skip_webkit=True), ['second-li', 'li-div']) self.ae(pcss(':lang("e")'), []) self.ae(pcss('li:nth-child(1)', 'li:first-child'), ['first-li']) self.ae(pcss('li:nth-child(3)', '#first-li ~ :nth-child(3)'), ['third-li']) self.ae(pcss('li:nth-child(10)'), []) self.ae(pcss('li:nth-child(2n)', 'li:nth-child(even)', 'li:nth-child(2n+0)'), ['second-li', 'fourth-li', 'sixth-li']) self.ae(pcss('li:nth-child(+2n+1)', 'li:nth-child(odd)'), ['first-li', 'third-li', 'fifth-li', 'seventh-li']) self.ae(pcss('li:nth-child(2n+4)'), ['fourth-li', 'sixth-li']) self.ae(pcss('li:nth-child(3n+1)'), ['first-li', 'fourth-li', 'seventh-li']) self.ae(pcss('li:nth-last-child(0)'), []) self.ae(pcss('li:nth-last-child(1)', 'li:last-child'), ['seventh-li']) self.ae(pcss('li:nth-last-child(2n)', 'li:nth-last-child(even)'), ['second-li', 'fourth-li', 'sixth-li']) self.ae(pcss('li:nth-last-child(2n+2)'), ['second-li', 'fourth-li', 'sixth-li']) self.ae(pcss('ol:first-of-type'), ['first-ol']) self.ae(pcss('ol:nth-child(1)'), []) self.ae(pcss('ol:nth-of-type(2)'), ['second-ol']) self.ae(pcss('ol:nth-last-of-type(1)'), ['second-ol']) self.ae(pcss('span:only-child'), ['foobar-span']) self.ae(pcss('li div:only-child'), ['li-div']) self.ae(pcss('div *:only-child'), ['li-div', 'foobar-span']) self.ae(pcss('p *:only-of-type', skip_webkit=True), ['p-em', 'fieldset']) self.ae(pcss('p:only-of-type', skip_webkit=True), ['paragraph']) self.ae(pcss('a:empty', 'a:EMpty'), ['name-anchor']) self.ae(pcss('li:empty'), ['third-li', 'fourth-li', 'fifth-li', 'sixth-li']) self.ae(pcss(':root', 'html:root', 'li:root'), ['html']) self.ae(pcss('* :root', 'p *:root'), []) self.ae(pcss('.a', '.b', '*.a', 'ol.a'), ['first-ol']) self.ae(pcss('.c', '*.c'), ['first-ol', 'third-li', 'fourth-li']) self.ae(pcss('ol *.c', 'ol li.c', 'li ~ li.c', 'ol > li.c'), [ 'third-li', 'fourth-li']) self.ae(pcss('#first-li', 'li#first-li', '*#first-li'), ['first-li']) self.ae(pcss('li div', 'li > div', 'div div'), ['li-div']) self.ae(pcss('div > div'), []) self.ae(pcss('div>.c', 'div > .c'), ['first-ol']) self.ae(pcss('div + div'), ['foobar-div']) self.ae(pcss('a ~ a'), ['tag-anchor', 'nofollow-anchor']) self.ae(pcss('a[rel="tag"] ~ a'), ['nofollow-anchor']) self.ae(pcss('ol#first-ol li:last-child'), ['seventh-li']) self.ae(pcss('ol#first-ol *:last-child'), ['li-div', 'seventh-li']) self.ae(pcss('#outer-div:first-child'), ['outer-div']) self.ae(pcss('#outer-div :first-child'), [ 'name-anchor', 'first-li', 'li-div', 'p-b', 'checkbox-fieldset-disabled', 'area-href']) self.ae(pcss('a[href]'), ['tag-anchor', 'nofollow-anchor']) self.ae(pcss(':not(*)'), []) self.ae(pcss('a:not([href])'), ['name-anchor']) self.ae(pcss('ol :Not(li[class])', skip_webkit=True), [ 'first-li', 'second-li', 'li-div', 'fifth-li', 'sixth-li', 'seventh-li']) self.ae(pcss(r'di\a0 v', r'div\['), []) self.ae(pcss(r'[h\a0 ref]', r'[h\]ref]'), []) self.assertRaises(ExpressionError, lambda : tuple(select('body:nth-child'))) select = Select(document, ignore_inappropriate_pseudo_classes=True) self.assertGreater(len(tuple(select('p:hover'))), 0) def test_select_shakespeare(self): document = html.document_fromstring(self.HTML_SHAKESPEARE) select = Select(document) def count(s): return sum(1 for r in select(s)) # Data borrowed from http://mootools.net/slickspeed/ # Changed from original; probably because I'm only self.ae(count('*'), 249) assert count('div:only-child') == 22 # ? assert count('div:nth-child(even)') == 106 assert count('div:nth-child(2n)') == 106 assert count('div:nth-child(odd)') == 137 assert count('div:nth-child(2n+1)') == 137 assert count('div:nth-child(n)') == 243 assert count('div:last-child') == 53 assert count('div:first-child') == 51 assert count('div > div') == 242 assert count('div + div') == 190 assert count('div ~ div') == 190 assert count('body') == 1 assert count('body div') == 243 assert count('div') == 243 assert count('div div') == 242 assert count('div div div') == 241 assert count('div, div, div') == 243 assert count('div, a, span') == 243 assert count('.dialog') == 51 assert count('div.dialog') == 51 assert count('div .dialog') == 51 assert count('div.character, div.dialog') == 99 assert count('div.direction.dialog') == 0 assert count('div.dialog.direction') == 0 assert count('div.dialog.scene') == 1 assert count('div.scene.scene') == 1 assert count('div.scene .scene') == 0 assert count('div.direction .dialog ') == 0 assert count('div .dialog .direction') == 4 assert count('div.dialog .dialog .direction') == 4 assert count('#speech5') == 1 assert count('div#speech5') == 1 assert count('div #speech5') == 1 assert count('div.scene div.dialog') == 49 assert count('div#scene1 div.dialog div') == 142 assert count('#scene1 #speech1') == 1 assert count('div[class]') == 103 assert count('div[class=dialog]') == 50 assert count('div[class^=dia]') == 51 assert count('div[class$=log]') == 50 assert count('div[class*=sce]') == 1 assert count('div[class|=dialog]') == 50 # ? Seems right assert count('div[class~=dialog]') == 51 # ? Seems right # }}} # Run tests {{{ def find_tests(): return unittest.defaultTestLoader.loadTestsFromTestCase(TestCSSSelectors) def run_tests(find_tests=find_tests, for_build=False): if not for_build: parser = argparse.ArgumentParser() parser.add_argument('name', nargs='?', default=None, help='The name of the test to run') args = parser.parse_args() if not for_build and args.name and args.name.startswith('.'): tests = find_tests() q = args.name[1:] if not q.startswith('test_'): q = 'test_' + q ans = None try: for test in tests: if test._testMethodName == q: ans = test raise StopIteration() except StopIteration: pass if ans is None: print('No test named %s found' % args.name) raise SystemExit(1) tests = ans else: tests = unittest.defaultTestLoader.loadTestsFromName(args.name) if not for_build and args.name else find_tests() r = unittest.TextTestRunner if for_build: r = r(verbosity=0, buffer=True, failfast=True) else: r = r(verbosity=4) result = r.run(tests) if for_build and result.errors or result.failures: raise SystemExit(1) if __name__ == '__main__': run_tests() # }}}
37,012
Python
.py
800
40.1175
129
0.599027
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,825
parser.py
kovidgoyal_calibre/src/css_selectors/parser.py
#!/usr/bin/env python # vim:fileencoding=utf-8 """ Tokenizer, parser and parsed objects for CSS selectors. :copyright: (c) 2007-2012 Ian Bicking and contributors. See AUTHORS for more details. :license: BSD, see LICENSE for more details. """ import operator import re import string import sys from css_selectors.errors import ExpressionError, SelectorSyntaxError from polyglot.builtins import codepoint_to_chr, unicode_type utab = {c:c+32 for c in range(ord(u'A'), ord(u'Z')+1)} if sys.version_info.major < 3: tab = string.maketrans(string.ascii_uppercase, string.ascii_lowercase) def ascii_lower(string): """Lower-case, but only in the ASCII range.""" return string.translate(utab if isinstance(string, unicode_type) else tab) def urepr(x): if isinstance(x, list): return '[%s]' % ', '.join((map(urepr, x))) ans = repr(x) if ans.startswith("u'") or ans.startswith('u"'): ans = ans[1:] return ans else: def ascii_lower(x): return x.translate(utab) urepr = repr # Parsed objects class Selector: """ Represents a parsed selector. """ def __init__(self, tree, pseudo_element=None): self.parsed_tree = tree if pseudo_element is not None and not isinstance( pseudo_element, FunctionalPseudoElement): pseudo_element = ascii_lower(pseudo_element) #: A :class:`FunctionalPseudoElement`, #: or the identifier for the pseudo-element as a string, # or ``None``. #: #: +-------------------------+----------------+--------------------------------+ #: | | Selector | Pseudo-element | #: +=========================+================+================================+ #: | CSS3 syntax | ``a::before`` | ``'before'`` | #: +-------------------------+----------------+--------------------------------+ #: | Older syntax | ``a:before`` | ``'before'`` | #: +-------------------------+----------------+--------------------------------+ #: | From the Lists3_ draft, | ``li::marker`` | ``'marker'`` | #: | not in Selectors3 | | | #: +-------------------------+----------------+--------------------------------+ #: | Invalid pseudo-class | ``li:marker`` | ``None`` | #: +-------------------------+----------------+--------------------------------+ #: | Functional | ``a::foo(2)`` | ``FunctionalPseudoElement(…)`` | #: +-------------------------+----------------+--------------------------------+ #: # : .. _Lists3: http://www.w3.org/TR/2011/WD-css3-lists-20110524/#marker-pseudoelement self.pseudo_element = pseudo_element def __repr__(self): if isinstance(self.pseudo_element, FunctionalPseudoElement): pseudo_element = repr(self.pseudo_element) if self.pseudo_element: pseudo_element = '::%s' % self.pseudo_element else: pseudo_element = '' return '%s[%r%s]' % ( self.__class__.__name__, self.parsed_tree, pseudo_element) def specificity(self): """Return the specificity_ of this selector as a tuple of 3 integers. .. _specificity: http://www.w3.org/TR/selectors/#specificity """ a, b, c = self.parsed_tree.specificity() if self.pseudo_element: c += 1 return a, b, c class Class: """ Represents selector.class_name """ def __init__(self, selector, class_name): self.selector = selector self.class_name = class_name def __repr__(self): return '%s[%r.%s]' % ( self.__class__.__name__, self.selector, self.class_name) def specificity(self): a, b, c = self.selector.specificity() b += 1 return a, b, c class FunctionalPseudoElement: """ Represents selector::name(arguments) .. attribute:: name The name (identifier) of the pseudo-element, as a string. .. attribute:: arguments The arguments of the pseudo-element, as a list of tokens. **Note:** tokens are not part of the public API, and may change between versions. Use at your own risks. """ def __init__(self, name, arguments): self.name = ascii_lower(name) self.arguments = arguments def __repr__(self): return '%s[::%s(%s)]' % ( self.__class__.__name__, self.name, urepr([token.value for token in self.arguments])) def argument_types(self): return [token.type for token in self.arguments] def specificity(self): a, b, c = self.selector.specificity() b += 1 return a, b, c class Function: """ Represents selector:name(expr) """ def __init__(self, selector, name, arguments): self.selector = selector self.name = ascii_lower(name) self.arguments = arguments self._parsed_arguments = None def __repr__(self): return '%s[%r:%s(%s)]' % ( self.__class__.__name__, self.selector, self.name, urepr([token.value for token in self.arguments])) def argument_types(self): return [token.type for token in self.arguments] @property def parsed_arguments(self): if self._parsed_arguments is None: try: self._parsed_arguments = parse_series(self.arguments) except ValueError: raise ExpressionError("Invalid series: '%r'" % self.arguments) return self._parsed_arguments def parse_arguments(self): if not self.arguments_parsed: self.arguments_parsed = True def specificity(self): a, b, c = self.selector.specificity() b += 1 return a, b, c class Pseudo: """ Represents selector:ident """ def __init__(self, selector, ident): self.selector = selector self.ident = ascii_lower(ident) def __repr__(self): return '%s[%r:%s]' % ( self.__class__.__name__, self.selector, self.ident) def specificity(self): a, b, c = self.selector.specificity() b += 1 return a, b, c class Negation: """ Represents selector:not(subselector) """ def __init__(self, selector, subselector): self.selector = selector self.subselector = subselector def __repr__(self): return '%s[%r:not(%r)]' % ( self.__class__.__name__, self.selector, self.subselector) def specificity(self): a1, b1, c1 = self.selector.specificity() a2, b2, c2 = self.subselector.specificity() return a1 + a2, b1 + b2, c1 + c2 class Attrib: """ Represents selector[namespace|attrib operator value] """ def __init__(self, selector, namespace, attrib, operator, value): self.selector = selector self.namespace = namespace self.attrib = attrib self.operator = operator self.value = value def __repr__(self): if self.namespace: attrib = '%s|%s' % (self.namespace, self.attrib) else: attrib = self.attrib if self.operator == 'exists': return '%s[%r[%s]]' % ( self.__class__.__name__, self.selector, attrib) else: return '%s[%r[%s %s %s]]' % ( self.__class__.__name__, self.selector, attrib, self.operator, urepr(self.value)) def specificity(self): a, b, c = self.selector.specificity() b += 1 return a, b, c class Element: """ Represents namespace|element `None` is for the universal selector '*' """ def __init__(self, namespace=None, element=None): self.namespace = namespace self.element = element def __repr__(self): element = self.element or '*' if self.namespace: element = '%s|%s' % (self.namespace, element) return '%s[%s]' % (self.__class__.__name__, element) def specificity(self): if self.element: return 0, 0, 1 else: return 0, 0, 0 class Hash: """ Represents selector#id """ def __init__(self, selector, id): self.selector = selector self.id = id def __repr__(self): return '%s[%r#%s]' % ( self.__class__.__name__, self.selector, self.id) def specificity(self): a, b, c = self.selector.specificity() a += 1 return a, b, c class CombinedSelector: def __init__(self, selector, combinator, subselector): assert selector is not None self.selector = selector self.combinator = combinator self.subselector = subselector def __repr__(self): if self.combinator == ' ': comb = '<followed>' else: comb = self.combinator return '%s[%r %s %r]' % ( self.__class__.__name__, self.selector, comb, self.subselector) def specificity(self): a1, b1, c1 = self.selector.specificity() a2, b2, c2 = self.subselector.specificity() return a1 + a2, b1 + b2, c1 + c2 # Parser # foo _el_re = re.compile(r'^[ \t\r\n\f]*([a-zA-Z]+)[ \t\r\n\f]*$') # foo#bar or #bar _id_re = re.compile(r'^[ \t\r\n\f]*([a-zA-Z]*)#([a-zA-Z0-9_-]+)[ \t\r\n\f]*$') # foo.bar or .bar _class_re = re.compile( r'^[ \t\r\n\f]*([a-zA-Z]*)\.([a-zA-Z][a-zA-Z0-9_-]*)[ \t\r\n\f]*$') def parse(css): """Parse a CSS *group of selectors*. :param css: A *group of selectors* as an Unicode string. :raises: :class:`SelectorSyntaxError` on invalid selectors. :returns: A list of parsed :class:`Selector` objects, one for each selector in the comma-separated group. """ # Fast path for simple cases match = _el_re.match(css) if match: return [Selector(Element(element=match.group(1)))] match = _id_re.match(css) if match is not None: return [Selector(Hash(Element(element=match.group(1) or None), match.group(2)))] match = _class_re.match(css) if match is not None: return [Selector(Class(Element(element=match.group(1) or None), match.group(2)))] stream = TokenStream(tokenize(css)) stream.source = css return list(parse_selector_group(stream)) # except SelectorSyntaxError: # e = sys.exc_info()[1] # message = "%s at %s -> %r" % ( # e, stream.used, stream.peek()) # e.msg = message # e.args = tuple([message]) # raise def parse_selector_group(stream): stream.skip_whitespace() while 1: yield Selector(*parse_selector(stream)) if stream.peek() == ('DELIM', ','): stream.next() stream.skip_whitespace() else: break def parse_selector(stream): result, pseudo_element = parse_simple_selector(stream) while 1: stream.skip_whitespace() peek = stream.peek() if peek in (('EOF', None), ('DELIM', ',')): break if pseudo_element: raise SelectorSyntaxError( 'Got pseudo-element ::%s not at the end of a selector' % pseudo_element) if peek.is_delim('+', '>', '~'): # A combinator combinator = stream.next().value stream.skip_whitespace() else: # By exclusion, the last parse_simple_selector() ended # at peek == ' ' combinator = ' ' next_selector, pseudo_element = parse_simple_selector(stream) result = CombinedSelector(result, combinator, next_selector) return result, pseudo_element special_pseudo_elements = ( 'first-line', 'first-letter', 'before', 'after') def parse_simple_selector(stream, inside_negation=False): stream.skip_whitespace() selector_start = len(stream.used) peek = stream.peek() if peek.type == 'IDENT' or peek == ('DELIM', '*'): if peek.type == 'IDENT': namespace = stream.next().value else: stream.next() namespace = None if stream.peek() == ('DELIM', '|'): stream.next() element = stream.next_ident_or_star() else: element = namespace namespace = None else: element = namespace = None result = Element(namespace, element) pseudo_element = None while 1: peek = stream.peek() if peek.type in ('S', 'EOF') or peek.is_delim(',', '+', '>', '~') or ( inside_negation and peek == ('DELIM', ')')): break if pseudo_element: raise SelectorSyntaxError( 'Got pseudo-element ::%s not at the end of a selector' % pseudo_element) if peek.type == 'HASH': result = Hash(result, stream.next().value) elif peek == ('DELIM', '.'): stream.next() result = Class(result, stream.next_ident()) elif peek == ('DELIM', '['): stream.next() result = parse_attrib(result, stream) elif peek == ('DELIM', ':'): stream.next() if stream.peek() == ('DELIM', ':'): stream.next() pseudo_element = stream.next_ident() if stream.peek() == ('DELIM', '('): stream.next() pseudo_element = FunctionalPseudoElement( pseudo_element, parse_arguments(stream)) continue ident = stream.next_ident() if ident.lower() in special_pseudo_elements: # Special case: CSS 2.1 pseudo-elements can have a single ':' # Any new pseudo-element must have two. pseudo_element = unicode_type(ident) continue if stream.peek() != ('DELIM', '('): result = Pseudo(result, ident) continue stream.next() stream.skip_whitespace() if ident.lower() == 'not': if inside_negation: raise SelectorSyntaxError('Got nested :not()') argument, argument_pseudo_element = parse_simple_selector( stream, inside_negation=True) next = stream.next() if argument_pseudo_element: raise SelectorSyntaxError( 'Got pseudo-element ::%s inside :not() at %s' % (argument_pseudo_element, next.pos)) if next != ('DELIM', ')'): raise SelectorSyntaxError("Expected ')', got %s" % (next,)) result = Negation(result, argument) else: result = Function(result, ident, parse_arguments(stream)) else: raise SelectorSyntaxError( "Expected selector, got %s" % (peek,)) if len(stream.used) == selector_start: raise SelectorSyntaxError( "Expected selector, got %s" % (stream.peek(),)) return result, pseudo_element def parse_arguments(stream): arguments = [] while 1: stream.skip_whitespace() next = stream.next() if next.type in ('IDENT', 'STRING', 'NUMBER') or next in [ ('DELIM', '+'), ('DELIM', '-')]: arguments.append(next) elif next == ('DELIM', ')'): return arguments else: raise SelectorSyntaxError( "Expected an argument, got %s" % (next,)) def parse_attrib(selector, stream): stream.skip_whitespace() attrib = stream.next_ident_or_star() if attrib is None and stream.peek() != ('DELIM', '|'): raise SelectorSyntaxError( "Expected '|', got %s" % (stream.peek(),)) if stream.peek() == ('DELIM', '|'): stream.next() if stream.peek() == ('DELIM', '='): namespace = None stream.next() op = '|=' else: namespace = attrib attrib = stream.next_ident() op = None else: namespace = op = None if op is None: stream.skip_whitespace() next = stream.next() if next == ('DELIM', ']'): return Attrib(selector, namespace, attrib, 'exists', None) elif next == ('DELIM', '='): op = '=' elif next.is_delim('^', '$', '*', '~', '|', '!') and ( stream.peek() == ('DELIM', '=')): op = next.value + '=' stream.next() else: raise SelectorSyntaxError( "Operator expected, got %s" % (next,)) stream.skip_whitespace() value = stream.next() if value.type not in ('IDENT', 'STRING'): raise SelectorSyntaxError( "Expected string or ident, got %s" % (value,)) stream.skip_whitespace() next = stream.next() if next != ('DELIM', ']'): raise SelectorSyntaxError( "Expected ']', got %s" % (next,)) return Attrib(selector, namespace, attrib, op, value.value) def parse_series(tokens): """ Parses the arguments for :nth-child() and friends. :raises: A list of tokens :returns: :``(a, b)`` """ for token in tokens: if token.type == 'STRING': raise ValueError('String tokens not allowed in series.') s = ''.join(token.value for token in tokens).strip() if s == 'odd': return (2, 1) elif s == 'even': return (2, 0) elif s == 'n': return (1, 0) if 'n' not in s: # Just b return (0, int(s)) a, b = s.split('n', 1) if not a: a = 1 elif a == '-' or a == '+': a = int(a+'1') else: a = int(a) if not b: b = 0 else: b = int(b) return (a, b) # Token objects class Token(tuple): def __new__(cls, type_, value, pos): obj = tuple.__new__(cls, (type_, value)) obj.pos = pos return obj def __repr__(self): return "<%s '%s' at %i>" % (self.type, self.value, self.pos) def is_delim(self, *values): return self.type == 'DELIM' and self.value in values type = property(operator.itemgetter(0)) value = property(operator.itemgetter(1)) class EOFToken(Token): def __new__(cls, pos): return Token.__new__(cls, 'EOF', None, pos) def __repr__(self): return '<%s at %i>' % (self.type, self.pos) # Tokenizer class TokenMacros: unicode_escape = r'\\([0-9a-f]{1,6})(?:\r\n|[ \n\r\t\f])?' escape = unicode_escape + r'|\\[^\n\r\f0-9a-f]' string_escape = r'\\(?:\n|\r\n|\r|\f)|' + escape nonascii = r'[^\0-\177]' nmchar = '[_a-z0-9-]|%s|%s' % (escape, nonascii) nmstart = '[_a-z]|%s|%s' % (escape, nonascii) def _compile(pattern): return re.compile(pattern % vars(TokenMacros), re.IGNORECASE).match _match_whitespace = _compile(r'[ \t\r\n\f]+') _match_number = _compile(r'[+-]?(?:[0-9]*\.[0-9]+|[0-9]+)') _match_hash = _compile('#(?:%(nmchar)s)+') _match_ident = _compile('-?(?:%(nmstart)s)(?:%(nmchar)s)*') _match_string_by_quote = { "'": _compile(r"([^\n\r\f\\']|%(string_escape)s)*"), '"': _compile(r'([^\n\r\f\\"]|%(string_escape)s)*'), } _sub_simple_escape = re.compile(r'\\(.)').sub _sub_unicode_escape = re.compile(TokenMacros.unicode_escape, re.I).sub _sub_newline_escape =re.compile(r'\\(?:\n|\r\n|\r|\f)').sub # Same as r'\1', but faster on CPython if hasattr(operator, 'methodcaller'): # Python 2.6+ _replace_simple = operator.methodcaller('group', 1) else: def _replace_simple(match): return match.group(1) def _replace_unicode(match): codepoint = int(match.group(1), 16) if codepoint > sys.maxunicode: codepoint = 0xFFFD return codepoint_to_chr(codepoint) def unescape_ident(value): value = _sub_unicode_escape(_replace_unicode, value) value = _sub_simple_escape(_replace_simple, value) return value def tokenize(s): pos = 0 len_s = len(s) while pos < len_s: match = _match_whitespace(s, pos=pos) if match: yield Token('S', ' ', pos) pos = match.end() continue match = _match_ident(s, pos=pos) if match: value = _sub_simple_escape(_replace_simple, _sub_unicode_escape(_replace_unicode, match.group())) yield Token('IDENT', value, pos) pos = match.end() continue match = _match_hash(s, pos=pos) if match: value = _sub_simple_escape(_replace_simple, _sub_unicode_escape(_replace_unicode, match.group()[1:])) yield Token('HASH', value, pos) pos = match.end() continue quote = s[pos] if quote in _match_string_by_quote: match = _match_string_by_quote[quote](s, pos=pos + 1) assert match, 'Should have found at least an empty match' end_pos = match.end() if end_pos == len_s: raise SelectorSyntaxError('Unclosed string at %s' % pos) if s[end_pos] != quote: raise SelectorSyntaxError('Invalid string at %s' % pos) value = _sub_simple_escape(_replace_simple, _sub_unicode_escape(_replace_unicode, _sub_newline_escape('', match.group()))) yield Token('STRING', value, pos) pos = end_pos + 1 continue match = _match_number(s, pos=pos) if match: value = match.group() yield Token('NUMBER', value, pos) pos = match.end() continue pos2 = pos + 2 if s[pos:pos2] == '/*': pos = s.find('*/', pos2) if pos == -1: pos = len_s else: pos += 2 continue yield Token('DELIM', s[pos], pos) pos += 1 assert pos == len_s yield EOFToken(pos) class TokenStream: def __init__(self, tokens, source=None): self.used = [] self.tokens = iter(tokens) self.source = source self.peeked = None self._peeking = False try: self.next_token = self.tokens.next except AttributeError: # Python 3 self.next_token = self.tokens.__next__ def next(self): if self._peeking: self._peeking = False self.used.append(self.peeked) return self.peeked else: next = self.next_token() self.used.append(next) return next def peek(self): if not self._peeking: self.peeked = self.next_token() self._peeking = True return self.peeked def next_ident(self): next = self.next() if next.type != 'IDENT': raise SelectorSyntaxError('Expected ident, got %s' % (next,)) return next.value def next_ident_or_star(self): next = self.next() if next.type == 'IDENT': return next.value elif next == ('DELIM', '*'): return None else: raise SelectorSyntaxError( "Expected ident or '*', got %s" % (next,)) def skip_whitespace(self): peek = self.peek() if peek.type == 'S': self.next()
23,816
Python
.py
644
28.076087
94
0.528272
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,826
core_name_map.py
kovidgoyal_calibre/src/qt/core_name_map.py
# autogenerated by __main__.py do not edit name_map = {'PYQT_VERSION': 'PyQt6.QtCore', 'PYQT_VERSION_STR': 'PyQt6.QtCore', 'QAbstractAnimation': 'PyQt6.QtCore', 'QAbstractButton': 'PyQt6.QtWidgets', 'QAbstractEventDispatcher': 'PyQt6.QtCore', 'QAbstractFileIconProvider': 'PyQt6.QtGui', 'QAbstractGraphicsShapeItem': 'PyQt6.QtWidgets', 'QAbstractItemDelegate': 'PyQt6.QtWidgets', 'QAbstractItemModel': 'PyQt6.QtCore', 'QAbstractItemView': 'PyQt6.QtWidgets', 'QAbstractListModel': 'PyQt6.QtCore', 'QAbstractNativeEventFilter': 'PyQt6.QtCore', 'QAbstractNetworkCache': 'PyQt6.QtNetwork', 'QAbstractOpenGLFunctions': 'PyQt6.QtOpenGL', 'QAbstractPrintDialog': 'PyQt6.QtPrintSupport', 'QAbstractProxyModel': 'PyQt6.QtCore', 'QAbstractScrollArea': 'PyQt6.QtWidgets', 'QAbstractSlider': 'PyQt6.QtWidgets', 'QAbstractSocket': 'PyQt6.QtNetwork', 'QAbstractSpinBox': 'PyQt6.QtWidgets', 'QAbstractTableModel': 'PyQt6.QtCore', 'QAbstractTextDocumentLayout': 'PyQt6.QtGui', 'QAction': 'PyQt6.QtGui', 'QActionEvent': 'PyQt6.QtGui', 'QActionGroup': 'PyQt6.QtGui', 'QAnimationGroup': 'PyQt6.QtCore', 'QApplication': 'PyQt6.QtWidgets', 'QAudio': 'PyQt6.QtMultimedia', 'QAudioBuffer': 'PyQt6.QtMultimedia', 'QAudioDecoder': 'PyQt6.QtMultimedia', 'QAudioDevice': 'PyQt6.QtMultimedia', 'QAudioFormat': 'PyQt6.QtMultimedia', 'QAudioInput': 'PyQt6.QtMultimedia', 'QAudioOutput': 'PyQt6.QtMultimedia', 'QAudioSink': 'PyQt6.QtMultimedia', 'QAudioSource': 'PyQt6.QtMultimedia', 'QAuthenticator': 'PyQt6.QtNetwork', 'QBackingStore': 'PyQt6.QtGui', 'QBasicTimer': 'PyQt6.QtCore', 'QBitArray': 'PyQt6.QtCore', 'QBitmap': 'PyQt6.QtGui', 'QBluetoothPermission': 'PyQt6.QtCore', 'QBoxLayout': 'PyQt6.QtWidgets', 'QBrush': 'PyQt6.QtGui', 'QBuffer': 'PyQt6.QtCore', 'QButtonGroup': 'PyQt6.QtWidgets', 'QByteArray': 'PyQt6.QtCore', 'QByteArrayMatcher': 'PyQt6.QtCore', 'QCalendar': 'PyQt6.QtCore', 'QCalendarPermission': 'PyQt6.QtCore', 'QCalendarWidget': 'PyQt6.QtWidgets', 'QCamera': 'PyQt6.QtMultimedia', 'QCameraDevice': 'PyQt6.QtMultimedia', 'QCameraFormat': 'PyQt6.QtMultimedia', 'QCameraPermission': 'PyQt6.QtCore', 'QCapturableWindow': 'PyQt6.QtMultimedia', 'QCborError': 'PyQt6.QtCore', 'QCborKnownTags': 'PyQt6.QtCore', 'QCborSimpleType': 'PyQt6.QtCore', 'QCborStreamReader': 'PyQt6.QtCore', 'QCborStreamWriter': 'PyQt6.QtCore', 'QCheckBox': 'PyQt6.QtWidgets', 'QChildEvent': 'PyQt6.QtCore', 'QChildWindowEvent': 'PyQt6.QtGui', 'QClipboard': 'PyQt6.QtGui', 'QCloseEvent': 'PyQt6.QtGui', 'QCollator': 'PyQt6.QtCore', 'QCollatorSortKey': 'PyQt6.QtCore', 'QColor': 'PyQt6.QtGui', 'QColorConstants': 'PyQt6.QtGui', 'QColorDialog': 'PyQt6.QtWidgets', 'QColorSpace': 'PyQt6.QtGui', 'QColorTransform': 'PyQt6.QtGui', 'QColumnView': 'PyQt6.QtWidgets', 'QComboBox': 'PyQt6.QtWidgets', 'QCommandLineOption': 'PyQt6.QtCore', 'QCommandLineParser': 'PyQt6.QtCore', 'QCommandLinkButton': 'PyQt6.QtWidgets', 'QCommonStyle': 'PyQt6.QtWidgets', 'QCompleter': 'PyQt6.QtWidgets', 'QConcatenateTablesProxyModel': 'PyQt6.QtCore', 'QConicalGradient': 'PyQt6.QtGui', 'QContactsPermission': 'PyQt6.QtCore', 'QContextMenuEvent': 'PyQt6.QtGui', 'QCoreApplication': 'PyQt6.QtCore', 'QCryptographicHash': 'PyQt6.QtCore', 'QCursor': 'PyQt6.QtGui', 'QDataStream': 'PyQt6.QtCore', 'QDataWidgetMapper': 'PyQt6.QtWidgets', 'QDate': 'PyQt6.QtCore', 'QDateEdit': 'PyQt6.QtWidgets', 'QDateTime': 'PyQt6.QtCore', 'QDateTimeEdit': 'PyQt6.QtWidgets', 'QDeadlineTimer': 'PyQt6.QtCore', 'QDesktopServices': 'PyQt6.QtGui', 'QDial': 'PyQt6.QtWidgets', 'QDialog': 'PyQt6.QtWidgets', 'QDialogButtonBox': 'PyQt6.QtWidgets', 'QDir': 'PyQt6.QtCore', 'QDirIterator': 'PyQt6.QtCore', 'QDnsDomainNameRecord': 'PyQt6.QtNetwork', 'QDnsHostAddressRecord': 'PyQt6.QtNetwork', 'QDnsLookup': 'PyQt6.QtNetwork', 'QDnsMailExchangeRecord': 'PyQt6.QtNetwork', 'QDnsServiceRecord': 'PyQt6.QtNetwork', 'QDnsTextRecord': 'PyQt6.QtNetwork', 'QDockWidget': 'PyQt6.QtWidgets', 'QDoubleSpinBox': 'PyQt6.QtWidgets', 'QDoubleValidator': 'PyQt6.QtGui', 'QDrag': 'PyQt6.QtGui', 'QDragEnterEvent': 'PyQt6.QtGui', 'QDragLeaveEvent': 'PyQt6.QtGui', 'QDragMoveEvent': 'PyQt6.QtGui', 'QDropEvent': 'PyQt6.QtGui', 'QDynamicPropertyChangeEvent': 'PyQt6.QtCore', 'QEasingCurve': 'PyQt6.QtCore', 'QElapsedTimer': 'PyQt6.QtCore', 'QEnterEvent': 'PyQt6.QtGui', 'QErrorMessage': 'PyQt6.QtWidgets', 'QEvent': 'PyQt6.QtCore', 'QEventLoop': 'PyQt6.QtCore', 'QEventLoopLocker': 'PyQt6.QtCore', 'QEventPoint': 'PyQt6.QtGui', 'QExposeEvent': 'PyQt6.QtGui', 'QFile': 'PyQt6.QtCore', 'QFileDevice': 'PyQt6.QtCore', 'QFileDialog': 'PyQt6.QtWidgets', 'QFileIconProvider': 'PyQt6.QtWidgets', 'QFileInfo': 'PyQt6.QtCore', 'QFileOpenEvent': 'PyQt6.QtGui', 'QFileSelector': 'PyQt6.QtCore', 'QFileSystemModel': 'PyQt6.QtGui', 'QFileSystemWatcher': 'PyQt6.QtCore', 'QFocusEvent': 'PyQt6.QtGui', 'QFocusFrame': 'PyQt6.QtWidgets', 'QFont': 'PyQt6.QtGui', 'QFontComboBox': 'PyQt6.QtWidgets', 'QFontDatabase': 'PyQt6.QtGui', 'QFontDialog': 'PyQt6.QtWidgets', 'QFontInfo': 'PyQt6.QtGui', 'QFontMetrics': 'PyQt6.QtGui', 'QFontMetricsF': 'PyQt6.QtGui', 'QFormLayout': 'PyQt6.QtWidgets', 'QFrame': 'PyQt6.QtWidgets', 'QGenericArgument': 'PyQt6.QtCore', 'QGenericReturnArgument': 'PyQt6.QtCore', 'QGesture': 'PyQt6.QtWidgets', 'QGestureEvent': 'PyQt6.QtWidgets', 'QGestureRecognizer': 'PyQt6.QtWidgets', 'QGlyphRun': 'PyQt6.QtGui', 'QGradient': 'PyQt6.QtGui', 'QGraphicsAnchor': 'PyQt6.QtWidgets', 'QGraphicsAnchorLayout': 'PyQt6.QtWidgets', 'QGraphicsBlurEffect': 'PyQt6.QtWidgets', 'QGraphicsColorizeEffect': 'PyQt6.QtWidgets', 'QGraphicsDropShadowEffect': 'PyQt6.QtWidgets', 'QGraphicsEffect': 'PyQt6.QtWidgets', 'QGraphicsEllipseItem': 'PyQt6.QtWidgets', 'QGraphicsGridLayout': 'PyQt6.QtWidgets', 'QGraphicsItem': 'PyQt6.QtWidgets', 'QGraphicsItemGroup': 'PyQt6.QtWidgets', 'QGraphicsLayout': 'PyQt6.QtWidgets', 'QGraphicsLayoutItem': 'PyQt6.QtWidgets', 'QGraphicsLineItem': 'PyQt6.QtWidgets', 'QGraphicsLinearLayout': 'PyQt6.QtWidgets', 'QGraphicsObject': 'PyQt6.QtWidgets', 'QGraphicsOpacityEffect': 'PyQt6.QtWidgets', 'QGraphicsPathItem': 'PyQt6.QtWidgets', 'QGraphicsPixmapItem': 'PyQt6.QtWidgets', 'QGraphicsPolygonItem': 'PyQt6.QtWidgets', 'QGraphicsProxyWidget': 'PyQt6.QtWidgets', 'QGraphicsRectItem': 'PyQt6.QtWidgets', 'QGraphicsRotation': 'PyQt6.QtWidgets', 'QGraphicsScale': 'PyQt6.QtWidgets', 'QGraphicsScene': 'PyQt6.QtWidgets', 'QGraphicsSceneContextMenuEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneDragDropEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneHelpEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneHoverEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneMouseEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneMoveEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneResizeEvent': 'PyQt6.QtWidgets', 'QGraphicsSceneWheelEvent': 'PyQt6.QtWidgets', 'QGraphicsSimpleTextItem': 'PyQt6.QtWidgets', 'QGraphicsTextItem': 'PyQt6.QtWidgets', 'QGraphicsTransform': 'PyQt6.QtWidgets', 'QGraphicsVideoItem': 'PyQt6.QtMultimediaWidgets', 'QGraphicsView': 'PyQt6.QtWidgets', 'QGraphicsWidget': 'PyQt6.QtWidgets', 'QGridLayout': 'PyQt6.QtWidgets', 'QGroupBox': 'PyQt6.QtWidgets', 'QGuiApplication': 'PyQt6.QtGui', 'QHBoxLayout': 'PyQt6.QtWidgets', 'QHeaderView': 'PyQt6.QtWidgets', 'QHelpEvent': 'PyQt6.QtGui', 'QHideEvent': 'PyQt6.QtGui', 'QHostAddress': 'PyQt6.QtNetwork', 'QHostInfo': 'PyQt6.QtNetwork', 'QHoverEvent': 'PyQt6.QtGui', 'QHstsPolicy': 'PyQt6.QtNetwork', 'QHttp1Configuration': 'PyQt6.QtNetwork', 'QHttp2Configuration': 'PyQt6.QtNetwork', 'QHttpHeaders': 'PyQt6.QtNetwork', 'QHttpMultiPart': 'PyQt6.QtNetwork', 'QHttpPart': 'PyQt6.QtNetwork', 'QIODevice': 'PyQt6.QtCore', 'QIODeviceBase': 'PyQt6.QtCore', 'QIcon': 'PyQt6.QtGui', 'QIconDragEvent': 'PyQt6.QtGui', 'QIconEngine': 'PyQt6.QtGui', 'QIdentityProxyModel': 'PyQt6.QtCore', 'QImage': 'PyQt6.QtGui', 'QImageCapture': 'PyQt6.QtMultimedia', 'QImageIOHandler': 'PyQt6.QtGui', 'QImageReader': 'PyQt6.QtGui', 'QImageWriter': 'PyQt6.QtGui', 'QInputDevice': 'PyQt6.QtGui', 'QInputDialog': 'PyQt6.QtWidgets', 'QInputEvent': 'PyQt6.QtGui', 'QInputMethod': 'PyQt6.QtGui', 'QInputMethodEvent': 'PyQt6.QtGui', 'QInputMethodQueryEvent': 'PyQt6.QtGui', 'QIntValidator': 'PyQt6.QtGui', 'QItemDelegate': 'PyQt6.QtWidgets', 'QItemEditorCreatorBase': 'PyQt6.QtWidgets', 'QItemEditorFactory': 'PyQt6.QtWidgets', 'QItemSelection': 'PyQt6.QtCore', 'QItemSelectionModel': 'PyQt6.QtCore', 'QItemSelectionRange': 'PyQt6.QtCore', 'QJsonDocument': 'PyQt6.QtCore', 'QJsonParseError': 'PyQt6.QtCore', 'QJsonValue': 'PyQt6.QtCore', 'QKeyCombination': 'PyQt6.QtCore', 'QKeyEvent': 'PyQt6.QtGui', 'QKeySequence': 'PyQt6.QtGui', 'QKeySequenceEdit': 'PyQt6.QtWidgets', 'QLCDNumber': 'PyQt6.QtWidgets', 'QLabel': 'PyQt6.QtWidgets', 'QLayout': 'PyQt6.QtWidgets', 'QLayoutItem': 'PyQt6.QtWidgets', 'QLibrary': 'PyQt6.QtCore', 'QLibraryInfo': 'PyQt6.QtCore', 'QLine': 'PyQt6.QtCore', 'QLineEdit': 'PyQt6.QtWidgets', 'QLineF': 'PyQt6.QtCore', 'QLinearGradient': 'PyQt6.QtGui', 'QListView': 'PyQt6.QtWidgets', 'QListWidget': 'PyQt6.QtWidgets', 'QListWidgetItem': 'PyQt6.QtWidgets', 'QLocalServer': 'PyQt6.QtNetwork', 'QLocalSocket': 'PyQt6.QtNetwork', 'QLocale': 'PyQt6.QtCore', 'QLocationPermission': 'PyQt6.QtCore', 'QLockFile': 'PyQt6.QtCore', 'QLoggingCategory': 'PyQt6.QtCore', 'QMainWindow': 'PyQt6.QtWidgets', 'QMargins': 'PyQt6.QtCore', 'QMarginsF': 'PyQt6.QtCore', 'QMatrix2x2': 'PyQt6.QtGui', 'QMatrix2x3': 'PyQt6.QtGui', 'QMatrix2x4': 'PyQt6.QtGui', 'QMatrix3x2': 'PyQt6.QtGui', 'QMatrix3x3': 'PyQt6.QtGui', 'QMatrix3x4': 'PyQt6.QtGui', 'QMatrix4x2': 'PyQt6.QtGui', 'QMatrix4x3': 'PyQt6.QtGui', 'QMatrix4x4': 'PyQt6.QtGui', 'QMdiArea': 'PyQt6.QtWidgets', 'QMdiSubWindow': 'PyQt6.QtWidgets', 'QMediaCaptureSession': 'PyQt6.QtMultimedia', 'QMediaDevices': 'PyQt6.QtMultimedia', 'QMediaFormat': 'PyQt6.QtMultimedia', 'QMediaMetaData': 'PyQt6.QtMultimedia', 'QMediaPlayer': 'PyQt6.QtMultimedia', 'QMediaRecorder': 'PyQt6.QtMultimedia', 'QMediaTimeRange': 'PyQt6.QtMultimedia', 'QMenu': 'PyQt6.QtWidgets', 'QMenuBar': 'PyQt6.QtWidgets', 'QMessageAuthenticationCode': 'PyQt6.QtCore', 'QMessageBox': 'PyQt6.QtWidgets', 'QMessageLogContext': 'PyQt6.QtCore', 'QMessageLogger': 'PyQt6.QtCore', 'QMetaClassInfo': 'PyQt6.QtCore', 'QMetaEnum': 'PyQt6.QtCore', 'QMetaMethod': 'PyQt6.QtCore', 'QMetaObject': 'PyQt6.QtCore', 'QMetaProperty': 'PyQt6.QtCore', 'QMetaType': 'PyQt6.QtCore', 'QMicrophonePermission': 'PyQt6.QtCore', 'QMimeData': 'PyQt6.QtCore', 'QMimeDatabase': 'PyQt6.QtCore', 'QMimeType': 'PyQt6.QtCore', 'QModelIndex': 'PyQt6.QtCore', 'QModelRoleData': 'PyQt6.QtCore', 'QModelRoleDataSpan': 'PyQt6.QtCore', 'QMouseEvent': 'PyQt6.QtGui', 'QMoveEvent': 'PyQt6.QtGui', 'QMovie': 'PyQt6.QtGui', 'QMutex': 'PyQt6.QtCore', 'QMutexLocker': 'PyQt6.QtCore', 'QNativeGestureEvent': 'PyQt6.QtGui', 'QNativeInterface': 'PyQt6.QtQuick', 'QNativeIpcKey': 'PyQt6.QtCore', 'QNetworkAccessManager': 'PyQt6.QtNetwork', 'QNetworkAddressEntry': 'PyQt6.QtNetwork', 'QNetworkCacheMetaData': 'PyQt6.QtNetwork', 'QNetworkCookie': 'PyQt6.QtNetwork', 'QNetworkCookieJar': 'PyQt6.QtNetwork', 'QNetworkDatagram': 'PyQt6.QtNetwork', 'QNetworkDiskCache': 'PyQt6.QtNetwork', 'QNetworkInformation': 'PyQt6.QtNetwork', 'QNetworkInterface': 'PyQt6.QtNetwork', 'QNetworkProxy': 'PyQt6.QtNetwork', 'QNetworkProxyFactory': 'PyQt6.QtNetwork', 'QNetworkProxyQuery': 'PyQt6.QtNetwork', 'QNetworkReply': 'PyQt6.QtNetwork', 'QNetworkRequest': 'PyQt6.QtNetwork', 'QObject': 'PyQt6.QtCore', 'QObjectCleanupHandler': 'PyQt6.QtCore', 'QOcspCertificateStatus': 'PyQt6.QtNetwork', 'QOcspResponse': 'PyQt6.QtNetwork', 'QOcspRevocationReason': 'PyQt6.QtNetwork', 'QOffscreenSurface': 'PyQt6.QtGui', 'QOpenGLBuffer': 'PyQt6.QtOpenGL', 'QOpenGLContext': 'PyQt6.QtGui', 'QOpenGLContextGroup': 'PyQt6.QtGui', 'QOpenGLDebugLogger': 'PyQt6.QtOpenGL', 'QOpenGLDebugMessage': 'PyQt6.QtOpenGL', 'QOpenGLFramebufferObject': 'PyQt6.QtOpenGL', 'QOpenGLFramebufferObjectFormat': 'PyQt6.QtOpenGL', 'QOpenGLFunctions_2_0': 'PyQt6.QtOpenGL', 'QOpenGLFunctions_2_1': 'PyQt6.QtOpenGL', 'QOpenGLFunctions_4_1_Core': 'PyQt6.QtOpenGL', 'QOpenGLPaintDevice': 'PyQt6.QtOpenGL', 'QOpenGLPixelTransferOptions': 'PyQt6.QtOpenGL', 'QOpenGLShader': 'PyQt6.QtOpenGL', 'QOpenGLShaderProgram': 'PyQt6.QtOpenGL', 'QOpenGLTexture': 'PyQt6.QtOpenGL', 'QOpenGLTextureBlitter': 'PyQt6.QtOpenGL', 'QOpenGLTimeMonitor': 'PyQt6.QtOpenGL', 'QOpenGLTimerQuery': 'PyQt6.QtOpenGL', 'QOpenGLVersionFunctionsFactory': 'PyQt6.QtOpenGL', 'QOpenGLVersionProfile': 'PyQt6.QtOpenGL', 'QOpenGLVertexArrayObject': 'PyQt6.QtOpenGL', 'QOpenGLWidget': 'PyQt6.QtOpenGLWidgets', 'QOpenGLWindow': 'PyQt6.QtOpenGL', 'QOperatingSystemVersion': 'PyQt6.QtCore', 'QOperatingSystemVersionBase': 'PyQt6.QtCore', 'QPageLayout': 'PyQt6.QtGui', 'QPageRanges': 'PyQt6.QtGui', 'QPageSetupDialog': 'PyQt6.QtPrintSupport', 'QPageSize': 'PyQt6.QtGui', 'QPagedPaintDevice': 'PyQt6.QtGui', 'QPaintDevice': 'PyQt6.QtGui', 'QPaintDeviceWindow': 'PyQt6.QtGui', 'QPaintEngine': 'PyQt6.QtGui', 'QPaintEngineState': 'PyQt6.QtGui', 'QPaintEvent': 'PyQt6.QtGui', 'QPainter': 'PyQt6.QtGui', 'QPainterPath': 'PyQt6.QtGui', 'QPainterPathStroker': 'PyQt6.QtGui', 'QPalette': 'PyQt6.QtGui', 'QPanGesture': 'PyQt6.QtWidgets', 'QParallelAnimationGroup': 'PyQt6.QtCore', 'QPasswordDigestor': 'PyQt6.QtNetwork', 'QPauseAnimation': 'PyQt6.QtCore', 'QPdfWriter': 'PyQt6.QtGui', 'QPen': 'PyQt6.QtGui', 'QPermission': 'PyQt6.QtCore', 'QPersistentModelIndex': 'PyQt6.QtCore', 'QPicture': 'PyQt6.QtGui', 'QPinchGesture': 'PyQt6.QtWidgets', 'QPixelFormat': 'PyQt6.QtGui', 'QPixmap': 'PyQt6.QtGui', 'QPixmapCache': 'PyQt6.QtGui', 'QPlainTextDocumentLayout': 'PyQt6.QtWidgets', 'QPlainTextEdit': 'PyQt6.QtWidgets', 'QPlatformSurfaceEvent': 'PyQt6.QtGui', 'QPluginLoader': 'PyQt6.QtCore', 'QPoint': 'PyQt6.QtCore', 'QPointF': 'PyQt6.QtCore', 'QPointerEvent': 'PyQt6.QtGui', 'QPointingDevice': 'PyQt6.QtGui', 'QPointingDeviceUniqueId': 'PyQt6.QtGui', 'QPolygon': 'PyQt6.QtGui', 'QPolygonF': 'PyQt6.QtGui', 'QPrintDialog': 'PyQt6.QtPrintSupport', 'QPrintEngine': 'PyQt6.QtPrintSupport', 'QPrintPreviewDialog': 'PyQt6.QtPrintSupport', 'QPrintPreviewWidget': 'PyQt6.QtPrintSupport', 'QPrinter': 'PyQt6.QtPrintSupport', 'QPrinterInfo': 'PyQt6.QtPrintSupport', 'QProcess': 'PyQt6.QtCore', 'QProcessEnvironment': 'PyQt6.QtCore', 'QProgressBar': 'PyQt6.QtWidgets', 'QProgressDialog': 'PyQt6.QtWidgets', 'QPropertyAnimation': 'PyQt6.QtCore', 'QProxyStyle': 'PyQt6.QtWidgets', 'QPushButton': 'PyQt6.QtWidgets', 'QQuaternion': 'PyQt6.QtGui', 'QQuickAsyncImageProvider': 'PyQt6.QtQuick', 'QQuickCloseEvent': 'PyQt6.QtQuick', 'QQuickFramebufferObject': 'PyQt6.QtQuick', 'QQuickGraphicsConfiguration': 'PyQt6.QtQuick', 'QQuickGraphicsDevice': 'PyQt6.QtQuick', 'QQuickImageProvider': 'PyQt6.QtQuick', 'QQuickImageResponse': 'PyQt6.QtQuick', 'QQuickItem': 'PyQt6.QtQuick', 'QQuickItemGrabResult': 'PyQt6.QtQuick', 'QQuickPaintedItem': 'PyQt6.QtQuick', 'QQuickRenderControl': 'PyQt6.QtQuick', 'QQuickRenderTarget': 'PyQt6.QtQuick', 'QQuickTextDocument': 'PyQt6.QtQuick', 'QQuickTextureFactory': 'PyQt6.QtQuick', 'QQuickView': 'PyQt6.QtQuick', 'QQuickWindow': 'PyQt6.QtQuick', 'QRadialGradient': 'PyQt6.QtGui', 'QRadioButton': 'PyQt6.QtWidgets', 'QRandomGenerator': 'PyQt6.QtCore', 'QRasterWindow': 'PyQt6.QtGui', 'QRawFont': 'PyQt6.QtGui', 'QReadLocker': 'PyQt6.QtCore', 'QReadWriteLock': 'PyQt6.QtCore', 'QRect': 'PyQt6.QtCore', 'QRectF': 'PyQt6.QtCore', 'QRecursiveMutex': 'PyQt6.QtCore', 'QRegion': 'PyQt6.QtGui', 'QRegularExpression': 'PyQt6.QtCore', 'QRegularExpressionMatch': 'PyQt6.QtCore', 'QRegularExpressionMatchIterator': 'PyQt6.QtCore', 'QRegularExpressionValidator': 'PyQt6.QtGui', 'QResizeEvent': 'PyQt6.QtGui', 'QResource': 'PyQt6.QtCore', 'QRgba64': 'PyQt6.QtGui', 'QRubberBand': 'PyQt6.QtWidgets', 'QRunnable': 'PyQt6.QtCore', 'QSGBasicGeometryNode': 'PyQt6.QtQuick', 'QSGClipNode': 'PyQt6.QtQuick', 'QSGDynamicTexture': 'PyQt6.QtQuick', 'QSGFlatColorMaterial': 'PyQt6.QtQuick', 'QSGGeometry': 'PyQt6.QtQuick', 'QSGGeometryNode': 'PyQt6.QtQuick', 'QSGImageNode': 'PyQt6.QtQuick', 'QSGMaterial': 'PyQt6.QtQuick', 'QSGMaterialShader': 'PyQt6.QtQuick', 'QSGMaterialType': 'PyQt6.QtQuick', 'QSGNode': 'PyQt6.QtQuick', 'QSGOpacityNode': 'PyQt6.QtQuick', 'QSGOpaqueTextureMaterial': 'PyQt6.QtQuick', 'QSGRectangleNode': 'PyQt6.QtQuick', 'QSGRenderNode': 'PyQt6.QtQuick', 'QSGRendererInterface': 'PyQt6.QtQuick', 'QSGSimpleRectNode': 'PyQt6.QtQuick', 'QSGSimpleTextureNode': 'PyQt6.QtQuick', 'QSGTextNode': 'PyQt6.QtQuick', 'QSGTexture': 'PyQt6.QtQuick', 'QSGTextureMaterial': 'PyQt6.QtQuick', 'QSGTextureProvider': 'PyQt6.QtQuick', 'QSGTransformNode': 'PyQt6.QtQuick', 'QSGVertexColorMaterial': 'PyQt6.QtQuick', 'QSaveFile': 'PyQt6.QtCore', 'QScreen': 'PyQt6.QtGui', 'QScreenCapture': 'PyQt6.QtMultimedia', 'QScrollArea': 'PyQt6.QtWidgets', 'QScrollBar': 'PyQt6.QtWidgets', 'QScrollEvent': 'PyQt6.QtGui', 'QScrollPrepareEvent': 'PyQt6.QtGui', 'QScroller': 'PyQt6.QtWidgets', 'QScrollerProperties': 'PyQt6.QtWidgets', 'QSemaphore': 'PyQt6.QtCore', 'QSemaphoreReleaser': 'PyQt6.QtCore', 'QSequentialAnimationGroup': 'PyQt6.QtCore', 'QSessionManager': 'PyQt6.QtGui', 'QSettings': 'PyQt6.QtCore', 'QSharedMemory': 'PyQt6.QtCore', 'QShortcut': 'PyQt6.QtGui', 'QShortcutEvent': 'PyQt6.QtGui', 'QShowEvent': 'PyQt6.QtGui', 'QSignalBlocker': 'PyQt6.QtCore', 'QSignalMapper': 'PyQt6.QtCore', 'QSinglePointEvent': 'PyQt6.QtGui', 'QSize': 'PyQt6.QtCore', 'QSizeF': 'PyQt6.QtCore', 'QSizeGrip': 'PyQt6.QtWidgets', 'QSizePolicy': 'PyQt6.QtWidgets', 'QSlider': 'PyQt6.QtWidgets', 'QSocketNotifier': 'PyQt6.QtCore', 'QSortFilterProxyModel': 'PyQt6.QtCore', 'QSoundEffect': 'PyQt6.QtMultimedia', 'QSpacerItem': 'PyQt6.QtWidgets', 'QSpinBox': 'PyQt6.QtWidgets', 'QSplashScreen': 'PyQt6.QtWidgets', 'QSplitter': 'PyQt6.QtWidgets', 'QSplitterHandle': 'PyQt6.QtWidgets', 'QSsl': 'PyQt6.QtNetwork', 'QSslCertificate': 'PyQt6.QtNetwork', 'QSslCertificateExtension': 'PyQt6.QtNetwork', 'QSslCipher': 'PyQt6.QtNetwork', 'QSslConfiguration': 'PyQt6.QtNetwork', 'QSslDiffieHellmanParameters': 'PyQt6.QtNetwork', 'QSslEllipticCurve': 'PyQt6.QtNetwork', 'QSslError': 'PyQt6.QtNetwork', 'QSslKey': 'PyQt6.QtNetwork', 'QSslPreSharedKeyAuthenticator': 'PyQt6.QtNetwork', 'QSslServer': 'PyQt6.QtNetwork', 'QSslSocket': 'PyQt6.QtNetwork', 'QStackedLayout': 'PyQt6.QtWidgets', 'QStackedWidget': 'PyQt6.QtWidgets', 'QStandardItem': 'PyQt6.QtGui', 'QStandardItemModel': 'PyQt6.QtGui', 'QStandardPaths': 'PyQt6.QtCore', 'QStaticText': 'PyQt6.QtGui', 'QStatusBar': 'PyQt6.QtWidgets', 'QStatusTipEvent': 'PyQt6.QtGui', 'QStorageInfo': 'PyQt6.QtCore', 'QStringConverter': 'PyQt6.QtCore', 'QStringConverterBase': 'PyQt6.QtCore', 'QStringDecoder': 'PyQt6.QtCore', 'QStringEncoder': 'PyQt6.QtCore', 'QStringListModel': 'PyQt6.QtCore', 'QStyle': 'PyQt6.QtWidgets', 'QStyleFactory': 'PyQt6.QtWidgets', 'QStyleHintReturn': 'PyQt6.QtWidgets', 'QStyleHintReturnMask': 'PyQt6.QtWidgets', 'QStyleHintReturnVariant': 'PyQt6.QtWidgets', 'QStyleHints': 'PyQt6.QtGui', 'QStyleOption': 'PyQt6.QtWidgets', 'QStyleOptionButton': 'PyQt6.QtWidgets', 'QStyleOptionComboBox': 'PyQt6.QtWidgets', 'QStyleOptionComplex': 'PyQt6.QtWidgets', 'QStyleOptionDockWidget': 'PyQt6.QtWidgets', 'QStyleOptionFocusRect': 'PyQt6.QtWidgets', 'QStyleOptionFrame': 'PyQt6.QtWidgets', 'QStyleOptionGraphicsItem': 'PyQt6.QtWidgets', 'QStyleOptionGroupBox': 'PyQt6.QtWidgets', 'QStyleOptionHeader': 'PyQt6.QtWidgets', 'QStyleOptionHeaderV2': 'PyQt6.QtWidgets', 'QStyleOptionMenuItem': 'PyQt6.QtWidgets', 'QStyleOptionProgressBar': 'PyQt6.QtWidgets', 'QStyleOptionRubberBand': 'PyQt6.QtWidgets', 'QStyleOptionSizeGrip': 'PyQt6.QtWidgets', 'QStyleOptionSlider': 'PyQt6.QtWidgets', 'QStyleOptionSpinBox': 'PyQt6.QtWidgets', 'QStyleOptionTab': 'PyQt6.QtWidgets', 'QStyleOptionTabBarBase': 'PyQt6.QtWidgets', 'QStyleOptionTabWidgetFrame': 'PyQt6.QtWidgets', 'QStyleOptionTitleBar': 'PyQt6.QtWidgets', 'QStyleOptionToolBar': 'PyQt6.QtWidgets', 'QStyleOptionToolBox': 'PyQt6.QtWidgets', 'QStyleOptionToolButton': 'PyQt6.QtWidgets', 'QStyleOptionViewItem': 'PyQt6.QtWidgets', 'QStylePainter': 'PyQt6.QtWidgets', 'QStyledItemDelegate': 'PyQt6.QtWidgets', 'QSurface': 'PyQt6.QtGui', 'QSurfaceFormat': 'PyQt6.QtGui', 'QSvgGenerator': 'PyQt6.QtSvg', 'QSvgRenderer': 'PyQt6.QtSvg', 'QSwipeGesture': 'PyQt6.QtWidgets', 'QSyntaxHighlighter': 'PyQt6.QtGui', 'QSysInfo': 'PyQt6.QtCore', 'QSystemSemaphore': 'PyQt6.QtCore', 'QSystemTrayIcon': 'PyQt6.QtWidgets', 'QT_TRANSLATE_NOOP': 'PyQt6.QtCore', 'QT_TR_NOOP': 'PyQt6.QtCore', 'QT_VERSION': 'PyQt6.QtCore', 'QT_VERSION_STR': 'PyQt6.QtCore', 'QTabBar': 'PyQt6.QtWidgets', 'QTabWidget': 'PyQt6.QtWidgets', 'QTableView': 'PyQt6.QtWidgets', 'QTableWidget': 'PyQt6.QtWidgets', 'QTableWidgetItem': 'PyQt6.QtWidgets', 'QTableWidgetSelectionRange': 'PyQt6.QtWidgets', 'QTabletEvent': 'PyQt6.QtGui', 'QTapAndHoldGesture': 'PyQt6.QtWidgets', 'QTapGesture': 'PyQt6.QtWidgets', 'QTcpServer': 'PyQt6.QtNetwork', 'QTcpSocket': 'PyQt6.QtNetwork', 'QTemporaryDir': 'PyQt6.QtCore', 'QTemporaryFile': 'PyQt6.QtCore', 'QTextBlock': 'PyQt6.QtGui', 'QTextBlockFormat': 'PyQt6.QtGui', 'QTextBlockGroup': 'PyQt6.QtGui', 'QTextBlockUserData': 'PyQt6.QtGui', 'QTextBoundaryFinder': 'PyQt6.QtCore', 'QTextBrowser': 'PyQt6.QtWidgets', 'QTextCharFormat': 'PyQt6.QtGui', 'QTextCursor': 'PyQt6.QtGui', 'QTextDocument': 'PyQt6.QtGui', 'QTextDocumentFragment': 'PyQt6.QtGui', 'QTextDocumentWriter': 'PyQt6.QtGui', 'QTextEdit': 'PyQt6.QtWidgets', 'QTextFormat': 'PyQt6.QtGui', 'QTextFragment': 'PyQt6.QtGui', 'QTextFrame': 'PyQt6.QtGui', 'QTextFrameFormat': 'PyQt6.QtGui', 'QTextImageFormat': 'PyQt6.QtGui', 'QTextInlineObject': 'PyQt6.QtGui', 'QTextItem': 'PyQt6.QtGui', 'QTextLayout': 'PyQt6.QtGui', 'QTextLength': 'PyQt6.QtGui', 'QTextLine': 'PyQt6.QtGui', 'QTextList': 'PyQt6.QtGui', 'QTextListFormat': 'PyQt6.QtGui', 'QTextObject': 'PyQt6.QtGui', 'QTextObjectInterface': 'PyQt6.QtGui', 'QTextOption': 'PyQt6.QtGui', 'QTextStream': 'PyQt6.QtCore', 'QTextStreamManipulator': 'PyQt6.QtCore', 'QTextTable': 'PyQt6.QtGui', 'QTextTableCell': 'PyQt6.QtGui', 'QTextTableCellFormat': 'PyQt6.QtGui', 'QTextTableFormat': 'PyQt6.QtGui', 'QTextToSpeech': 'PyQt6.QtTextToSpeech', 'QThread': 'PyQt6.QtCore', 'QThreadPool': 'PyQt6.QtCore', 'QTime': 'PyQt6.QtCore', 'QTimeEdit': 'PyQt6.QtWidgets', 'QTimeLine': 'PyQt6.QtCore', 'QTimeZone': 'PyQt6.QtCore', 'QTimer': 'PyQt6.QtCore', 'QTimerEvent': 'PyQt6.QtCore', 'QToolBar': 'PyQt6.QtWidgets', 'QToolBox': 'PyQt6.QtWidgets', 'QToolButton': 'PyQt6.QtWidgets', 'QToolTip': 'PyQt6.QtWidgets', 'QTouchEvent': 'PyQt6.QtGui', 'QTransform': 'PyQt6.QtGui', 'QTranslator': 'PyQt6.QtCore', 'QTransposeProxyModel': 'PyQt6.QtCore', 'QTreeView': 'PyQt6.QtWidgets', 'QTreeWidget': 'PyQt6.QtWidgets', 'QTreeWidgetItem': 'PyQt6.QtWidgets', 'QTreeWidgetItemIterator': 'PyQt6.QtWidgets', 'QTypeRevision': 'PyQt6.QtCore', 'QUdpSocket': 'PyQt6.QtNetwork', 'QUndoCommand': 'PyQt6.QtGui', 'QUndoGroup': 'PyQt6.QtGui', 'QUndoStack': 'PyQt6.QtGui', 'QUndoView': 'PyQt6.QtWidgets', 'QUrl': 'PyQt6.QtCore', 'QUrlQuery': 'PyQt6.QtCore', 'QUuid': 'PyQt6.QtCore', 'QVBoxLayout': 'PyQt6.QtWidgets', 'QValidator': 'PyQt6.QtGui', 'QVariant': 'PyQt6.QtCore', 'QVariantAnimation': 'PyQt6.QtCore', 'QVector2D': 'PyQt6.QtGui', 'QVector3D': 'PyQt6.QtGui', 'QVector4D': 'PyQt6.QtGui', 'QVersionNumber': 'PyQt6.QtCore', 'QVideoFrame': 'PyQt6.QtMultimedia', 'QVideoFrameFormat': 'PyQt6.QtMultimedia', 'QVideoSink': 'PyQt6.QtMultimedia', 'QVideoWidget': 'PyQt6.QtMultimediaWidgets', 'QVoice': 'PyQt6.QtTextToSpeech', 'QWIDGETSIZE_MAX': 'PyQt6.QtWidgets', 'QWaitCondition': 'PyQt6.QtCore', 'QWhatsThis': 'PyQt6.QtWidgets', 'QWhatsThisClickedEvent': 'PyQt6.QtGui', 'QWheelEvent': 'PyQt6.QtGui', 'QWidget': 'PyQt6.QtWidgets', 'QWidgetAction': 'PyQt6.QtWidgets', 'QWidgetItem': 'PyQt6.QtWidgets', 'QWindow': 'PyQt6.QtGui', 'QWindowCapture': 'PyQt6.QtMultimedia', 'QWindowStateChangeEvent': 'PyQt6.QtGui', 'QWizard': 'PyQt6.QtWidgets', 'QWizardPage': 'PyQt6.QtWidgets', 'QWriteLocker': 'PyQt6.QtCore', 'QXmlStreamAttribute': 'PyQt6.QtCore', 'QXmlStreamAttributes': 'PyQt6.QtCore', 'QXmlStreamEntityDeclaration': 'PyQt6.QtCore', 'QXmlStreamEntityResolver': 'PyQt6.QtCore', 'QXmlStreamNamespaceDeclaration': 'PyQt6.QtCore', 'QXmlStreamNotationDeclaration': 'PyQt6.QtCore', 'QXmlStreamReader': 'PyQt6.QtCore', 'QXmlStreamWriter': 'PyQt6.QtCore', 'Q_ARG': 'PyQt6.QtCore', 'Q_RETURN_ARG': 'PyQt6.QtCore', 'Qt': 'PyQt6.QtCore', 'QtCore': 'PyQt6.QtCore', 'QtGui': 'PyQt6.QtGui', 'QtMsgType': 'PyQt6.QtCore', 'QtMultimedia': 'PyQt6.QtMultimedia', 'QtMultimediaWidgets': 'PyQt6.QtMultimediaWidgets', 'QtNetwork': 'PyQt6.QtNetwork', 'QtOpenGL': 'PyQt6.QtOpenGL', 'QtOpenGLWidgets': 'PyQt6.QtOpenGLWidgets', 'QtPrintSupport': 'PyQt6.QtPrintSupport', 'QtQuick': 'PyQt6.QtQuick', 'QtSvg': 'PyQt6.QtSvg', 'QtTextToSpeech': 'PyQt6.QtTextToSpeech', 'QtVideo': 'PyQt6.QtMultimedia', 'QtWidgets': 'PyQt6.QtWidgets', 'pyqtBoundSignal': 'PyQt6.QtCore', 'pyqtClassInfo': 'PyQt6.QtCore', 'pyqtEnum': 'PyQt6.QtCore', 'pyqtPickleProtocol': 'PyQt6.QtCore', 'pyqtProperty': 'PyQt6.QtCore', 'pyqtRemoveInputHook': 'PyQt6.QtCore', 'pyqtRestoreInputHook': 'PyQt6.QtCore', 'pyqtSetPickleProtocol': 'PyQt6.QtCore', 'pyqtSignal': 'PyQt6.QtCore', 'pyqtSlot': 'PyQt6.QtCore', 'qAbs': 'PyQt6.QtCore', 'qAddPostRoutine': 'PyQt6.QtCore', 'qAddPreRoutine': 'PyQt6.QtCore', 'qAlpha': 'PyQt6.QtGui', 'qBlue': 'PyQt6.QtGui', 'qChecksum': 'PyQt6.QtCore', 'qCompress': 'PyQt6.QtCore', 'qCritical': 'PyQt6.QtCore', 'qDebug': 'PyQt6.QtCore', 'qDrawBorderPixmap': 'PyQt6.QtWidgets', 'qDrawPlainRect': 'PyQt6.QtWidgets', 'qDrawPlainRoundedRect': 'PyQt6.QtWidgets', 'qDrawShadeLine': 'PyQt6.QtWidgets', 'qDrawShadePanel': 'PyQt6.QtWidgets', 'qDrawShadeRect': 'PyQt6.QtWidgets', 'qDrawWinButton': 'PyQt6.QtWidgets', 'qDrawWinPanel': 'PyQt6.QtWidgets', 'qEnvironmentVariable': 'PyQt6.QtCore', 'qEnvironmentVariableIntValue': 'PyQt6.QtCore', 'qEnvironmentVariableIsEmpty': 'PyQt6.QtCore', 'qEnvironmentVariableIsSet': 'PyQt6.QtCore', 'qFatal': 'PyQt6.QtCore', 'qFloatDistance': 'PyQt6.QtCore', 'qFormatLogMessage': 'PyQt6.QtCore', 'qFuzzyCompare': 'PyQt6.QtCore', 'qFuzzyIsNull': 'PyQt6.QtCore', 'qGray': 'PyQt6.QtGui', 'qGreen': 'PyQt6.QtGui', 'qInf': 'PyQt6.QtCore', 'qInfo': 'PyQt6.QtCore', 'qInstallMessageHandler': 'PyQt6.QtCore', 'qIsFinite': 'PyQt6.QtCore', 'qIsInf': 'PyQt6.QtCore', 'qIsNaN': 'PyQt6.QtCore', 'qPixelFormatAlpha': 'PyQt6.QtGui', 'qPixelFormatCmyk': 'PyQt6.QtGui', 'qPixelFormatGrayscale': 'PyQt6.QtGui', 'qPixelFormatHsl': 'PyQt6.QtGui', 'qPixelFormatHsv': 'PyQt6.QtGui', 'qPixelFormatRgba': 'PyQt6.QtGui', 'qPixelFormatYuv': 'PyQt6.QtGui', 'qPremultiply': 'PyQt6.QtGui', 'qQNaN': 'PyQt6.QtCore', 'qRed': 'PyQt6.QtGui', 'qRegisterResourceData': 'PyQt6.QtCore', 'qRemovePostRoutine': 'PyQt6.QtCore', 'qRgb': 'PyQt6.QtGui', 'qRgba': 'PyQt6.QtGui', 'qRgba64': 'PyQt6.QtGui', 'qRound': 'PyQt6.QtCore', 'qRound64': 'PyQt6.QtCore', 'qSNaN': 'PyQt6.QtCore', 'qSetFieldWidth': 'PyQt6.QtCore', 'qSetMessagePattern': 'PyQt6.QtCore', 'qSetPadChar': 'PyQt6.QtCore', 'qSetRealNumberPrecision': 'PyQt6.QtCore', 'qUncompress': 'PyQt6.QtCore', 'qUnpremultiply': 'PyQt6.QtGui', 'qUnregisterResourceData': 'PyQt6.QtCore', 'qVersion': 'PyQt6.QtCore', 'qWarning': 'PyQt6.QtCore', 'qYieldCpu': 'PyQt6.QtCore', 'qt_set_sequence_auto_mnemonic': 'PyQt6.QtGui', 'sip': 'PyQt6.sip'} module_names = frozenset(('QtCore', 'QtGui', 'QtWidgets', 'QtNetwork', 'QtSvg', 'QtPrintSupport', 'QtOpenGL', 'QtOpenGLWidgets', 'QtQuick', 'QtMultimedia', 'QtMultimediaWidgets', 'QtTextToSpeech', 'sip') )
28,190
Python
.py
784
33.961735
53
0.743852
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,827
__main__.py
kovidgoyal_calibre/src/qt/__main__.py
#!/usr/bin/env python # vim:fileencoding=utf-8 # License: GPL v3 Copyright: 2021, Kovid Goyal <kovid at kovidgoyal.net> import importlib import os from pprint import pprint QT_WRAPPER = 'PyQt6' base = os.path.dirname(os.path.abspath(__file__)) module_lists = { 'core': ( 'QtCore', 'QtGui', 'QtWidgets', 'QtNetwork', 'QtSvg', 'QtPrintSupport', 'QtOpenGL', 'QtOpenGLWidgets', 'QtQuick', 'QtMultimedia', 'QtMultimediaWidgets', 'QtTextToSpeech', ), 'webengine': ( 'QtWebEngineCore', 'QtWebEngineWidgets', ), 'dbus': ( 'QtDBus', ) } def scan(name): module_names = module_lists[name] name_map = {} types = [] for mod_name in module_names: mod = importlib.import_module(f'{QT_WRAPPER}.{mod_name}') full_name = name_map[mod_name] = mod.__name__ types.append(f'import {full_name}') for obj_name in sorted(dir(mod)): if not obj_name.startswith('_') and obj_name not in name_map: name_map[obj_name] = full_name types.append(f'{obj_name} = {full_name}.{obj_name}') with open(f'{base}/{name}.py', 'w') as f: print('# autogenerated by __main__.py do not edit', file=f) print(f'from .{name}_name_map import module_names, name_map', file=f) print('''from .loader import dynamic_load already_imported = {} qt_modules = {} def __getattr__(name): return dynamic_load(name, name_map, already_imported, qt_modules, module_names) ''', file=f) with open(f'{base}/{name}.pyi', 'w') as f: print('# autogenerated by __main__.py do not edit', file=f) f.write('\n'.join(types)) if name == 'core': module_names += ('sip',) mod = importlib.import_module(f'{QT_WRAPPER}.sip') name_map['sip'] = mod.__name__ with open(f'{base}/{name}_name_map.py', 'w') as f: print('# autogenerated by __main__.py do not edit', file=f) print('name_map =', end=' ', file=f) pprint(name_map, stream=f) print('module_names = frozenset(', end='', file=f) pprint(module_names, stream=f) print(')', file=f) top_level_module_names = () for name in module_lists.keys(): top_level_module_names += module_lists[name] scan(name) with open(f'{base}/__init__.py', 'w') as f: print('# autogenerated by __main__.py do not edit', file=f) print(f'{top_level_module_names=}', file=f) print(f''' def __getattr__(name): if name in top_level_module_names: import importlib return importlib.import_module("{QT_WRAPPER}." + name) raise AttributeError(name) ''', file=f)
2,720
Python
.py
80
27.725
83
0.592395
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,828
dbus.pyi
kovidgoyal_calibre/src/qt/dbus.pyi
# autogenerated by __main__.py do not edit import PyQt6.QtDBus QDBus = PyQt6.QtDBus.QDBus QDBusAbstractAdaptor = PyQt6.QtDBus.QDBusAbstractAdaptor QDBusAbstractInterface = PyQt6.QtDBus.QDBusAbstractInterface QDBusArgument = PyQt6.QtDBus.QDBusArgument QDBusConnection = PyQt6.QtDBus.QDBusConnection QDBusConnectionInterface = PyQt6.QtDBus.QDBusConnectionInterface QDBusError = PyQt6.QtDBus.QDBusError QDBusInterface = PyQt6.QtDBus.QDBusInterface QDBusMessage = PyQt6.QtDBus.QDBusMessage QDBusObjectPath = PyQt6.QtDBus.QDBusObjectPath QDBusPendingCall = PyQt6.QtDBus.QDBusPendingCall QDBusPendingCallWatcher = PyQt6.QtDBus.QDBusPendingCallWatcher QDBusPendingReply = PyQt6.QtDBus.QDBusPendingReply QDBusReply = PyQt6.QtDBus.QDBusReply QDBusServiceWatcher = PyQt6.QtDBus.QDBusServiceWatcher QDBusSignature = PyQt6.QtDBus.QDBusSignature QDBusUnixFileDescriptor = PyQt6.QtDBus.QDBusUnixFileDescriptor QDBusVariant = PyQt6.QtDBus.QDBusVariant
936
Python
.py
20
45.85
64
0.886587
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,829
webengine_name_map.py
kovidgoyal_calibre/src/qt/webengine_name_map.py
# autogenerated by __main__.py do not edit name_map = {'PYQT_WEBENGINE_VERSION': 'PyQt6.QtWebEngineCore', 'PYQT_WEBENGINE_VERSION_STR': 'PyQt6.QtWebEngineCore', 'QWebEngineCertificateError': 'PyQt6.QtWebEngineCore', 'QWebEngineClientCertificateSelection': 'PyQt6.QtWebEngineCore', 'QWebEngineClientCertificateStore': 'PyQt6.QtWebEngineCore', 'QWebEngineContextMenuRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineCookieStore': 'PyQt6.QtWebEngineCore', 'QWebEngineDesktopMediaRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineDownloadRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineFileSystemAccessRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineFindTextResult': 'PyQt6.QtWebEngineCore', 'QWebEngineFullScreenRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineGlobalSettings': 'PyQt6.QtWebEngineCore', 'QWebEngineHistory': 'PyQt6.QtWebEngineCore', 'QWebEngineHistoryItem': 'PyQt6.QtWebEngineCore', 'QWebEngineHistoryModel': 'PyQt6.QtWebEngineCore', 'QWebEngineHttpRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineLoadingInfo': 'PyQt6.QtWebEngineCore', 'QWebEngineNavigationRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineNewWindowRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineNotification': 'PyQt6.QtWebEngineCore', 'QWebEnginePage': 'PyQt6.QtWebEngineCore', 'QWebEngineProfile': 'PyQt6.QtWebEngineCore', 'QWebEngineQuotaRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineRegisterProtocolHandlerRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineScript': 'PyQt6.QtWebEngineCore', 'QWebEngineScriptCollection': 'PyQt6.QtWebEngineCore', 'QWebEngineSettings': 'PyQt6.QtWebEngineCore', 'QWebEngineUrlRequestInfo': 'PyQt6.QtWebEngineCore', 'QWebEngineUrlRequestInterceptor': 'PyQt6.QtWebEngineCore', 'QWebEngineUrlRequestJob': 'PyQt6.QtWebEngineCore', 'QWebEngineUrlScheme': 'PyQt6.QtWebEngineCore', 'QWebEngineUrlSchemeHandler': 'PyQt6.QtWebEngineCore', 'QWebEngineView': 'PyQt6.QtWebEngineWidgets', 'QWebEngineWebAuthPinRequest': 'PyQt6.QtWebEngineCore', 'QWebEngineWebAuthUxRequest': 'PyQt6.QtWebEngineCore', 'QtWebEngineCore': 'PyQt6.QtWebEngineCore', 'QtWebEngineWidgets': 'PyQt6.QtWebEngineWidgets', 'qWebEngineChromiumSecurityPatchVersion': 'PyQt6.QtWebEngineCore', 'qWebEngineChromiumVersion': 'PyQt6.QtWebEngineCore', 'qWebEngineGetDomainAndRegistry': 'PyQt6.QtWebEngineCore', 'qWebEngineProcessName': 'PyQt6.QtWebEngineCore', 'qWebEngineVersion': 'PyQt6.QtWebEngineCore'} module_names = frozenset(('QtWebEngineCore', 'QtWebEngineWidgets') )
2,456
Python
.py
46
51.478261
69
0.824896
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,830
dbus.py
kovidgoyal_calibre/src/qt/dbus.py
# autogenerated by __main__.py do not edit from .dbus_name_map import module_names, name_map from .loader import dynamic_load already_imported = {} qt_modules = {} def __getattr__(name): return dynamic_load(name, name_map, already_imported, qt_modules, module_names)
273
Python
.py
7
37.142857
83
0.746212
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,831
dbus_name_map.py
kovidgoyal_calibre/src/qt/dbus_name_map.py
# autogenerated by __main__.py do not edit name_map = {'QDBus': 'PyQt6.QtDBus', 'QDBusAbstractAdaptor': 'PyQt6.QtDBus', 'QDBusAbstractInterface': 'PyQt6.QtDBus', 'QDBusArgument': 'PyQt6.QtDBus', 'QDBusConnection': 'PyQt6.QtDBus', 'QDBusConnectionInterface': 'PyQt6.QtDBus', 'QDBusError': 'PyQt6.QtDBus', 'QDBusInterface': 'PyQt6.QtDBus', 'QDBusMessage': 'PyQt6.QtDBus', 'QDBusObjectPath': 'PyQt6.QtDBus', 'QDBusPendingCall': 'PyQt6.QtDBus', 'QDBusPendingCallWatcher': 'PyQt6.QtDBus', 'QDBusPendingReply': 'PyQt6.QtDBus', 'QDBusReply': 'PyQt6.QtDBus', 'QDBusServiceWatcher': 'PyQt6.QtDBus', 'QDBusSignature': 'PyQt6.QtDBus', 'QDBusUnixFileDescriptor': 'PyQt6.QtDBus', 'QDBusVariant': 'PyQt6.QtDBus', 'QtDBus': 'PyQt6.QtDBus'} module_names = frozenset(('QtDBus',) )
782
Python
.py
22
33.727273
44
0.739474
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,832
loader.py
kovidgoyal_calibre/src/qt/loader.py
#!/usr/bin/env python # vim:fileencoding=utf-8 # License: GPL v3 Copyright: 2021, Kovid Goyal <kovid at kovidgoyal.net> import sys from importlib import import_module def dynamic_load(name, name_map, already_imported, qt_modules, module_names=()): ans = already_imported.get(name, already_imported) if ans is not already_imported: return ans mod_name = name_map.get(name) if mod_name is not None: mod = qt_modules.get(mod_name) if mod is None: try: mod = qt_modules[mod_name] = import_module(mod_name) except ImportError as err: mod = qt_modules[mod_name] = False print( 'Failed to import PyQt module:', mod_name, 'with error:', err, file=sys.stderr ) if mod is not False: if name in module_names: q = mod else: q = getattr(mod, name, qt_modules) if q is not qt_modules: already_imported[name] = q return q raise AttributeError(f'The object {name} is not a known Qt object')
1,218
Python
.py
33
25.515152
80
0.545685
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,833
core.pyi
kovidgoyal_calibre/src/qt/core.pyi
# autogenerated by __main__.py do not edit import PyQt6.QtCore PYQT_VERSION = PyQt6.QtCore.PYQT_VERSION PYQT_VERSION_STR = PyQt6.QtCore.PYQT_VERSION_STR QAbstractAnimation = PyQt6.QtCore.QAbstractAnimation QAbstractEventDispatcher = PyQt6.QtCore.QAbstractEventDispatcher QAbstractItemModel = PyQt6.QtCore.QAbstractItemModel QAbstractListModel = PyQt6.QtCore.QAbstractListModel QAbstractNativeEventFilter = PyQt6.QtCore.QAbstractNativeEventFilter QAbstractProxyModel = PyQt6.QtCore.QAbstractProxyModel QAbstractTableModel = PyQt6.QtCore.QAbstractTableModel QAnimationGroup = PyQt6.QtCore.QAnimationGroup QBasicTimer = PyQt6.QtCore.QBasicTimer QBitArray = PyQt6.QtCore.QBitArray QBluetoothPermission = PyQt6.QtCore.QBluetoothPermission QBuffer = PyQt6.QtCore.QBuffer QByteArray = PyQt6.QtCore.QByteArray QByteArrayMatcher = PyQt6.QtCore.QByteArrayMatcher QCalendar = PyQt6.QtCore.QCalendar QCalendarPermission = PyQt6.QtCore.QCalendarPermission QCameraPermission = PyQt6.QtCore.QCameraPermission QCborError = PyQt6.QtCore.QCborError QCborKnownTags = PyQt6.QtCore.QCborKnownTags QCborSimpleType = PyQt6.QtCore.QCborSimpleType QCborStreamReader = PyQt6.QtCore.QCborStreamReader QCborStreamWriter = PyQt6.QtCore.QCborStreamWriter QChildEvent = PyQt6.QtCore.QChildEvent QCollator = PyQt6.QtCore.QCollator QCollatorSortKey = PyQt6.QtCore.QCollatorSortKey QCommandLineOption = PyQt6.QtCore.QCommandLineOption QCommandLineParser = PyQt6.QtCore.QCommandLineParser QConcatenateTablesProxyModel = PyQt6.QtCore.QConcatenateTablesProxyModel QContactsPermission = PyQt6.QtCore.QContactsPermission QCoreApplication = PyQt6.QtCore.QCoreApplication QCryptographicHash = PyQt6.QtCore.QCryptographicHash QDataStream = PyQt6.QtCore.QDataStream QDate = PyQt6.QtCore.QDate QDateTime = PyQt6.QtCore.QDateTime QDeadlineTimer = PyQt6.QtCore.QDeadlineTimer QDir = PyQt6.QtCore.QDir QDirIterator = PyQt6.QtCore.QDirIterator QDynamicPropertyChangeEvent = PyQt6.QtCore.QDynamicPropertyChangeEvent QEasingCurve = PyQt6.QtCore.QEasingCurve QElapsedTimer = PyQt6.QtCore.QElapsedTimer QEvent = PyQt6.QtCore.QEvent QEventLoop = PyQt6.QtCore.QEventLoop QEventLoopLocker = PyQt6.QtCore.QEventLoopLocker QFile = PyQt6.QtCore.QFile QFileDevice = PyQt6.QtCore.QFileDevice QFileInfo = PyQt6.QtCore.QFileInfo QFileSelector = PyQt6.QtCore.QFileSelector QFileSystemWatcher = PyQt6.QtCore.QFileSystemWatcher QGenericArgument = PyQt6.QtCore.QGenericArgument QGenericReturnArgument = PyQt6.QtCore.QGenericReturnArgument QIODevice = PyQt6.QtCore.QIODevice QIODeviceBase = PyQt6.QtCore.QIODeviceBase QIdentityProxyModel = PyQt6.QtCore.QIdentityProxyModel QItemSelection = PyQt6.QtCore.QItemSelection QItemSelectionModel = PyQt6.QtCore.QItemSelectionModel QItemSelectionRange = PyQt6.QtCore.QItemSelectionRange QJsonDocument = PyQt6.QtCore.QJsonDocument QJsonParseError = PyQt6.QtCore.QJsonParseError QJsonValue = PyQt6.QtCore.QJsonValue QKeyCombination = PyQt6.QtCore.QKeyCombination QLibrary = PyQt6.QtCore.QLibrary QLibraryInfo = PyQt6.QtCore.QLibraryInfo QLine = PyQt6.QtCore.QLine QLineF = PyQt6.QtCore.QLineF QLocale = PyQt6.QtCore.QLocale QLocationPermission = PyQt6.QtCore.QLocationPermission QLockFile = PyQt6.QtCore.QLockFile QLoggingCategory = PyQt6.QtCore.QLoggingCategory QMargins = PyQt6.QtCore.QMargins QMarginsF = PyQt6.QtCore.QMarginsF QMessageAuthenticationCode = PyQt6.QtCore.QMessageAuthenticationCode QMessageLogContext = PyQt6.QtCore.QMessageLogContext QMessageLogger = PyQt6.QtCore.QMessageLogger QMetaClassInfo = PyQt6.QtCore.QMetaClassInfo QMetaEnum = PyQt6.QtCore.QMetaEnum QMetaMethod = PyQt6.QtCore.QMetaMethod QMetaObject = PyQt6.QtCore.QMetaObject QMetaProperty = PyQt6.QtCore.QMetaProperty QMetaType = PyQt6.QtCore.QMetaType QMicrophonePermission = PyQt6.QtCore.QMicrophonePermission QMimeData = PyQt6.QtCore.QMimeData QMimeDatabase = PyQt6.QtCore.QMimeDatabase QMimeType = PyQt6.QtCore.QMimeType QModelIndex = PyQt6.QtCore.QModelIndex QModelRoleData = PyQt6.QtCore.QModelRoleData QModelRoleDataSpan = PyQt6.QtCore.QModelRoleDataSpan QMutex = PyQt6.QtCore.QMutex QMutexLocker = PyQt6.QtCore.QMutexLocker QNativeIpcKey = PyQt6.QtCore.QNativeIpcKey QObject = PyQt6.QtCore.QObject QObjectCleanupHandler = PyQt6.QtCore.QObjectCleanupHandler QOperatingSystemVersion = PyQt6.QtCore.QOperatingSystemVersion QOperatingSystemVersionBase = PyQt6.QtCore.QOperatingSystemVersionBase QParallelAnimationGroup = PyQt6.QtCore.QParallelAnimationGroup QPauseAnimation = PyQt6.QtCore.QPauseAnimation QPermission = PyQt6.QtCore.QPermission QPersistentModelIndex = PyQt6.QtCore.QPersistentModelIndex QPluginLoader = PyQt6.QtCore.QPluginLoader QPoint = PyQt6.QtCore.QPoint QPointF = PyQt6.QtCore.QPointF QProcess = PyQt6.QtCore.QProcess QProcessEnvironment = PyQt6.QtCore.QProcessEnvironment QPropertyAnimation = PyQt6.QtCore.QPropertyAnimation QRandomGenerator = PyQt6.QtCore.QRandomGenerator QReadLocker = PyQt6.QtCore.QReadLocker QReadWriteLock = PyQt6.QtCore.QReadWriteLock QRect = PyQt6.QtCore.QRect QRectF = PyQt6.QtCore.QRectF QRecursiveMutex = PyQt6.QtCore.QRecursiveMutex QRegularExpression = PyQt6.QtCore.QRegularExpression QRegularExpressionMatch = PyQt6.QtCore.QRegularExpressionMatch QRegularExpressionMatchIterator = PyQt6.QtCore.QRegularExpressionMatchIterator QResource = PyQt6.QtCore.QResource QRunnable = PyQt6.QtCore.QRunnable QSaveFile = PyQt6.QtCore.QSaveFile QSemaphore = PyQt6.QtCore.QSemaphore QSemaphoreReleaser = PyQt6.QtCore.QSemaphoreReleaser QSequentialAnimationGroup = PyQt6.QtCore.QSequentialAnimationGroup QSettings = PyQt6.QtCore.QSettings QSharedMemory = PyQt6.QtCore.QSharedMemory QSignalBlocker = PyQt6.QtCore.QSignalBlocker QSignalMapper = PyQt6.QtCore.QSignalMapper QSize = PyQt6.QtCore.QSize QSizeF = PyQt6.QtCore.QSizeF QSocketNotifier = PyQt6.QtCore.QSocketNotifier QSortFilterProxyModel = PyQt6.QtCore.QSortFilterProxyModel QStandardPaths = PyQt6.QtCore.QStandardPaths QStorageInfo = PyQt6.QtCore.QStorageInfo QStringConverter = PyQt6.QtCore.QStringConverter QStringConverterBase = PyQt6.QtCore.QStringConverterBase QStringDecoder = PyQt6.QtCore.QStringDecoder QStringEncoder = PyQt6.QtCore.QStringEncoder QStringListModel = PyQt6.QtCore.QStringListModel QSysInfo = PyQt6.QtCore.QSysInfo QSystemSemaphore = PyQt6.QtCore.QSystemSemaphore QT_TRANSLATE_NOOP = PyQt6.QtCore.QT_TRANSLATE_NOOP QT_TR_NOOP = PyQt6.QtCore.QT_TR_NOOP QT_VERSION = PyQt6.QtCore.QT_VERSION QT_VERSION_STR = PyQt6.QtCore.QT_VERSION_STR QTemporaryDir = PyQt6.QtCore.QTemporaryDir QTemporaryFile = PyQt6.QtCore.QTemporaryFile QTextBoundaryFinder = PyQt6.QtCore.QTextBoundaryFinder QTextStream = PyQt6.QtCore.QTextStream QTextStreamManipulator = PyQt6.QtCore.QTextStreamManipulator QThread = PyQt6.QtCore.QThread QThreadPool = PyQt6.QtCore.QThreadPool QTime = PyQt6.QtCore.QTime QTimeLine = PyQt6.QtCore.QTimeLine QTimeZone = PyQt6.QtCore.QTimeZone QTimer = PyQt6.QtCore.QTimer QTimerEvent = PyQt6.QtCore.QTimerEvent QTranslator = PyQt6.QtCore.QTranslator QTransposeProxyModel = PyQt6.QtCore.QTransposeProxyModel QTypeRevision = PyQt6.QtCore.QTypeRevision QUrl = PyQt6.QtCore.QUrl QUrlQuery = PyQt6.QtCore.QUrlQuery QUuid = PyQt6.QtCore.QUuid QVariant = PyQt6.QtCore.QVariant QVariantAnimation = PyQt6.QtCore.QVariantAnimation QVersionNumber = PyQt6.QtCore.QVersionNumber QWaitCondition = PyQt6.QtCore.QWaitCondition QWriteLocker = PyQt6.QtCore.QWriteLocker QXmlStreamAttribute = PyQt6.QtCore.QXmlStreamAttribute QXmlStreamAttributes = PyQt6.QtCore.QXmlStreamAttributes QXmlStreamEntityDeclaration = PyQt6.QtCore.QXmlStreamEntityDeclaration QXmlStreamEntityResolver = PyQt6.QtCore.QXmlStreamEntityResolver QXmlStreamNamespaceDeclaration = PyQt6.QtCore.QXmlStreamNamespaceDeclaration QXmlStreamNotationDeclaration = PyQt6.QtCore.QXmlStreamNotationDeclaration QXmlStreamReader = PyQt6.QtCore.QXmlStreamReader QXmlStreamWriter = PyQt6.QtCore.QXmlStreamWriter Q_ARG = PyQt6.QtCore.Q_ARG Q_RETURN_ARG = PyQt6.QtCore.Q_RETURN_ARG Qt = PyQt6.QtCore.Qt QtMsgType = PyQt6.QtCore.QtMsgType pyqtBoundSignal = PyQt6.QtCore.pyqtBoundSignal pyqtClassInfo = PyQt6.QtCore.pyqtClassInfo pyqtEnum = PyQt6.QtCore.pyqtEnum pyqtPickleProtocol = PyQt6.QtCore.pyqtPickleProtocol pyqtProperty = PyQt6.QtCore.pyqtProperty pyqtRemoveInputHook = PyQt6.QtCore.pyqtRemoveInputHook pyqtRestoreInputHook = PyQt6.QtCore.pyqtRestoreInputHook pyqtSetPickleProtocol = PyQt6.QtCore.pyqtSetPickleProtocol pyqtSignal = PyQt6.QtCore.pyqtSignal pyqtSlot = PyQt6.QtCore.pyqtSlot qAbs = PyQt6.QtCore.qAbs qAddPostRoutine = PyQt6.QtCore.qAddPostRoutine qAddPreRoutine = PyQt6.QtCore.qAddPreRoutine qChecksum = PyQt6.QtCore.qChecksum qCompress = PyQt6.QtCore.qCompress qCritical = PyQt6.QtCore.qCritical qDebug = PyQt6.QtCore.qDebug qEnvironmentVariable = PyQt6.QtCore.qEnvironmentVariable qEnvironmentVariableIntValue = PyQt6.QtCore.qEnvironmentVariableIntValue qEnvironmentVariableIsEmpty = PyQt6.QtCore.qEnvironmentVariableIsEmpty qEnvironmentVariableIsSet = PyQt6.QtCore.qEnvironmentVariableIsSet qFatal = PyQt6.QtCore.qFatal qFloatDistance = PyQt6.QtCore.qFloatDistance qFormatLogMessage = PyQt6.QtCore.qFormatLogMessage qFuzzyCompare = PyQt6.QtCore.qFuzzyCompare qFuzzyIsNull = PyQt6.QtCore.qFuzzyIsNull qInf = PyQt6.QtCore.qInf qInfo = PyQt6.QtCore.qInfo qInstallMessageHandler = PyQt6.QtCore.qInstallMessageHandler qIsFinite = PyQt6.QtCore.qIsFinite qIsInf = PyQt6.QtCore.qIsInf qIsNaN = PyQt6.QtCore.qIsNaN qQNaN = PyQt6.QtCore.qQNaN qRegisterResourceData = PyQt6.QtCore.qRegisterResourceData qRemovePostRoutine = PyQt6.QtCore.qRemovePostRoutine qRound = PyQt6.QtCore.qRound qRound64 = PyQt6.QtCore.qRound64 qSNaN = PyQt6.QtCore.qSNaN qSetFieldWidth = PyQt6.QtCore.qSetFieldWidth qSetMessagePattern = PyQt6.QtCore.qSetMessagePattern qSetPadChar = PyQt6.QtCore.qSetPadChar qSetRealNumberPrecision = PyQt6.QtCore.qSetRealNumberPrecision qUncompress = PyQt6.QtCore.qUncompress qUnregisterResourceData = PyQt6.QtCore.qUnregisterResourceData qVersion = PyQt6.QtCore.qVersion qWarning = PyQt6.QtCore.qWarning qYieldCpu = PyQt6.QtCore.qYieldCpu import PyQt6.QtGui QAbstractFileIconProvider = PyQt6.QtGui.QAbstractFileIconProvider QAbstractTextDocumentLayout = PyQt6.QtGui.QAbstractTextDocumentLayout QAction = PyQt6.QtGui.QAction QActionEvent = PyQt6.QtGui.QActionEvent QActionGroup = PyQt6.QtGui.QActionGroup QBackingStore = PyQt6.QtGui.QBackingStore QBitmap = PyQt6.QtGui.QBitmap QBrush = PyQt6.QtGui.QBrush QChildWindowEvent = PyQt6.QtGui.QChildWindowEvent QClipboard = PyQt6.QtGui.QClipboard QCloseEvent = PyQt6.QtGui.QCloseEvent QColor = PyQt6.QtGui.QColor QColorConstants = PyQt6.QtGui.QColorConstants QColorSpace = PyQt6.QtGui.QColorSpace QColorTransform = PyQt6.QtGui.QColorTransform QConicalGradient = PyQt6.QtGui.QConicalGradient QContextMenuEvent = PyQt6.QtGui.QContextMenuEvent QCursor = PyQt6.QtGui.QCursor QDesktopServices = PyQt6.QtGui.QDesktopServices QDoubleValidator = PyQt6.QtGui.QDoubleValidator QDrag = PyQt6.QtGui.QDrag QDragEnterEvent = PyQt6.QtGui.QDragEnterEvent QDragLeaveEvent = PyQt6.QtGui.QDragLeaveEvent QDragMoveEvent = PyQt6.QtGui.QDragMoveEvent QDropEvent = PyQt6.QtGui.QDropEvent QEnterEvent = PyQt6.QtGui.QEnterEvent QEventPoint = PyQt6.QtGui.QEventPoint QExposeEvent = PyQt6.QtGui.QExposeEvent QFileOpenEvent = PyQt6.QtGui.QFileOpenEvent QFileSystemModel = PyQt6.QtGui.QFileSystemModel QFocusEvent = PyQt6.QtGui.QFocusEvent QFont = PyQt6.QtGui.QFont QFontDatabase = PyQt6.QtGui.QFontDatabase QFontInfo = PyQt6.QtGui.QFontInfo QFontMetrics = PyQt6.QtGui.QFontMetrics QFontMetricsF = PyQt6.QtGui.QFontMetricsF QGlyphRun = PyQt6.QtGui.QGlyphRun QGradient = PyQt6.QtGui.QGradient QGuiApplication = PyQt6.QtGui.QGuiApplication QHelpEvent = PyQt6.QtGui.QHelpEvent QHideEvent = PyQt6.QtGui.QHideEvent QHoverEvent = PyQt6.QtGui.QHoverEvent QIcon = PyQt6.QtGui.QIcon QIconDragEvent = PyQt6.QtGui.QIconDragEvent QIconEngine = PyQt6.QtGui.QIconEngine QImage = PyQt6.QtGui.QImage QImageIOHandler = PyQt6.QtGui.QImageIOHandler QImageReader = PyQt6.QtGui.QImageReader QImageWriter = PyQt6.QtGui.QImageWriter QInputDevice = PyQt6.QtGui.QInputDevice QInputEvent = PyQt6.QtGui.QInputEvent QInputMethod = PyQt6.QtGui.QInputMethod QInputMethodEvent = PyQt6.QtGui.QInputMethodEvent QInputMethodQueryEvent = PyQt6.QtGui.QInputMethodQueryEvent QIntValidator = PyQt6.QtGui.QIntValidator QKeyEvent = PyQt6.QtGui.QKeyEvent QKeySequence = PyQt6.QtGui.QKeySequence QLinearGradient = PyQt6.QtGui.QLinearGradient QMatrix2x2 = PyQt6.QtGui.QMatrix2x2 QMatrix2x3 = PyQt6.QtGui.QMatrix2x3 QMatrix2x4 = PyQt6.QtGui.QMatrix2x4 QMatrix3x2 = PyQt6.QtGui.QMatrix3x2 QMatrix3x3 = PyQt6.QtGui.QMatrix3x3 QMatrix3x4 = PyQt6.QtGui.QMatrix3x4 QMatrix4x2 = PyQt6.QtGui.QMatrix4x2 QMatrix4x3 = PyQt6.QtGui.QMatrix4x3 QMatrix4x4 = PyQt6.QtGui.QMatrix4x4 QMouseEvent = PyQt6.QtGui.QMouseEvent QMoveEvent = PyQt6.QtGui.QMoveEvent QMovie = PyQt6.QtGui.QMovie QNativeGestureEvent = PyQt6.QtGui.QNativeGestureEvent QOffscreenSurface = PyQt6.QtGui.QOffscreenSurface QOpenGLContext = PyQt6.QtGui.QOpenGLContext QOpenGLContextGroup = PyQt6.QtGui.QOpenGLContextGroup QPageLayout = PyQt6.QtGui.QPageLayout QPageRanges = PyQt6.QtGui.QPageRanges QPageSize = PyQt6.QtGui.QPageSize QPagedPaintDevice = PyQt6.QtGui.QPagedPaintDevice QPaintDevice = PyQt6.QtGui.QPaintDevice QPaintDeviceWindow = PyQt6.QtGui.QPaintDeviceWindow QPaintEngine = PyQt6.QtGui.QPaintEngine QPaintEngineState = PyQt6.QtGui.QPaintEngineState QPaintEvent = PyQt6.QtGui.QPaintEvent QPainter = PyQt6.QtGui.QPainter QPainterPath = PyQt6.QtGui.QPainterPath QPainterPathStroker = PyQt6.QtGui.QPainterPathStroker QPalette = PyQt6.QtGui.QPalette QPdfWriter = PyQt6.QtGui.QPdfWriter QPen = PyQt6.QtGui.QPen QPicture = PyQt6.QtGui.QPicture QPixelFormat = PyQt6.QtGui.QPixelFormat QPixmap = PyQt6.QtGui.QPixmap QPixmapCache = PyQt6.QtGui.QPixmapCache QPlatformSurfaceEvent = PyQt6.QtGui.QPlatformSurfaceEvent QPointerEvent = PyQt6.QtGui.QPointerEvent QPointingDevice = PyQt6.QtGui.QPointingDevice QPointingDeviceUniqueId = PyQt6.QtGui.QPointingDeviceUniqueId QPolygon = PyQt6.QtGui.QPolygon QPolygonF = PyQt6.QtGui.QPolygonF QQuaternion = PyQt6.QtGui.QQuaternion QRadialGradient = PyQt6.QtGui.QRadialGradient QRasterWindow = PyQt6.QtGui.QRasterWindow QRawFont = PyQt6.QtGui.QRawFont QRegion = PyQt6.QtGui.QRegion QRegularExpressionValidator = PyQt6.QtGui.QRegularExpressionValidator QResizeEvent = PyQt6.QtGui.QResizeEvent QRgba64 = PyQt6.QtGui.QRgba64 QScreen = PyQt6.QtGui.QScreen QScrollEvent = PyQt6.QtGui.QScrollEvent QScrollPrepareEvent = PyQt6.QtGui.QScrollPrepareEvent QSessionManager = PyQt6.QtGui.QSessionManager QShortcut = PyQt6.QtGui.QShortcut QShortcutEvent = PyQt6.QtGui.QShortcutEvent QShowEvent = PyQt6.QtGui.QShowEvent QSinglePointEvent = PyQt6.QtGui.QSinglePointEvent QStandardItem = PyQt6.QtGui.QStandardItem QStandardItemModel = PyQt6.QtGui.QStandardItemModel QStaticText = PyQt6.QtGui.QStaticText QStatusTipEvent = PyQt6.QtGui.QStatusTipEvent QStyleHints = PyQt6.QtGui.QStyleHints QSurface = PyQt6.QtGui.QSurface QSurfaceFormat = PyQt6.QtGui.QSurfaceFormat QSyntaxHighlighter = PyQt6.QtGui.QSyntaxHighlighter QTabletEvent = PyQt6.QtGui.QTabletEvent QTextBlock = PyQt6.QtGui.QTextBlock QTextBlockFormat = PyQt6.QtGui.QTextBlockFormat QTextBlockGroup = PyQt6.QtGui.QTextBlockGroup QTextBlockUserData = PyQt6.QtGui.QTextBlockUserData QTextCharFormat = PyQt6.QtGui.QTextCharFormat QTextCursor = PyQt6.QtGui.QTextCursor QTextDocument = PyQt6.QtGui.QTextDocument QTextDocumentFragment = PyQt6.QtGui.QTextDocumentFragment QTextDocumentWriter = PyQt6.QtGui.QTextDocumentWriter QTextFormat = PyQt6.QtGui.QTextFormat QTextFragment = PyQt6.QtGui.QTextFragment QTextFrame = PyQt6.QtGui.QTextFrame QTextFrameFormat = PyQt6.QtGui.QTextFrameFormat QTextImageFormat = PyQt6.QtGui.QTextImageFormat QTextInlineObject = PyQt6.QtGui.QTextInlineObject QTextItem = PyQt6.QtGui.QTextItem QTextLayout = PyQt6.QtGui.QTextLayout QTextLength = PyQt6.QtGui.QTextLength QTextLine = PyQt6.QtGui.QTextLine QTextList = PyQt6.QtGui.QTextList QTextListFormat = PyQt6.QtGui.QTextListFormat QTextObject = PyQt6.QtGui.QTextObject QTextObjectInterface = PyQt6.QtGui.QTextObjectInterface QTextOption = PyQt6.QtGui.QTextOption QTextTable = PyQt6.QtGui.QTextTable QTextTableCell = PyQt6.QtGui.QTextTableCell QTextTableCellFormat = PyQt6.QtGui.QTextTableCellFormat QTextTableFormat = PyQt6.QtGui.QTextTableFormat QTouchEvent = PyQt6.QtGui.QTouchEvent QTransform = PyQt6.QtGui.QTransform QUndoCommand = PyQt6.QtGui.QUndoCommand QUndoGroup = PyQt6.QtGui.QUndoGroup QUndoStack = PyQt6.QtGui.QUndoStack QValidator = PyQt6.QtGui.QValidator QVector2D = PyQt6.QtGui.QVector2D QVector3D = PyQt6.QtGui.QVector3D QVector4D = PyQt6.QtGui.QVector4D QWhatsThisClickedEvent = PyQt6.QtGui.QWhatsThisClickedEvent QWheelEvent = PyQt6.QtGui.QWheelEvent QWindow = PyQt6.QtGui.QWindow QWindowStateChangeEvent = PyQt6.QtGui.QWindowStateChangeEvent qAlpha = PyQt6.QtGui.qAlpha qBlue = PyQt6.QtGui.qBlue qGray = PyQt6.QtGui.qGray qGreen = PyQt6.QtGui.qGreen qPixelFormatAlpha = PyQt6.QtGui.qPixelFormatAlpha qPixelFormatCmyk = PyQt6.QtGui.qPixelFormatCmyk qPixelFormatGrayscale = PyQt6.QtGui.qPixelFormatGrayscale qPixelFormatHsl = PyQt6.QtGui.qPixelFormatHsl qPixelFormatHsv = PyQt6.QtGui.qPixelFormatHsv qPixelFormatRgba = PyQt6.QtGui.qPixelFormatRgba qPixelFormatYuv = PyQt6.QtGui.qPixelFormatYuv qPremultiply = PyQt6.QtGui.qPremultiply qRed = PyQt6.QtGui.qRed qRgb = PyQt6.QtGui.qRgb qRgba = PyQt6.QtGui.qRgba qRgba64 = PyQt6.QtGui.qRgba64 qUnpremultiply = PyQt6.QtGui.qUnpremultiply qt_set_sequence_auto_mnemonic = PyQt6.QtGui.qt_set_sequence_auto_mnemonic import PyQt6.QtWidgets QAbstractButton = PyQt6.QtWidgets.QAbstractButton QAbstractGraphicsShapeItem = PyQt6.QtWidgets.QAbstractGraphicsShapeItem QAbstractItemDelegate = PyQt6.QtWidgets.QAbstractItemDelegate QAbstractItemView = PyQt6.QtWidgets.QAbstractItemView QAbstractScrollArea = PyQt6.QtWidgets.QAbstractScrollArea QAbstractSlider = PyQt6.QtWidgets.QAbstractSlider QAbstractSpinBox = PyQt6.QtWidgets.QAbstractSpinBox QApplication = PyQt6.QtWidgets.QApplication QBoxLayout = PyQt6.QtWidgets.QBoxLayout QButtonGroup = PyQt6.QtWidgets.QButtonGroup QCalendarWidget = PyQt6.QtWidgets.QCalendarWidget QCheckBox = PyQt6.QtWidgets.QCheckBox QColorDialog = PyQt6.QtWidgets.QColorDialog QColumnView = PyQt6.QtWidgets.QColumnView QComboBox = PyQt6.QtWidgets.QComboBox QCommandLinkButton = PyQt6.QtWidgets.QCommandLinkButton QCommonStyle = PyQt6.QtWidgets.QCommonStyle QCompleter = PyQt6.QtWidgets.QCompleter QDataWidgetMapper = PyQt6.QtWidgets.QDataWidgetMapper QDateEdit = PyQt6.QtWidgets.QDateEdit QDateTimeEdit = PyQt6.QtWidgets.QDateTimeEdit QDial = PyQt6.QtWidgets.QDial QDialog = PyQt6.QtWidgets.QDialog QDialogButtonBox = PyQt6.QtWidgets.QDialogButtonBox QDockWidget = PyQt6.QtWidgets.QDockWidget QDoubleSpinBox = PyQt6.QtWidgets.QDoubleSpinBox QErrorMessage = PyQt6.QtWidgets.QErrorMessage QFileDialog = PyQt6.QtWidgets.QFileDialog QFileIconProvider = PyQt6.QtWidgets.QFileIconProvider QFocusFrame = PyQt6.QtWidgets.QFocusFrame QFontComboBox = PyQt6.QtWidgets.QFontComboBox QFontDialog = PyQt6.QtWidgets.QFontDialog QFormLayout = PyQt6.QtWidgets.QFormLayout QFrame = PyQt6.QtWidgets.QFrame QGesture = PyQt6.QtWidgets.QGesture QGestureEvent = PyQt6.QtWidgets.QGestureEvent QGestureRecognizer = PyQt6.QtWidgets.QGestureRecognizer QGraphicsAnchor = PyQt6.QtWidgets.QGraphicsAnchor QGraphicsAnchorLayout = PyQt6.QtWidgets.QGraphicsAnchorLayout QGraphicsBlurEffect = PyQt6.QtWidgets.QGraphicsBlurEffect QGraphicsColorizeEffect = PyQt6.QtWidgets.QGraphicsColorizeEffect QGraphicsDropShadowEffect = PyQt6.QtWidgets.QGraphicsDropShadowEffect QGraphicsEffect = PyQt6.QtWidgets.QGraphicsEffect QGraphicsEllipseItem = PyQt6.QtWidgets.QGraphicsEllipseItem QGraphicsGridLayout = PyQt6.QtWidgets.QGraphicsGridLayout QGraphicsItem = PyQt6.QtWidgets.QGraphicsItem QGraphicsItemGroup = PyQt6.QtWidgets.QGraphicsItemGroup QGraphicsLayout = PyQt6.QtWidgets.QGraphicsLayout QGraphicsLayoutItem = PyQt6.QtWidgets.QGraphicsLayoutItem QGraphicsLineItem = PyQt6.QtWidgets.QGraphicsLineItem QGraphicsLinearLayout = PyQt6.QtWidgets.QGraphicsLinearLayout QGraphicsObject = PyQt6.QtWidgets.QGraphicsObject QGraphicsOpacityEffect = PyQt6.QtWidgets.QGraphicsOpacityEffect QGraphicsPathItem = PyQt6.QtWidgets.QGraphicsPathItem QGraphicsPixmapItem = PyQt6.QtWidgets.QGraphicsPixmapItem QGraphicsPolygonItem = PyQt6.QtWidgets.QGraphicsPolygonItem QGraphicsProxyWidget = PyQt6.QtWidgets.QGraphicsProxyWidget QGraphicsRectItem = PyQt6.QtWidgets.QGraphicsRectItem QGraphicsRotation = PyQt6.QtWidgets.QGraphicsRotation QGraphicsScale = PyQt6.QtWidgets.QGraphicsScale QGraphicsScene = PyQt6.QtWidgets.QGraphicsScene QGraphicsSceneContextMenuEvent = PyQt6.QtWidgets.QGraphicsSceneContextMenuEvent QGraphicsSceneDragDropEvent = PyQt6.QtWidgets.QGraphicsSceneDragDropEvent QGraphicsSceneEvent = PyQt6.QtWidgets.QGraphicsSceneEvent QGraphicsSceneHelpEvent = PyQt6.QtWidgets.QGraphicsSceneHelpEvent QGraphicsSceneHoverEvent = PyQt6.QtWidgets.QGraphicsSceneHoverEvent QGraphicsSceneMouseEvent = PyQt6.QtWidgets.QGraphicsSceneMouseEvent QGraphicsSceneMoveEvent = PyQt6.QtWidgets.QGraphicsSceneMoveEvent QGraphicsSceneResizeEvent = PyQt6.QtWidgets.QGraphicsSceneResizeEvent QGraphicsSceneWheelEvent = PyQt6.QtWidgets.QGraphicsSceneWheelEvent QGraphicsSimpleTextItem = PyQt6.QtWidgets.QGraphicsSimpleTextItem QGraphicsTextItem = PyQt6.QtWidgets.QGraphicsTextItem QGraphicsTransform = PyQt6.QtWidgets.QGraphicsTransform QGraphicsView = PyQt6.QtWidgets.QGraphicsView QGraphicsWidget = PyQt6.QtWidgets.QGraphicsWidget QGridLayout = PyQt6.QtWidgets.QGridLayout QGroupBox = PyQt6.QtWidgets.QGroupBox QHBoxLayout = PyQt6.QtWidgets.QHBoxLayout QHeaderView = PyQt6.QtWidgets.QHeaderView QInputDialog = PyQt6.QtWidgets.QInputDialog QItemDelegate = PyQt6.QtWidgets.QItemDelegate QItemEditorCreatorBase = PyQt6.QtWidgets.QItemEditorCreatorBase QItemEditorFactory = PyQt6.QtWidgets.QItemEditorFactory QKeySequenceEdit = PyQt6.QtWidgets.QKeySequenceEdit QLCDNumber = PyQt6.QtWidgets.QLCDNumber QLabel = PyQt6.QtWidgets.QLabel QLayout = PyQt6.QtWidgets.QLayout QLayoutItem = PyQt6.QtWidgets.QLayoutItem QLineEdit = PyQt6.QtWidgets.QLineEdit QListView = PyQt6.QtWidgets.QListView QListWidget = PyQt6.QtWidgets.QListWidget QListWidgetItem = PyQt6.QtWidgets.QListWidgetItem QMainWindow = PyQt6.QtWidgets.QMainWindow QMdiArea = PyQt6.QtWidgets.QMdiArea QMdiSubWindow = PyQt6.QtWidgets.QMdiSubWindow QMenu = PyQt6.QtWidgets.QMenu QMenuBar = PyQt6.QtWidgets.QMenuBar QMessageBox = PyQt6.QtWidgets.QMessageBox QPanGesture = PyQt6.QtWidgets.QPanGesture QPinchGesture = PyQt6.QtWidgets.QPinchGesture QPlainTextDocumentLayout = PyQt6.QtWidgets.QPlainTextDocumentLayout QPlainTextEdit = PyQt6.QtWidgets.QPlainTextEdit QProgressBar = PyQt6.QtWidgets.QProgressBar QProgressDialog = PyQt6.QtWidgets.QProgressDialog QProxyStyle = PyQt6.QtWidgets.QProxyStyle QPushButton = PyQt6.QtWidgets.QPushButton QRadioButton = PyQt6.QtWidgets.QRadioButton QRubberBand = PyQt6.QtWidgets.QRubberBand QScrollArea = PyQt6.QtWidgets.QScrollArea QScrollBar = PyQt6.QtWidgets.QScrollBar QScroller = PyQt6.QtWidgets.QScroller QScrollerProperties = PyQt6.QtWidgets.QScrollerProperties QSizeGrip = PyQt6.QtWidgets.QSizeGrip QSizePolicy = PyQt6.QtWidgets.QSizePolicy QSlider = PyQt6.QtWidgets.QSlider QSpacerItem = PyQt6.QtWidgets.QSpacerItem QSpinBox = PyQt6.QtWidgets.QSpinBox QSplashScreen = PyQt6.QtWidgets.QSplashScreen QSplitter = PyQt6.QtWidgets.QSplitter QSplitterHandle = PyQt6.QtWidgets.QSplitterHandle QStackedLayout = PyQt6.QtWidgets.QStackedLayout QStackedWidget = PyQt6.QtWidgets.QStackedWidget QStatusBar = PyQt6.QtWidgets.QStatusBar QStyle = PyQt6.QtWidgets.QStyle QStyleFactory = PyQt6.QtWidgets.QStyleFactory QStyleHintReturn = PyQt6.QtWidgets.QStyleHintReturn QStyleHintReturnMask = PyQt6.QtWidgets.QStyleHintReturnMask QStyleHintReturnVariant = PyQt6.QtWidgets.QStyleHintReturnVariant QStyleOption = PyQt6.QtWidgets.QStyleOption QStyleOptionButton = PyQt6.QtWidgets.QStyleOptionButton QStyleOptionComboBox = PyQt6.QtWidgets.QStyleOptionComboBox QStyleOptionComplex = PyQt6.QtWidgets.QStyleOptionComplex QStyleOptionDockWidget = PyQt6.QtWidgets.QStyleOptionDockWidget QStyleOptionFocusRect = PyQt6.QtWidgets.QStyleOptionFocusRect QStyleOptionFrame = PyQt6.QtWidgets.QStyleOptionFrame QStyleOptionGraphicsItem = PyQt6.QtWidgets.QStyleOptionGraphicsItem QStyleOptionGroupBox = PyQt6.QtWidgets.QStyleOptionGroupBox QStyleOptionHeader = PyQt6.QtWidgets.QStyleOptionHeader QStyleOptionHeaderV2 = PyQt6.QtWidgets.QStyleOptionHeaderV2 QStyleOptionMenuItem = PyQt6.QtWidgets.QStyleOptionMenuItem QStyleOptionProgressBar = PyQt6.QtWidgets.QStyleOptionProgressBar QStyleOptionRubberBand = PyQt6.QtWidgets.QStyleOptionRubberBand QStyleOptionSizeGrip = PyQt6.QtWidgets.QStyleOptionSizeGrip QStyleOptionSlider = PyQt6.QtWidgets.QStyleOptionSlider QStyleOptionSpinBox = PyQt6.QtWidgets.QStyleOptionSpinBox QStyleOptionTab = PyQt6.QtWidgets.QStyleOptionTab QStyleOptionTabBarBase = PyQt6.QtWidgets.QStyleOptionTabBarBase QStyleOptionTabWidgetFrame = PyQt6.QtWidgets.QStyleOptionTabWidgetFrame QStyleOptionTitleBar = PyQt6.QtWidgets.QStyleOptionTitleBar QStyleOptionToolBar = PyQt6.QtWidgets.QStyleOptionToolBar QStyleOptionToolBox = PyQt6.QtWidgets.QStyleOptionToolBox QStyleOptionToolButton = PyQt6.QtWidgets.QStyleOptionToolButton QStyleOptionViewItem = PyQt6.QtWidgets.QStyleOptionViewItem QStylePainter = PyQt6.QtWidgets.QStylePainter QStyledItemDelegate = PyQt6.QtWidgets.QStyledItemDelegate QSwipeGesture = PyQt6.QtWidgets.QSwipeGesture QSystemTrayIcon = PyQt6.QtWidgets.QSystemTrayIcon QTabBar = PyQt6.QtWidgets.QTabBar QTabWidget = PyQt6.QtWidgets.QTabWidget QTableView = PyQt6.QtWidgets.QTableView QTableWidget = PyQt6.QtWidgets.QTableWidget QTableWidgetItem = PyQt6.QtWidgets.QTableWidgetItem QTableWidgetSelectionRange = PyQt6.QtWidgets.QTableWidgetSelectionRange QTapAndHoldGesture = PyQt6.QtWidgets.QTapAndHoldGesture QTapGesture = PyQt6.QtWidgets.QTapGesture QTextBrowser = PyQt6.QtWidgets.QTextBrowser QTextEdit = PyQt6.QtWidgets.QTextEdit QTimeEdit = PyQt6.QtWidgets.QTimeEdit QToolBar = PyQt6.QtWidgets.QToolBar QToolBox = PyQt6.QtWidgets.QToolBox QToolButton = PyQt6.QtWidgets.QToolButton QToolTip = PyQt6.QtWidgets.QToolTip QTreeView = PyQt6.QtWidgets.QTreeView QTreeWidget = PyQt6.QtWidgets.QTreeWidget QTreeWidgetItem = PyQt6.QtWidgets.QTreeWidgetItem QTreeWidgetItemIterator = PyQt6.QtWidgets.QTreeWidgetItemIterator QUndoView = PyQt6.QtWidgets.QUndoView QVBoxLayout = PyQt6.QtWidgets.QVBoxLayout QWIDGETSIZE_MAX = PyQt6.QtWidgets.QWIDGETSIZE_MAX QWhatsThis = PyQt6.QtWidgets.QWhatsThis QWidget = PyQt6.QtWidgets.QWidget QWidgetAction = PyQt6.QtWidgets.QWidgetAction QWidgetItem = PyQt6.QtWidgets.QWidgetItem QWizard = PyQt6.QtWidgets.QWizard QWizardPage = PyQt6.QtWidgets.QWizardPage qDrawBorderPixmap = PyQt6.QtWidgets.qDrawBorderPixmap qDrawPlainRect = PyQt6.QtWidgets.qDrawPlainRect qDrawPlainRoundedRect = PyQt6.QtWidgets.qDrawPlainRoundedRect qDrawShadeLine = PyQt6.QtWidgets.qDrawShadeLine qDrawShadePanel = PyQt6.QtWidgets.qDrawShadePanel qDrawShadeRect = PyQt6.QtWidgets.qDrawShadeRect qDrawWinButton = PyQt6.QtWidgets.qDrawWinButton qDrawWinPanel = PyQt6.QtWidgets.qDrawWinPanel import PyQt6.QtNetwork QAbstractNetworkCache = PyQt6.QtNetwork.QAbstractNetworkCache QAbstractSocket = PyQt6.QtNetwork.QAbstractSocket QAuthenticator = PyQt6.QtNetwork.QAuthenticator QDnsDomainNameRecord = PyQt6.QtNetwork.QDnsDomainNameRecord QDnsHostAddressRecord = PyQt6.QtNetwork.QDnsHostAddressRecord QDnsLookup = PyQt6.QtNetwork.QDnsLookup QDnsMailExchangeRecord = PyQt6.QtNetwork.QDnsMailExchangeRecord QDnsServiceRecord = PyQt6.QtNetwork.QDnsServiceRecord QDnsTextRecord = PyQt6.QtNetwork.QDnsTextRecord QHostAddress = PyQt6.QtNetwork.QHostAddress QHostInfo = PyQt6.QtNetwork.QHostInfo QHstsPolicy = PyQt6.QtNetwork.QHstsPolicy QHttp1Configuration = PyQt6.QtNetwork.QHttp1Configuration QHttp2Configuration = PyQt6.QtNetwork.QHttp2Configuration QHttpHeaders = PyQt6.QtNetwork.QHttpHeaders QHttpMultiPart = PyQt6.QtNetwork.QHttpMultiPart QHttpPart = PyQt6.QtNetwork.QHttpPart QLocalServer = PyQt6.QtNetwork.QLocalServer QLocalSocket = PyQt6.QtNetwork.QLocalSocket QNetworkAccessManager = PyQt6.QtNetwork.QNetworkAccessManager QNetworkAddressEntry = PyQt6.QtNetwork.QNetworkAddressEntry QNetworkCacheMetaData = PyQt6.QtNetwork.QNetworkCacheMetaData QNetworkCookie = PyQt6.QtNetwork.QNetworkCookie QNetworkCookieJar = PyQt6.QtNetwork.QNetworkCookieJar QNetworkDatagram = PyQt6.QtNetwork.QNetworkDatagram QNetworkDiskCache = PyQt6.QtNetwork.QNetworkDiskCache QNetworkInformation = PyQt6.QtNetwork.QNetworkInformation QNetworkInterface = PyQt6.QtNetwork.QNetworkInterface QNetworkProxy = PyQt6.QtNetwork.QNetworkProxy QNetworkProxyFactory = PyQt6.QtNetwork.QNetworkProxyFactory QNetworkProxyQuery = PyQt6.QtNetwork.QNetworkProxyQuery QNetworkReply = PyQt6.QtNetwork.QNetworkReply QNetworkRequest = PyQt6.QtNetwork.QNetworkRequest QOcspCertificateStatus = PyQt6.QtNetwork.QOcspCertificateStatus QOcspResponse = PyQt6.QtNetwork.QOcspResponse QOcspRevocationReason = PyQt6.QtNetwork.QOcspRevocationReason QPasswordDigestor = PyQt6.QtNetwork.QPasswordDigestor QSsl = PyQt6.QtNetwork.QSsl QSslCertificate = PyQt6.QtNetwork.QSslCertificate QSslCertificateExtension = PyQt6.QtNetwork.QSslCertificateExtension QSslCipher = PyQt6.QtNetwork.QSslCipher QSslConfiguration = PyQt6.QtNetwork.QSslConfiguration QSslDiffieHellmanParameters = PyQt6.QtNetwork.QSslDiffieHellmanParameters QSslEllipticCurve = PyQt6.QtNetwork.QSslEllipticCurve QSslError = PyQt6.QtNetwork.QSslError QSslKey = PyQt6.QtNetwork.QSslKey QSslPreSharedKeyAuthenticator = PyQt6.QtNetwork.QSslPreSharedKeyAuthenticator QSslServer = PyQt6.QtNetwork.QSslServer QSslSocket = PyQt6.QtNetwork.QSslSocket QTcpServer = PyQt6.QtNetwork.QTcpServer QTcpSocket = PyQt6.QtNetwork.QTcpSocket QUdpSocket = PyQt6.QtNetwork.QUdpSocket import PyQt6.QtSvg QSvgGenerator = PyQt6.QtSvg.QSvgGenerator QSvgRenderer = PyQt6.QtSvg.QSvgRenderer import PyQt6.QtPrintSupport QAbstractPrintDialog = PyQt6.QtPrintSupport.QAbstractPrintDialog QPageSetupDialog = PyQt6.QtPrintSupport.QPageSetupDialog QPrintDialog = PyQt6.QtPrintSupport.QPrintDialog QPrintEngine = PyQt6.QtPrintSupport.QPrintEngine QPrintPreviewDialog = PyQt6.QtPrintSupport.QPrintPreviewDialog QPrintPreviewWidget = PyQt6.QtPrintSupport.QPrintPreviewWidget QPrinter = PyQt6.QtPrintSupport.QPrinter QPrinterInfo = PyQt6.QtPrintSupport.QPrinterInfo import PyQt6.QtOpenGL QAbstractOpenGLFunctions = PyQt6.QtOpenGL.QAbstractOpenGLFunctions QOpenGLBuffer = PyQt6.QtOpenGL.QOpenGLBuffer QOpenGLDebugLogger = PyQt6.QtOpenGL.QOpenGLDebugLogger QOpenGLDebugMessage = PyQt6.QtOpenGL.QOpenGLDebugMessage QOpenGLFramebufferObject = PyQt6.QtOpenGL.QOpenGLFramebufferObject QOpenGLFramebufferObjectFormat = PyQt6.QtOpenGL.QOpenGLFramebufferObjectFormat QOpenGLFunctions_2_0 = PyQt6.QtOpenGL.QOpenGLFunctions_2_0 QOpenGLFunctions_2_1 = PyQt6.QtOpenGL.QOpenGLFunctions_2_1 QOpenGLFunctions_4_1_Core = PyQt6.QtOpenGL.QOpenGLFunctions_4_1_Core QOpenGLPaintDevice = PyQt6.QtOpenGL.QOpenGLPaintDevice QOpenGLPixelTransferOptions = PyQt6.QtOpenGL.QOpenGLPixelTransferOptions QOpenGLShader = PyQt6.QtOpenGL.QOpenGLShader QOpenGLShaderProgram = PyQt6.QtOpenGL.QOpenGLShaderProgram QOpenGLTexture = PyQt6.QtOpenGL.QOpenGLTexture QOpenGLTextureBlitter = PyQt6.QtOpenGL.QOpenGLTextureBlitter QOpenGLTimeMonitor = PyQt6.QtOpenGL.QOpenGLTimeMonitor QOpenGLTimerQuery = PyQt6.QtOpenGL.QOpenGLTimerQuery QOpenGLVersionFunctionsFactory = PyQt6.QtOpenGL.QOpenGLVersionFunctionsFactory QOpenGLVersionProfile = PyQt6.QtOpenGL.QOpenGLVersionProfile QOpenGLVertexArrayObject = PyQt6.QtOpenGL.QOpenGLVertexArrayObject QOpenGLWindow = PyQt6.QtOpenGL.QOpenGLWindow import PyQt6.QtOpenGLWidgets QOpenGLWidget = PyQt6.QtOpenGLWidgets.QOpenGLWidget import PyQt6.QtQuick QNativeInterface = PyQt6.QtQuick.QNativeInterface QQuickAsyncImageProvider = PyQt6.QtQuick.QQuickAsyncImageProvider QQuickCloseEvent = PyQt6.QtQuick.QQuickCloseEvent QQuickFramebufferObject = PyQt6.QtQuick.QQuickFramebufferObject QQuickGraphicsConfiguration = PyQt6.QtQuick.QQuickGraphicsConfiguration QQuickGraphicsDevice = PyQt6.QtQuick.QQuickGraphicsDevice QQuickImageProvider = PyQt6.QtQuick.QQuickImageProvider QQuickImageResponse = PyQt6.QtQuick.QQuickImageResponse QQuickItem = PyQt6.QtQuick.QQuickItem QQuickItemGrabResult = PyQt6.QtQuick.QQuickItemGrabResult QQuickPaintedItem = PyQt6.QtQuick.QQuickPaintedItem QQuickRenderControl = PyQt6.QtQuick.QQuickRenderControl QQuickRenderTarget = PyQt6.QtQuick.QQuickRenderTarget QQuickTextDocument = PyQt6.QtQuick.QQuickTextDocument QQuickTextureFactory = PyQt6.QtQuick.QQuickTextureFactory QQuickView = PyQt6.QtQuick.QQuickView QQuickWindow = PyQt6.QtQuick.QQuickWindow QSGBasicGeometryNode = PyQt6.QtQuick.QSGBasicGeometryNode QSGClipNode = PyQt6.QtQuick.QSGClipNode QSGDynamicTexture = PyQt6.QtQuick.QSGDynamicTexture QSGFlatColorMaterial = PyQt6.QtQuick.QSGFlatColorMaterial QSGGeometry = PyQt6.QtQuick.QSGGeometry QSGGeometryNode = PyQt6.QtQuick.QSGGeometryNode QSGImageNode = PyQt6.QtQuick.QSGImageNode QSGMaterial = PyQt6.QtQuick.QSGMaterial QSGMaterialShader = PyQt6.QtQuick.QSGMaterialShader QSGMaterialType = PyQt6.QtQuick.QSGMaterialType QSGNode = PyQt6.QtQuick.QSGNode QSGOpacityNode = PyQt6.QtQuick.QSGOpacityNode QSGOpaqueTextureMaterial = PyQt6.QtQuick.QSGOpaqueTextureMaterial QSGRectangleNode = PyQt6.QtQuick.QSGRectangleNode QSGRenderNode = PyQt6.QtQuick.QSGRenderNode QSGRendererInterface = PyQt6.QtQuick.QSGRendererInterface QSGSimpleRectNode = PyQt6.QtQuick.QSGSimpleRectNode QSGSimpleTextureNode = PyQt6.QtQuick.QSGSimpleTextureNode QSGTextNode = PyQt6.QtQuick.QSGTextNode QSGTexture = PyQt6.QtQuick.QSGTexture QSGTextureMaterial = PyQt6.QtQuick.QSGTextureMaterial QSGTextureProvider = PyQt6.QtQuick.QSGTextureProvider QSGTransformNode = PyQt6.QtQuick.QSGTransformNode QSGVertexColorMaterial = PyQt6.QtQuick.QSGVertexColorMaterial import PyQt6.QtMultimedia QAudio = PyQt6.QtMultimedia.QAudio QAudioBuffer = PyQt6.QtMultimedia.QAudioBuffer QAudioDecoder = PyQt6.QtMultimedia.QAudioDecoder QAudioDevice = PyQt6.QtMultimedia.QAudioDevice QAudioFormat = PyQt6.QtMultimedia.QAudioFormat QAudioInput = PyQt6.QtMultimedia.QAudioInput QAudioOutput = PyQt6.QtMultimedia.QAudioOutput QAudioSink = PyQt6.QtMultimedia.QAudioSink QAudioSource = PyQt6.QtMultimedia.QAudioSource QCamera = PyQt6.QtMultimedia.QCamera QCameraDevice = PyQt6.QtMultimedia.QCameraDevice QCameraFormat = PyQt6.QtMultimedia.QCameraFormat QCapturableWindow = PyQt6.QtMultimedia.QCapturableWindow QImageCapture = PyQt6.QtMultimedia.QImageCapture QMediaCaptureSession = PyQt6.QtMultimedia.QMediaCaptureSession QMediaDevices = PyQt6.QtMultimedia.QMediaDevices QMediaFormat = PyQt6.QtMultimedia.QMediaFormat QMediaMetaData = PyQt6.QtMultimedia.QMediaMetaData QMediaPlayer = PyQt6.QtMultimedia.QMediaPlayer QMediaRecorder = PyQt6.QtMultimedia.QMediaRecorder QMediaTimeRange = PyQt6.QtMultimedia.QMediaTimeRange QScreenCapture = PyQt6.QtMultimedia.QScreenCapture QSoundEffect = PyQt6.QtMultimedia.QSoundEffect QVideoFrame = PyQt6.QtMultimedia.QVideoFrame QVideoFrameFormat = PyQt6.QtMultimedia.QVideoFrameFormat QVideoSink = PyQt6.QtMultimedia.QVideoSink QWindowCapture = PyQt6.QtMultimedia.QWindowCapture QtVideo = PyQt6.QtMultimedia.QtVideo import PyQt6.QtMultimediaWidgets QGraphicsVideoItem = PyQt6.QtMultimediaWidgets.QGraphicsVideoItem QVideoWidget = PyQt6.QtMultimediaWidgets.QVideoWidget import PyQt6.QtTextToSpeech QTextToSpeech = PyQt6.QtTextToSpeech.QTextToSpeech QVoice = PyQt6.QtTextToSpeech.QVoice
35,459
Python
.py
769
45.111834
79
0.888559
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,834
__init__.py
kovidgoyal_calibre/src/qt/__init__.py
# autogenerated by __main__.py do not edit top_level_module_names=('QtCore', 'QtGui', 'QtWidgets', 'QtNetwork', 'QtSvg', 'QtPrintSupport', 'QtOpenGL', 'QtOpenGLWidgets', 'QtQuick', 'QtMultimedia', 'QtMultimediaWidgets', 'QtTextToSpeech', 'QtWebEngineCore', 'QtWebEngineWidgets', 'QtDBus') def __getattr__(name): if name in top_level_module_names: import importlib return importlib.import_module("PyQt6." + name) raise AttributeError(name)
466
Python
.py
7
61.714286
245
0.714912
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,835
webengine.pyi
kovidgoyal_calibre/src/qt/webengine.pyi
# autogenerated by __main__.py do not edit import PyQt6.QtWebEngineCore PYQT_WEBENGINE_VERSION = PyQt6.QtWebEngineCore.PYQT_WEBENGINE_VERSION PYQT_WEBENGINE_VERSION_STR = PyQt6.QtWebEngineCore.PYQT_WEBENGINE_VERSION_STR QWebEngineCertificateError = PyQt6.QtWebEngineCore.QWebEngineCertificateError QWebEngineClientCertificateSelection = PyQt6.QtWebEngineCore.QWebEngineClientCertificateSelection QWebEngineClientCertificateStore = PyQt6.QtWebEngineCore.QWebEngineClientCertificateStore QWebEngineContextMenuRequest = PyQt6.QtWebEngineCore.QWebEngineContextMenuRequest QWebEngineCookieStore = PyQt6.QtWebEngineCore.QWebEngineCookieStore QWebEngineDesktopMediaRequest = PyQt6.QtWebEngineCore.QWebEngineDesktopMediaRequest QWebEngineDownloadRequest = PyQt6.QtWebEngineCore.QWebEngineDownloadRequest QWebEngineFileSystemAccessRequest = PyQt6.QtWebEngineCore.QWebEngineFileSystemAccessRequest QWebEngineFindTextResult = PyQt6.QtWebEngineCore.QWebEngineFindTextResult QWebEngineFullScreenRequest = PyQt6.QtWebEngineCore.QWebEngineFullScreenRequest QWebEngineGlobalSettings = PyQt6.QtWebEngineCore.QWebEngineGlobalSettings QWebEngineHistory = PyQt6.QtWebEngineCore.QWebEngineHistory QWebEngineHistoryItem = PyQt6.QtWebEngineCore.QWebEngineHistoryItem QWebEngineHistoryModel = PyQt6.QtWebEngineCore.QWebEngineHistoryModel QWebEngineHttpRequest = PyQt6.QtWebEngineCore.QWebEngineHttpRequest QWebEngineLoadingInfo = PyQt6.QtWebEngineCore.QWebEngineLoadingInfo QWebEngineNavigationRequest = PyQt6.QtWebEngineCore.QWebEngineNavigationRequest QWebEngineNewWindowRequest = PyQt6.QtWebEngineCore.QWebEngineNewWindowRequest QWebEngineNotification = PyQt6.QtWebEngineCore.QWebEngineNotification QWebEnginePage = PyQt6.QtWebEngineCore.QWebEnginePage QWebEngineProfile = PyQt6.QtWebEngineCore.QWebEngineProfile QWebEngineQuotaRequest = PyQt6.QtWebEngineCore.QWebEngineQuotaRequest QWebEngineRegisterProtocolHandlerRequest = PyQt6.QtWebEngineCore.QWebEngineRegisterProtocolHandlerRequest QWebEngineScript = PyQt6.QtWebEngineCore.QWebEngineScript QWebEngineScriptCollection = PyQt6.QtWebEngineCore.QWebEngineScriptCollection QWebEngineSettings = PyQt6.QtWebEngineCore.QWebEngineSettings QWebEngineUrlRequestInfo = PyQt6.QtWebEngineCore.QWebEngineUrlRequestInfo QWebEngineUrlRequestInterceptor = PyQt6.QtWebEngineCore.QWebEngineUrlRequestInterceptor QWebEngineUrlRequestJob = PyQt6.QtWebEngineCore.QWebEngineUrlRequestJob QWebEngineUrlScheme = PyQt6.QtWebEngineCore.QWebEngineUrlScheme QWebEngineUrlSchemeHandler = PyQt6.QtWebEngineCore.QWebEngineUrlSchemeHandler QWebEngineWebAuthPinRequest = PyQt6.QtWebEngineCore.QWebEngineWebAuthPinRequest QWebEngineWebAuthUxRequest = PyQt6.QtWebEngineCore.QWebEngineWebAuthUxRequest qWebEngineChromiumSecurityPatchVersion = PyQt6.QtWebEngineCore.qWebEngineChromiumSecurityPatchVersion qWebEngineChromiumVersion = PyQt6.QtWebEngineCore.qWebEngineChromiumVersion qWebEngineGetDomainAndRegistry = PyQt6.QtWebEngineCore.qWebEngineGetDomainAndRegistry qWebEngineProcessName = PyQt6.QtWebEngineCore.qWebEngineProcessName qWebEngineVersion = PyQt6.QtWebEngineCore.qWebEngineVersion import PyQt6.QtWebEngineWidgets QWebEngineView = PyQt6.QtWebEngineWidgets.QWebEngineView
3,180
Python
.py
44
71.295455
105
0.926363
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,836
core.py
kovidgoyal_calibre/src/qt/core.py
# autogenerated by __main__.py do not edit from .core_name_map import module_names, name_map from .loader import dynamic_load already_imported = {} qt_modules = {} def __getattr__(name): return dynamic_load(name, name_map, already_imported, qt_modules, module_names)
273
Python
.py
7
37.142857
83
0.746212
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,837
webengine.py
kovidgoyal_calibre/src/qt/webengine.py
# autogenerated by __main__.py do not edit from .webengine_name_map import module_names, name_map from .loader import dynamic_load already_imported = {} qt_modules = {} def __getattr__(name): return dynamic_load(name, name_map, already_imported, qt_modules, module_names)
278
Python
.py
7
37.857143
83
0.750929
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,838
token_data.py
kovidgoyal_calibre/src/tinycss/token_data.py
# coding: utf8 """ tinycss.token_data ------------------ Shared data for both implementations (Cython and Python) of the tokenizer. :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ import functools import operator import re import string import sys # * Raw strings with the r'' notation are used so that \ do not need # to be escaped. # * Names and regexps are separated by a tabulation. # * Macros are re-ordered so that only previous definitions are needed. # * {} are used for macro substitution with ``string.Formatter``, # so other uses of { or } have been doubled. # * The syntax is otherwise compatible with re.compile. # * Some parentheses were added to add capturing groups. # (in unicode, DIMENSION and URI) # *** Willful violation: *** # Numbers can take a + or - sign, but the sign is a separate DELIM token. # Since comments are allowed anywhere between tokens, this makes # the following this is valid. It means 10 negative pixels: # margin-top: -/**/10px # This makes parsing numbers a pain, so instead we’ll do the same is Firefox # and make the sign part as of the 'num' macro. The above CSS will be invalid. # See discussion: # http://lists.w3.org/Archives/Public/www-style/2011Oct/0028.html MACROS = r''' nl \n|\r\n|\r|\f w [ \t\r\n\f]* nonascii [^\0-\237] unicode \\([0-9a-f]{{1,6}})(\r\n|[ \n\r\t\f])? simple_escape [^\n\r\f0-9a-f] escape {unicode}|\\{simple_escape} nmstart [_a-z]|{nonascii}|{escape} nmchar [_a-z0-9-]|{nonascii}|{escape} name {nmchar}+ ident [-]?{nmstart}{nmchar}* num [-+]?(?:[0-9]*\.[0-9]+|[0-9]+) string1 \"([^\n\r\f\\"]|\\{nl}|{escape})*\" string2 \'([^\n\r\f\\']|\\{nl}|{escape})*\' string {string1}|{string2} badstring1 \"([^\n\r\f\\"]|\\{nl}|{escape})*\\? badstring2 \'([^\n\r\f\\']|\\{nl}|{escape})*\\? badstring {badstring1}|{badstring2} badcomment1 \/\*[^*]*\*+([^/*][^*]*\*+)* badcomment2 \/\*[^*]*(\*+[^/*][^*]*)* badcomment {badcomment1}|{badcomment2} baduri1 url\({w}([!#$%&*-~]|{nonascii}|{escape})*{w} baduri2 url\({w}{string}{w} baduri3 url\({w}{badstring} baduri {baduri1}|{baduri2}|{baduri3} '''.replace(r'\0', '\0').replace(r'\237', '\237') # Removed these tokens. Instead, they’re tokenized as two DELIM each. # INCLUDES ~= # DASHMATCH |= # They are only used in selectors but selectors3 also have ^=, *= and $=. # We don’t actually parse selectors anyway # Re-ordered so that the longest match is always the first. # For example, "url('foo')" matches URI, BAD_URI, FUNCTION and IDENT, # but URI would always be a longer match than the others. TOKENS = r''' S [ \t\r\n\f]+ URI url\({w}({string}|([!#$%&*-\[\]-~]|{nonascii}|{escape})*){w}\) BAD_URI {baduri} FUNCTION {ident}\( UNICODE-RANGE u\+[0-9a-f?]{{1,6}}(-[0-9a-f]{{1,6}})? IDENT {ident} ATKEYWORD @{ident} HASH #{name} DIMENSION ({num})({ident}) PERCENTAGE {num}% NUMBER {num} STRING {string} BAD_STRING {badstring} COMMENT \/\*[^*]*\*+([^/*][^*]*\*+)*\/ BAD_COMMENT {badcomment} : : ; ; { \{{ } \}} ( \( ) \) [ \[ ] \] CDO <!-- CDC --> ''' # Strings with {macro} expanded COMPILED_MACROS = {} COMPILED_TOKEN_REGEXPS = [] # [(name, regexp.match)] ordered COMPILED_TOKEN_INDEXES = {} # {name: i} helper for the C speedups # Indexed by codepoint value of the first character of a token. # Codepoints >= 160 (aka nonascii) all use the index 160. # values are (i, name, regexp.match) TOKEN_DISPATCH = [] try: unichr except NameError: # Python 3 unichr = chr unicode = str def _init(): """Import-time initialization.""" COMPILED_MACROS.clear() for line in MACROS.splitlines(): if line.strip(): name, value = line.split('\t') COMPILED_MACROS[name.strip()] = '(?:%s)' \ % value.format(**COMPILED_MACROS) COMPILED_TOKEN_REGEXPS[:] = ( ( name.strip(), re.compile( value.format(**COMPILED_MACROS), # Case-insensitive when matching eg. uRL(foo) # but preserve the case in extracted groups re.I ).match ) for line in TOKENS.splitlines() if line.strip() for name, value in [line.split('\t')] ) COMPILED_TOKEN_INDEXES.clear() for i, (name, regexp) in enumerate(COMPILED_TOKEN_REGEXPS): COMPILED_TOKEN_INDEXES[name] = i dispatch = [[] for i in range(161)] for chars, names in [ (' \t\r\n\f', ['S']), ('uU', ['URI', 'BAD_URI', 'UNICODE-RANGE']), # \ is an escape outside of another token (string.ascii_letters + '\\_-' + unichr(160), ['FUNCTION', 'IDENT']), (string.digits + '.+-', ['DIMENSION', 'PERCENTAGE', 'NUMBER']), ('@', ['ATKEYWORD']), ('#', ['HASH']), ('\'"', ['STRING', 'BAD_STRING']), ('/', ['COMMENT', 'BAD_COMMENT']), ('<', ['CDO']), ('-', ['CDC']), ]: for char in chars: dispatch[ord(char)].extend(names) for char in ':;{}()[]': dispatch[ord(char)] = [char] TOKEN_DISPATCH[:] = ( [ (index,) + COMPILED_TOKEN_REGEXPS[index] for name in names for index in [COMPILED_TOKEN_INDEXES[name]] ] for names in dispatch ) _init() def _unicode_replace(match, int=int, unichr=unichr, maxunicode=sys.maxunicode): codepoint = int(match.group(1), 16) if codepoint <= maxunicode: return unichr(codepoint) else: return '\N{REPLACEMENT CHARACTER}' # U+FFFD UNICODE_UNESCAPE = functools.partial( re.compile(COMPILED_MACROS['unicode'], re.I).sub, _unicode_replace) NEWLINE_UNESCAPE = functools.partial( re.compile(r'()\\' + COMPILED_MACROS['nl']).sub, '') SIMPLE_UNESCAPE = functools.partial( re.compile(r'\\(%s)' % COMPILED_MACROS['simple_escape'], re.I).sub, # Same as r'\1', but faster on CPython operator.methodcaller('group', 1)) def FIND_NEWLINES(x): return list(re.compile(COMPILED_MACROS['nl']).finditer(x)) class Token: r"""A single atomic token. .. attribute:: is_container Always ``False``. Helps to tell :class:`Token` apart from :class:`ContainerToken`. .. attribute:: type The type of token as a string: ``S`` A sequence of white space ``IDENT`` An identifier: a name that does not start with a digit. A name is a sequence of letters, digits, ``_``, ``-``, escaped characters and non-ASCII characters. Eg: ``margin-left`` ``HASH`` ``#`` followed immediately by a name. Eg: ``#ff8800`` ``ATKEYWORD`` ``@`` followed immediately by an identifier. Eg: ``@page`` ``URI`` Eg: ``url(foo)`` The content may or may not be quoted. ``UNICODE-RANGE`` ``U+`` followed by one or two hexadecimal Unicode codepoints. Eg: ``U+20-00FF`` ``INTEGER`` An integer with an optional ``+`` or ``-`` sign ``NUMBER`` A non-integer number with an optional ``+`` or ``-`` sign ``DIMENSION`` An integer or number followed immediately by an identifier (the unit). Eg: ``12px`` ``PERCENTAGE`` An integer or number followed immediately by ``%`` ``STRING`` A string, quoted with ``"`` or ``'`` ``:`` or ``;`` That character. ``DELIM`` A single character not matched in another token. Eg: ``,`` See the source of the :mod:`.token_data` module for the precise regular expressions that match various tokens. Note that other token types exist in the early tokenization steps, but these are ignored, are syntax errors, or are later transformed into :class:`ContainerToken` or :class:`FunctionToken`. .. attribute:: value The parsed value: * INTEGER, NUMBER, PERCENTAGE or DIMENSION tokens: the numeric value as an int or float. * STRING tokens: the unescaped string without quotes * URI tokens: the unescaped URI without quotes or ``url(`` and ``)`` markers. * IDENT, ATKEYWORD or HASH tokens: the unescaped token, with ``@`` or ``#`` markers left as-is * Other tokens: same as :attr:`as_css` *Unescaped* refers to the various escaping methods based on the backslash ``\`` character in CSS syntax. .. attribute:: unit * DIMENSION tokens: the normalized (unescaped, lower-case) unit name as a string. eg. ``'px'`` * PERCENTAGE tokens: the string ``'%'`` * Other tokens: ``None`` .. attribute:: line The line number in the CSS source of the start of this token. .. attribute:: column The column number (inside a source line) of the start of this token. """ is_container = False __slots__ = 'type', '_as_css', 'value', 'unit', 'line', 'column' def __init__(self, type_, css_value, value, unit, line, column): self.type = type_ self._as_css = css_value self.value = value self.unit = unit self.line = line self.column = column def as_css(self): """ Return as an Unicode string the CSS representation of the token, as parsed in the source. """ return self._as_css def __repr__(self): return ('<Token {0.type} at {0.line}:{0.column} {0.value!r}{1}>' .format(self, self.unit or '')) class ContainerToken: """A token that contains other (nested) tokens. .. attribute:: is_container Always ``True``. Helps to tell :class:`ContainerToken` apart from :class:`Token`. .. attribute:: type The type of token as a string. One of ``{``, ``(``, ``[`` or ``FUNCTION``. For ``FUNCTION``, the object is actually a :class:`FunctionToken`. .. attribute:: unit Always ``None``. Included to make :class:`ContainerToken` behave more like :class:`Token`. .. attribute:: content A list of :class:`Token` or nested :class:`ContainerToken`, not including the opening or closing token. .. attribute:: line The line number in the CSS source of the start of this token. .. attribute:: column The column number (inside a source line) of the start of this token. """ is_container = True unit = None __slots__ = 'type', '_css_start', '_css_end', 'content', 'line', 'column' def __init__(self, type_, css_start, css_end, content, line, column): self.type = type_ self._css_start = css_start self._css_end = css_end self.content = content self.line = line self.column = column def as_css(self): """ Return as an Unicode string the CSS representation of the token, as parsed in the source. """ parts = [self._css_start] parts.extend(token.as_css() for token in self.content) parts.append(self._css_end) return ''.join(parts) format_string = '<ContainerToken {0.type} at {0.line}:{0.column}>' def __repr__(self): return (self.format_string + ' {0.content}').format(self) class FunctionToken(ContainerToken): """A specialized :class:`ContainerToken` for a ``FUNCTION`` group. Has an additional attribute: .. attribute:: function_name The unescaped name of the function, with the ``(`` marker removed. """ __slots__ = 'function_name', def __init__(self, type_, css_start, css_end, function_name, content, line, column): super(FunctionToken, self).__init__( type_, css_start, css_end, content, line, column) # Remove the ( marker: self.function_name = function_name[:-1] format_string = ('<FunctionToken {0.function_name}() at ' '{0.line}:{0.column}>') class TokenList(list): """ A mixed list of :class:`~.token_data.Token` and :class:`~.token_data.ContainerToken` objects. This is a subclass of the builtin :class:`~builtins.list` type. It can be iterated, indexed and sliced as usual, but also has some additional API: """ @property def line(self): """The line number in the CSS source of the first token.""" return self[0].line @property def column(self): """The column number (inside a source line) of the first token.""" return self[0].column def as_css(self): """ Return as an Unicode string the CSS representation of the tokens, as parsed in the source. """ return ''.join(token.as_css() for token in self) def load_c_tokenizer(): from calibre_extensions import tokenizer tokens = list(':;(){}[]') + ['DELIM', 'INTEGER', 'STRING'] tokenizer.init( COMPILED_TOKEN_REGEXPS, UNICODE_UNESCAPE, NEWLINE_UNESCAPE, SIMPLE_UNESCAPE, FIND_NEWLINES, TOKEN_DISPATCH, COMPILED_TOKEN_INDEXES, *tokens) return tokenizer
13,353
Python
.py
345
31.744928
79
0.593357
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,839
parsing.py
kovidgoyal_calibre/src/tinycss/parsing.py
# coding: utf8 """ tinycss.parsing --------------- Utilities for parsing lists of tokens. :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ # TODO: unit tests def split_on_comma(tokens): """Split a list of tokens on commas, ie ``,`` DELIM tokens. Only "top-level" comma tokens are splitting points, not commas inside a function or other :class:`ContainerToken`. :param tokens: An iterable of :class:`~.token_data.Token` or :class:`~.token_data.ContainerToken`. :returns: A list of lists of tokens """ parts = [] this_part = [] for token in tokens: if token.type == 'DELIM' and token.value == ',': parts.append(this_part) this_part = [] else: this_part.append(token) parts.append(this_part) return parts def strip_whitespace(tokens): """Remove whitespace at the beginning and end of a token list. Whitespace tokens in-between other tokens in the list are preserved. :param tokens: A list of :class:`~.token_data.Token` or :class:`~.token_data.ContainerToken`. :return: A new sub-sequence of the list. """ for i, token in enumerate(tokens): if token.type != 'S': break else: return [] # only whitespace tokens = tokens[i:] while tokens and tokens[-1].type == 'S': tokens.pop() return tokens def remove_whitespace(tokens): """Remove any top-level whitespace in a token list. Whitespace tokens inside recursive :class:`~.token_data.ContainerToken` are preserved. :param tokens: A list of :class:`~.token_data.Token` or :class:`~.token_data.ContainerToken`. :return: A new sub-sequence of the list. """ return [token for token in tokens if token.type != 'S'] def validate_value(tokens): """Validate a property value. :param tokens: an iterable of tokens :raises: :class:`ParseError` if there is any invalid token for the 'value' production of the core grammar. """ for token in tokens: type_ = token.type if type_ == '{': validate_block(token.content, 'property value') else: validate_any(token, 'property value') def validate_block(tokens, context): """ :raises: :class:`ParseError` if there is any invalid token for the 'block' production of the core grammar. :param tokens: an iterable of tokens :param context: a string for the 'unexpected in ...' message """ for token in tokens: type_ = token.type if type_ == '{': validate_block(token.content, context) elif type_ not in (';', 'ATKEYWORD'): validate_any(token, context) def validate_any(token, context): """ :raises: :class:`ParseError` if this is an invalid token for the 'any' production of the core grammar. :param token: a single token :param context: a string for the 'unexpected in ...' message """ type_ = token.type if type_ in ('FUNCTION', '(', '['): for token in token.content: validate_any(token, type_) elif type_ not in ('S', 'IDENT', 'DIMENSION', 'PERCENTAGE', 'NUMBER', 'INTEGER', 'URI', 'DELIM', 'STRING', 'HASH', ':', 'UNICODE-RANGE'): if type_ in ('}', ')', ']'): adjective = 'unmatched' else: adjective = 'unexpected' raise ParseError(token, '{0} {1} token in {2}'.format(adjective, type_, context)) class ParseError(ValueError): """Details about a CSS syntax error. Usually indicates that something (a rule or a declaration) was ignored and will not appear as a parsed object. This exception is typically logged in a list rather than being propagated to the user API. .. attribute:: line Source line where the error occurred. .. attribute:: column Column in the source line where the error occurred. .. attribute:: reason What happened (a string). """ def __init__(self, subject, reason): self.line = subject.line self.column = subject.column self.reason = reason super(ParseError, self).__init__( 'Parse error at {0.line}:{0.column}, {0.reason}'.format(self))
4,461
Python
.py
126
28.126984
77
0.609819
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,840
color3.py
kovidgoyal_calibre/src/tinycss/color3.py
# coding: utf8 """ tinycss.colors3 --------------- Parser for CSS 3 color values http://www.w3.org/TR/css3-color/ This module does not provide anything that integrates in a parser class, only functions that parse single tokens from (eg.) a property value. :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ import collections import itertools import re from .tokenizer import tokenize_grouped class RGBA(collections.namedtuple('RGBA', ['red', 'green', 'blue', 'alpha'])): """An RGBA color. A tuple of four floats in the 0..1 range: ``(r, g, b, a)``. Also has ``red``, ``green``, ``blue`` and ``alpha`` attributes to access the same values. """ def parse_color_string(css_string): """Parse a CSS string as a color value. This is a convenience wrapper around :func:`parse_color` in case you have a string that is not from a CSS stylesheet. :param css_string: An unicode string in CSS syntax. :returns: Same as :func:`parse_color`. """ tokens = list(tokenize_grouped(css_string.strip())) if len(tokens) == 1: return parse_color(tokens[0]) def parse_color(token): """Parse single token as a color value. :param token: A single :class:`~.token_data.Token` or :class:`~.token_data.ContainerToken`, as found eg. in a property value. :returns: * ``None``, if the token is not a valid CSS 3 color value. (No exception is raised.) * For the *currentColor* keyword: the string ``'currentColor'`` * Every other values (including keywords, HSL and HSLA) is converted to RGBA and returned as an :class:`RGBA` object (a 4-tuple with attribute access). The alpha channel is clipped to [0, 1], but R, G, or B can be out of range (eg. ``rgb(-51, 306, 0)`` is represented as ``(-.2, 1.2, 0, 1)``.) """ if token.type == 'IDENT': return COLOR_KEYWORDS.get(token.value.lower()) elif token.type == 'HASH': for multiplier, regexp in HASH_REGEXPS: match = regexp(token.value) if match: r, g, b = [int(group * multiplier, 16) / 255 for group in match.groups()] return RGBA(r, g, b, 1.) elif token.type == 'FUNCTION': args = parse_comma_separated(token.content) if args: name = token.function_name.lower() if name == 'rgb': return parse_rgb(args, alpha=1.) elif name == 'rgba': alpha = parse_alpha(args[3:]) if alpha is not None: return parse_rgb(args[:3], alpha) elif name == 'hsl': return parse_hsl(args, alpha=1.) elif name == 'hsla': alpha = parse_alpha(args[3:]) if alpha is not None: return parse_hsl(args[:3], alpha) def parse_alpha(args): """ If args is a list of a single INTEGER or NUMBER token, return its value clipped to the 0..1 range Otherwise, return None. """ if len(args) == 1 and args[0].type in ('NUMBER', 'INTEGER'): return min(1, max(0, args[0].value)) def parse_rgb(args, alpha): """ If args is a list of 3 INTEGER tokens or 3 PERCENTAGE tokens, return RGB values as a tuple of 3 floats in 0..1. Otherwise, return None. """ types = [arg.type for arg in args] if types == ['INTEGER', 'INTEGER', 'INTEGER']: r, g, b = [arg.value / 255 for arg in args[:3]] return RGBA(r, g, b, alpha) elif types == ['PERCENTAGE', 'PERCENTAGE', 'PERCENTAGE']: r, g, b = [arg.value / 100 for arg in args[:3]] return RGBA(r, g, b, alpha) def parse_hsl(args, alpha): """ If args is a list of 1 INTEGER token and 2 PERCENTAGE tokens, return RGB values as a tuple of 3 floats in 0..1. Otherwise, return None. """ types = [arg.type for arg in args] if types == ['INTEGER', 'PERCENTAGE', 'PERCENTAGE']: hsl = [arg.value for arg in args[:3]] r, g, b = hsl_to_rgb(*hsl) return RGBA(r, g, b, alpha) def hsl_to_rgb(hue, saturation, lightness): """ :param hue: degrees :param saturation: percentage :param lightness: percentage :returns: (r, g, b) as floats in the 0..1 range """ hue = (hue / 360) % 1 saturation = min(1, max(0, saturation / 100)) lightness = min(1, max(0, lightness / 100)) # Translated from ABC: http://www.w3.org/TR/css3-color/#hsl-color def hue_to_rgb(m1, m2, h): if h < 0: h += 1 if h > 1: h -= 1 if h * 6 < 1: return m1 + (m2 - m1) * h * 6 if h * 2 < 1: return m2 if h * 3 < 2: return m1 + (m2 - m1) * (2 / 3 - h) * 6 return m1 if lightness <= 0.5: m2 = lightness * (saturation + 1) else: m2 = lightness + saturation - lightness * saturation m1 = lightness * 2 - m2 return ( hue_to_rgb(m1, m2, hue + 1 / 3), hue_to_rgb(m1, m2, hue), hue_to_rgb(m1, m2, hue - 1 / 3), ) def parse_comma_separated(tokens): """Parse a list of tokens (typically the content of a function token) as arguments made of a single token each, separated by mandatory commas, with optional white space around each argument. return the argument list without commas or white space; or None if the function token content do not match the description above. """ tokens = [token for token in tokens if token.type != 'S'] if not tokens: return [] if len(tokens) % 2 == 1 and all( token.type == 'DELIM' and token.value == ',' for token in tokens[1::2]): return tokens[::2] HASH_REGEXPS = ( (2, re.compile(r'^#([\da-f])([\da-f])([\da-f])$', re.I).match), (1, re.compile(r'^#([\da-f]{2})([\da-f]{2})([\da-f]{2})$', re.I).match), ) # (r, g, b) in 0..255 BASIC_COLOR_KEYWORDS = [ ('black', (0, 0, 0)), ('silver', (192, 192, 192)), ('gray', (128, 128, 128)), ('white', (255, 255, 255)), ('maroon', (128, 0, 0)), ('red', (255, 0, 0)), ('purple', (128, 0, 128)), ('fuchsia', (255, 0, 255)), ('green', (0, 128, 0)), ('lime', (0, 255, 0)), ('olive', (128, 128, 0)), ('yellow', (255, 255, 0)), ('navy', (0, 0, 128)), ('blue', (0, 0, 255)), ('teal', (0, 128, 128)), ('aqua', (0, 255, 255)), ] # (r, g, b) in 0..255 EXTENDED_COLOR_KEYWORDS = [ ('aliceblue', (240, 248, 255)), ('antiquewhite', (250, 235, 215)), ('aqua', (0, 255, 255)), ('aquamarine', (127, 255, 212)), ('azure', (240, 255, 255)), ('beige', (245, 245, 220)), ('bisque', (255, 228, 196)), ('black', (0, 0, 0)), ('blanchedalmond', (255, 235, 205)), ('blue', (0, 0, 255)), ('blueviolet', (138, 43, 226)), ('brown', (165, 42, 42)), ('burlywood', (222, 184, 135)), ('cadetblue', (95, 158, 160)), ('chartreuse', (127, 255, 0)), ('chocolate', (210, 105, 30)), ('coral', (255, 127, 80)), ('cornflowerblue', (100, 149, 237)), ('cornsilk', (255, 248, 220)), ('crimson', (220, 20, 60)), ('cyan', (0, 255, 255)), ('darkblue', (0, 0, 139)), ('darkcyan', (0, 139, 139)), ('darkgoldenrod', (184, 134, 11)), ('darkgray', (169, 169, 169)), ('darkgreen', (0, 100, 0)), ('darkgrey', (169, 169, 169)), ('darkkhaki', (189, 183, 107)), ('darkmagenta', (139, 0, 139)), ('darkolivegreen', (85, 107, 47)), ('darkorange', (255, 140, 0)), ('darkorchid', (153, 50, 204)), ('darkred', (139, 0, 0)), ('darksalmon', (233, 150, 122)), ('darkseagreen', (143, 188, 143)), ('darkslateblue', (72, 61, 139)), ('darkslategray', (47, 79, 79)), ('darkslategrey', (47, 79, 79)), ('darkturquoise', (0, 206, 209)), ('darkviolet', (148, 0, 211)), ('deeppink', (255, 20, 147)), ('deepskyblue', (0, 191, 255)), ('dimgray', (105, 105, 105)), ('dimgrey', (105, 105, 105)), ('dodgerblue', (30, 144, 255)), ('firebrick', (178, 34, 34)), ('floralwhite', (255, 250, 240)), ('forestgreen', (34, 139, 34)), ('fuchsia', (255, 0, 255)), ('gainsboro', (220, 220, 220)), ('ghostwhite', (248, 248, 255)), ('gold', (255, 215, 0)), ('goldenrod', (218, 165, 32)), ('gray', (128, 128, 128)), ('green', (0, 128, 0)), ('greenyellow', (173, 255, 47)), ('grey', (128, 128, 128)), ('honeydew', (240, 255, 240)), ('hotpink', (255, 105, 180)), ('indianred', (205, 92, 92)), ('indigo', (75, 0, 130)), ('ivory', (255, 255, 240)), ('khaki', (240, 230, 140)), ('lavender', (230, 230, 250)), ('lavenderblush', (255, 240, 245)), ('lawngreen', (124, 252, 0)), ('lemonchiffon', (255, 250, 205)), ('lightblue', (173, 216, 230)), ('lightcoral', (240, 128, 128)), ('lightcyan', (224, 255, 255)), ('lightgoldenrodyellow', (250, 250, 210)), ('lightgray', (211, 211, 211)), ('lightgreen', (144, 238, 144)), ('lightgrey', (211, 211, 211)), ('lightpink', (255, 182, 193)), ('lightsalmon', (255, 160, 122)), ('lightseagreen', (32, 178, 170)), ('lightskyblue', (135, 206, 250)), ('lightslategray', (119, 136, 153)), ('lightslategrey', (119, 136, 153)), ('lightsteelblue', (176, 196, 222)), ('lightyellow', (255, 255, 224)), ('lime', (0, 255, 0)), ('limegreen', (50, 205, 50)), ('linen', (250, 240, 230)), ('magenta', (255, 0, 255)), ('maroon', (128, 0, 0)), ('mediumaquamarine', (102, 205, 170)), ('mediumblue', (0, 0, 205)), ('mediumorchid', (186, 85, 211)), ('mediumpurple', (147, 112, 219)), ('mediumseagreen', (60, 179, 113)), ('mediumslateblue', (123, 104, 238)), ('mediumspringgreen', (0, 250, 154)), ('mediumturquoise', (72, 209, 204)), ('mediumvioletred', (199, 21, 133)), ('midnightblue', (25, 25, 112)), ('mintcream', (245, 255, 250)), ('mistyrose', (255, 228, 225)), ('moccasin', (255, 228, 181)), ('navajowhite', (255, 222, 173)), ('navy', (0, 0, 128)), ('oldlace', (253, 245, 230)), ('olive', (128, 128, 0)), ('olivedrab', (107, 142, 35)), ('orange', (255, 165, 0)), ('orangered', (255, 69, 0)), ('orchid', (218, 112, 214)), ('palegoldenrod', (238, 232, 170)), ('palegreen', (152, 251, 152)), ('paleturquoise', (175, 238, 238)), ('palevioletred', (219, 112, 147)), ('papayawhip', (255, 239, 213)), ('peachpuff', (255, 218, 185)), ('peru', (205, 133, 63)), ('pink', (255, 192, 203)), ('plum', (221, 160, 221)), ('powderblue', (176, 224, 230)), ('purple', (128, 0, 128)), ('red', (255, 0, 0)), ('rosybrown', (188, 143, 143)), ('royalblue', (65, 105, 225)), ('saddlebrown', (139, 69, 19)), ('salmon', (250, 128, 114)), ('sandybrown', (244, 164, 96)), ('seagreen', (46, 139, 87)), ('seashell', (255, 245, 238)), ('sienna', (160, 82, 45)), ('silver', (192, 192, 192)), ('skyblue', (135, 206, 235)), ('slateblue', (106, 90, 205)), ('slategray', (112, 128, 144)), ('slategrey', (112, 128, 144)), ('snow', (255, 250, 250)), ('springgreen', (0, 255, 127)), ('steelblue', (70, 130, 180)), ('tan', (210, 180, 140)), ('teal', (0, 128, 128)), ('thistle', (216, 191, 216)), ('tomato', (255, 99, 71)), ('turquoise', (64, 224, 208)), ('violet', (238, 130, 238)), ('wheat', (245, 222, 179)), ('white', (255, 255, 255)), ('whitesmoke', (245, 245, 245)), ('yellow', (255, 255, 0)), ('yellowgreen', (154, 205, 50)), ] # (r, g, b, a) in 0..1 or a string marker SPECIAL_COLOR_KEYWORDS = { 'currentcolor': 'currentColor', 'transparent': RGBA(0., 0., 0., 0.), } # RGBA namedtuples of (r, g, b, a) in 0..1 or a string marker COLOR_KEYWORDS = SPECIAL_COLOR_KEYWORDS.copy() COLOR_KEYWORDS.update( # 255 maps to 1, 0 to 0, the rest is linear. (keyword, RGBA(r / 255., g / 255., b / 255., 1.)) for keyword, (r, g, b) in itertools.chain( BASIC_COLOR_KEYWORDS, EXTENDED_COLOR_KEYWORDS))
12,242
Python
.py
339
30.032448
78
0.540388
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,841
media3.py
kovidgoyal_calibre/src/tinycss/media3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from polyglot.builtins import error_message from tinycss.css21 import CSS21Parser from tinycss.parsing import ParseError, remove_whitespace, split_on_comma class MediaQuery: __slots__ = 'media_type', 'expressions', 'negated' def __init__(self, media_type='all', expressions=(), negated=False): self.media_type = media_type self.expressions = expressions self.negated = negated def __repr__(self): return '<MediaQuery type=%s negated=%s expressions=%s>' % ( self.media_type, self.negated, self.expressions) def __eq__(self, other): return self.media_type == getattr(other, 'media_type', None) and \ self.negated == getattr(other, 'negated', None) and \ self.expressions == getattr(other, 'expressions', None) class MalformedExpression(Exception): def __init__(self, tok, msg): Exception.__init__(self, msg) self.tok = tok class CSSMedia3Parser(CSS21Parser): ''' Parse media queries as defined by the CSS 3 media module ''' def parse_media(self, tokens, errors): if not tokens: return [MediaQuery('all')] queries = [] for part in split_on_comma(remove_whitespace(tokens)): negated = False media_type = None expressions = [] try: for i, tok in enumerate(part): if i == 0 and tok.type == 'IDENT': val = tok.value.lower() if val == 'only': continue # ignore leading ONLY if val == 'not': negated = True continue if media_type is None and tok.type == 'IDENT': media_type = tok.value continue elif media_type is None: media_type = 'all' if tok.type == 'IDENT' and tok.value.lower() == 'and': continue if not tok.is_container: raise MalformedExpression(tok, 'expected a media expression not a %s' % tok.type) if tok.type != '(': raise MalformedExpression(tok, 'media expressions must be in parentheses not %s' % tok.type) content = remove_whitespace(tok.content) if len(content) == 0: raise MalformedExpression(tok, 'media expressions cannot be empty') if content[0].type != 'IDENT': raise MalformedExpression(content[0], 'expected a media feature not a %s' % tok.type) media_feature, expr = content[0].value, None if len(content) > 1: if len(content) < 3: raise MalformedExpression(content[1], 'malformed media feature definition') if content[1].type != ':': raise MalformedExpression(content[1], 'expected a :') expr = content[2:] if len(expr) == 1: expr = expr[0] elif len(expr) == 3 and (expr[0].type, expr[1].type, expr[1].value, expr[2].type) == ( 'INTEGER', 'DELIM', '/', 'INTEGER'): # This should really be moved into token_data, but # since RATIO is not part of CSS 2.1 and does not # occur anywhere else, we special case it here. r = expr[0] r.value = (expr[0].value, expr[2].value) r.type = 'RATIO' r._as_css = expr[0]._as_css + expr[1]._as_css + expr[2]._as_css expr = r else: raise MalformedExpression(expr[0], 'malformed media feature definition') expressions.append((media_feature, expr)) except MalformedExpression as err: errors.append(ParseError(err.tok, error_message(err))) media_type, negated, expressions = 'all', True, () queries.append(MediaQuery(media_type or 'all', expressions=tuple(expressions), negated=negated)) return queries
4,582
Python
.py
86
36
116
0.50983
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,842
decoding.py
kovidgoyal_calibre/src/tinycss/decoding.py
# coding: utf8 """ tinycss.decoding ---------------- Decoding stylesheets from bytes to Unicode. http://www.w3.org/TR/CSS21/syndata.html#charset :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ import operator import re from polyglot.binary import from_hex_bytes __all__ = ['decode'] # Everything else is implementation detail def decode(css_bytes, protocol_encoding=None, linking_encoding=None, document_encoding=None): """ Determine the character encoding from the passed metadata and the ``@charset`` rule in the stylesheet (if any); and decode accordingly. If no encoding information is available or decoding fails, decoding defaults to UTF-8 and then fall back on ISO-8859-1. :param css_bytes: a CSS stylesheet as a byte string :param protocol_encoding: The "charset" parameter of a "Content-Type" HTTP header (if any), or similar metadata for other protocols. :param linking_encoding: ``<link charset="">`` or other metadata from the linking mechanism (if any) :param document_encoding: Encoding of the referring style sheet or document (if any) :return: A tuple of an Unicode string, with any BOM removed, and the encoding that was used. """ if protocol_encoding: css_unicode = try_encoding(css_bytes, protocol_encoding) if css_unicode is not None: return css_unicode, protocol_encoding for encoding, pattern in ENCODING_MAGIC_NUMBERS: match = pattern(css_bytes) if match: has_at_charset = isinstance(encoding, tuple) if has_at_charset: extract, endianness = encoding encoding = extract(match.group(1)) # Get an ASCII-only unicode value. # This is the only thing that works on both Python 2 and 3 # for bytes.decode() # Non-ASCII encoding names are invalid anyway, # but make sure they stay invalid. encoding = encoding.decode('ascii', 'replace') encoding = encoding.replace('\ufffd', '?') if encoding.replace('-', '').replace('_', '').lower() in [ 'utf16', 'utf32']: encoding += endianness encoding = encoding.encode('ascii', 'replace').decode('ascii') css_unicode = try_encoding(css_bytes, encoding) if css_unicode and not (has_at_charset and not css_unicode.startswith('@charset "')): return css_unicode, encoding break for encoding in [linking_encoding, document_encoding]: if encoding: css_unicode = try_encoding(css_bytes, encoding) if css_unicode is not None: return css_unicode, encoding css_unicode = try_encoding(css_bytes, 'UTF-8') if css_unicode is not None: return css_unicode, 'UTF-8' return try_encoding(css_bytes, 'ISO-8859-1', fallback=False), 'ISO-8859-1' def try_encoding(css_bytes, encoding, fallback=True): if fallback: try: css_unicode = css_bytes.decode(encoding) # LookupError means unknown encoding except (UnicodeDecodeError, LookupError): return None else: css_unicode = css_bytes.decode(encoding) if css_unicode and css_unicode[0] == '\ufeff': # Remove any Byte Order Mark css_unicode = css_unicode[1:] return css_unicode def hex2re(hex_data): return re.escape(from_hex_bytes(hex_data.replace(' ', '').encode('ascii'))) class Slicer: """Slice()[start:stop:end] == slice(start, stop, end)""" def __getitem__(self, slice_): return operator.itemgetter(slice_) Slice = Slicer() # List of (bom_size, encoding, pattern) # bom_size is in bytes and can be zero # encoding is a string or (slice_, endianness) for "as specified" # slice_ is a slice object.How to extract the specified ENCODING_MAGIC_NUMBERS = [ ((Slice[:], ''), re.compile( hex2re('EF BB BF 40 63 68 61 72 73 65 74 20 22') + b'([^\x22]*?)' + hex2re('22 3B')).match), ('UTF-8', re.compile( hex2re('EF BB BF')).match), ((Slice[:], ''), re.compile( hex2re('40 63 68 61 72 73 65 74 20 22') + b'([^\x22]*?)' + hex2re('22 3B')).match), ((Slice[1::2], '-BE'), re.compile( hex2re('FE FF 00 40 00 63 00 68 00 61 00 72 00 73 00 65 00' '74 00 20 00 22') + b'((\x00[^\x22])*?)' + hex2re('00 22 00 3B')).match), ((Slice[1::2], '-BE'), re.compile( hex2re('00 40 00 63 00 68 00 61 00 72 00 73 00 65 00 74 00' '20 00 22') + b'((\x00[^\x22])*?)' + hex2re('00 22 00 3B')).match), ((Slice[::2], '-LE'), re.compile( hex2re('FF FE 40 00 63 00 68 00 61 00 72 00 73 00 65 00 74' '00 20 00 22 00') + b'(([^\x22]\x00)*?)' + hex2re('22 00 3B 00')).match), ((Slice[::2], '-LE'), re.compile( hex2re('40 00 63 00 68 00 61 00 72 00 73 00 65 00 74 00 20' '00 22 00') + b'(([^\x22]\x00)*?)' + hex2re('22 00 3B 00')).match), ((Slice[3::4], '-BE'), re.compile( hex2re('00 00 FE FF 00 00 00 40 00 00 00 63 00 00 00 68 00' '00 00 61 00 00 00 72 00 00 00 73 00 00 00 65 00 00' '00 74 00 00 00 20 00 00 00 22') + b'((\x00\x00\x00[^\x22])*?)' + hex2re('00 00 00 22 00 00 00 3B')).match), ((Slice[3::4], '-BE'), re.compile( hex2re('00 00 00 40 00 00 00 63 00 00 00 68 00 00 00 61 00' '00 00 72 00 00 00 73 00 00 00 65 00 00 00 74 00 00' '00 20 00 00 00 22') + b'((\x00\x00\x00[^\x22])*?)' + hex2re('00 00 00 22 00 00 00 3B')).match), # Python does not support 2143 or 3412 endianness, AFAIK. # I guess we could fix it up ourselves but meh. Patches welcome. # ((Slice[2::4], '-2143'), re.compile( # hex2re('00 00 FF FE 00 00 40 00 00 00 63 00 00 00 68 00 00' # '00 61 00 00 00 72 00 00 00 73 00 00 00 65 00 00 00' # '74 00 00 00 20 00 00 00 22 00') # + b'((\x00\x00[^\x22]\x00)*?)' # + hex2re('00 00 22 00 00 00 3B 00')).match), # ((Slice[2::4], '-2143'), re.compile( # hex2re('00 00 40 00 00 00 63 00 00 00 68 00 00 00 61 00 00' # '00 72 00 00 00 73 00 00 00 65 00 00 00 74 00 00 00' # '20 00 00 00 22 00') # + b'((\x00\x00[^\x22]\x00)*?)' # + hex2re('00 00 22 00 00 00 3B 00')).match), # ((Slice[1::4], '-3412'), re.compile( # hex2re('FE FF 00 00 00 40 00 00 00 63 00 00 00 68 00 00 00' # '61 00 00 00 72 00 00 00 73 00 00 00 65 00 00 00 74' # '00 00 00 20 00 00 00 22 00 00') # + b'((\x00[^\x22]\x00\x00)*?)' # + hex2re('00 22 00 00 00 3B 00 00')).match), # ((Slice[1::4], '-3412'), re.compile( # hex2re('00 40 00 00 00 63 00 00 00 68 00 00 00 61 00 00 00' # '72 00 00 00 73 00 00 00 65 00 00 00 74 00 00 00 20' # '00 00 00 22 00 00') # + b'((\x00[^\x22]\x00\x00)*?)' # + hex2re('00 22 00 00 00 3B 00 00')).match), ((Slice[::4], '-LE'), re.compile( hex2re('FF FE 00 00 40 00 00 00 63 00 00 00 68 00 00 00 61' '00 00 00 72 00 00 00 73 00 00 00 65 00 00 00 74 00' '00 00 20 00 00 00 22 00 00 00') + b'(([^\x22]\x00\x00\x00)*?)' + hex2re('22 00 00 00 3B 00 00 00')).match), ((Slice[::4], '-LE'), re.compile( hex2re('40 00 00 00 63 00 00 00 68 00 00 00 61 00 00 00 72' '00 00 00 73 00 00 00 65 00 00 00 74 00 00 00 20 00' '00 00 22 00 00 00') + b'(([^\x22]\x00\x00\x00)*?)' + hex2re('22 00 00 00 3B 00 00 00')).match), ('UTF-32-BE', re.compile( hex2re('00 00 FE FF')).match), ('UTF-32-LE', re.compile( hex2re('FF FE 00 00')).match), # ('UTF-32-2143', re.compile( # hex2re('00 00 FF FE')).match), # ('UTF-32-3412', re.compile( # hex2re('FE FF 00 00')).match), ('UTF-16-BE', re.compile( hex2re('FE FF')).match), ('UTF-16-LE', re.compile( hex2re('FF FE')).match), # Some of there are supported by Python, but I didn’t bother. # You know the story with patches ... # # as specified, transcoded from EBCDIC to ASCII # ('as_specified-EBCDIC', re.compile( # hex2re('7C 83 88 81 99 A2 85 A3 40 7F') # + b'([^\x7F]*?)' # + hex2re('7F 5E')).match), # # as specified, transcoded from IBM1026 to ASCII # ('as_specified-IBM1026', re.compile( # hex2re('AE 83 88 81 99 A2 85 A3 40 FC') # + b'([^\xFC]*?)' # + hex2re('FC 5E')).match), # # as specified, transcoded from GSM 03.38 to ASCII # ('as_specified-GSM_03.38', re.compile( # hex2re('00 63 68 61 72 73 65 74 20 22') # + b'([^\x22]*?)' # + hex2re('22 3B')).match), ]
9,085
Python
.py
205
37.097561
79
0.557858
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,843
css21.py
kovidgoyal_calibre/src/tinycss/css21.py
# coding: utf8 """ tinycss.css21 ------------- Parser for CSS 2.1 http://www.w3.org/TR/CSS21/syndata.html :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ from itertools import chain, islice from tinycss.decoding import decode from tinycss.parsing import ParseError, remove_whitespace, split_on_comma, strip_whitespace, validate_any, validate_value from tinycss.token_data import TokenList from tinycss.tokenizer import tokenize_grouped # stylesheet : [ CDO | CDC | S | statement ]*; # statement : ruleset | at-rule; # at-rule : ATKEYWORD S* any* [ block | ';' S* ]; # block : '{' S* [ any | block | ATKEYWORD S* | ';' S* ]* '}' S*; # ruleset : selector? '{' S* declaration? [ ';' S* declaration? ]* '}' S*; # selector : any+; # declaration : property S* ':' S* value; # property : IDENT; # value : [ any | block | ATKEYWORD S* ]+; # any : [ IDENT | NUMBER | PERCENTAGE | DIMENSION | STRING # | DELIM | URI | HASH | UNICODE-RANGE | INCLUDES # | DASHMATCH | ':' | FUNCTION S* [any|unused]* ')' # | '(' S* [any|unused]* ')' | '[' S* [any|unused]* ']' # ] S*; # unused : block | ATKEYWORD S* | ';' S* | CDO S* | CDC S*; class Stylesheet: """ A parsed CSS stylesheet. .. attribute:: rules A mixed list, in source order, of :class:`RuleSet` and various at-rules such as :class:`ImportRule`, :class:`MediaRule` and :class:`PageRule`. Use their :obj:`at_keyword` attribute to distinguish them. .. attribute:: errors A list of :class:`~.parsing.ParseError`. Invalid rules and declarations are ignored, with the details logged in this list. .. attribute:: encoding The character encoding that was used to decode the stylesheet from bytes, or ``None`` for Unicode stylesheets. """ def __init__(self, rules, errors, encoding): self.rules = rules self.errors = errors self.encoding = encoding def __repr__(self): return '<{0.__class__.__name__} {1} rules {2} errors>'.format( self, len(self.rules), len(self.errors)) class AtRule: """ An unparsed at-rule. .. attribute:: at_keyword The normalized (lower-case) at-keyword as a string. Eg: ``'@page'`` .. attribute:: head The part of the at-rule between the at-keyword and the ``{`` marking the body, or the ``;`` marking the end of an at-rule without a body. A :class:`~.token_data.TokenList`. .. attribute:: body The content of the body between ``{`` and ``}`` as a :class:`~.token_data.TokenList`, or ``None`` if there is no body (ie. if the rule ends with ``;``). The head was validated against the core grammar but **not** the body, as the body might contain declarations. In case of an error in a declaration, parsing should continue from the next declaration. The whole rule should not be ignored as it would be for an error in the head. These at-rules are expected to be parsed further before reaching the user API. """ __slots__ = 'at_keyword', 'head', 'body', 'line', 'column' def __init__(self, at_keyword, head, body, line, column): self.at_keyword = at_keyword self.head = TokenList(head) self.body = TokenList(body) if body is not None else body self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} {0.line}:{0.column} {0.at_keyword}>' .format(self)) class RuleSet: """A ruleset. .. attribute:: at_keyword Always ``None``. Helps to tell rulesets apart from at-rules. .. attribute:: selector The selector as a :class:`~.token_data.TokenList`. In CSS 3, this is actually called a selector group. ``rule.selector.as_css()`` gives the selector as a string. This string can be used with *cssselect*, see :ref:`selectors3`. .. attribute:: declarations The list of :class:`Declaration`, in source order. """ at_keyword = None __slots__ = 'selector', 'declarations', 'line', 'column' def __init__(self, selector, declarations, line, column): self.selector = TokenList(selector) self.declarations = declarations self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} at {0.line}:{0.column} {1}>' .format(self, self.selector.as_css())) class Declaration: """A property declaration. .. attribute:: name The property name as a normalized (lower-case) string. .. attribute:: value The property value as a :class:`~.token_data.TokenList`. The value is not parsed. UAs using tinycss may only support some properties or some values and tinycss does not know which. They need to parse values themselves and ignore declarations with unknown or unsupported properties or values, and fall back on any previous declaration. :mod:`tinycss.color3` parses color values, but other values will need specific parsing/validation code. .. attribute:: priority Either the string ``'important'`` or ``None``. """ __slots__ = 'name', 'value', 'priority', 'line', 'column' def __init__(self, name, value, priority, line, column): self.name = name self.value = TokenList(value) self.priority = priority self.line = line self.column = column def __repr__(self): priority = ' !' + self.priority if self.priority else '' return ('<{0.__class__.__name__} {0.line}:{0.column}' ' {0.name}: {1}{2}>'.format( self, self.value.as_css(), priority)) class PageRule: """A parsed CSS 2.1 @page rule. .. attribute:: at_keyword Always ``'@page'`` .. attribute:: selector The page selector. In CSS 2.1 this is either ``None`` (no selector), or the string ``'first'``, ``'left'`` or ``'right'`` for the pseudo class of the same name. .. attribute:: specificity Specificity of the page selector. This is a tuple of four integers, but these tuples are mostly meant to be compared to each other. .. attribute:: declarations A list of :class:`Declaration`, in source order. .. attribute:: at_rules The list of parsed at-rules inside the @page block, in source order. Always empty for CSS 2.1. """ at_keyword = '@page' __slots__ = 'selector', 'specificity', 'declarations', 'at_rules', 'line', 'column' def __init__(self, selector, specificity, declarations, at_rules, line, column): self.selector = selector self.specificity = specificity self.declarations = declarations self.at_rules = at_rules self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} {0.line}:{0.column}' ' {0.selector}>'.format(self)) class MediaRule: """A parsed @media rule. .. attribute:: at_keyword Always ``'@media'`` .. attribute:: media For CSS 2.1 without media queries: the media types as a list of strings. .. attribute:: rules The list :class:`RuleSet` and various at-rules inside the @media block, in source order. """ at_keyword = '@media' __slots__ = 'media', 'rules', 'line', 'column' def __init__(self, media, rules, line, column): self.media = media self.rules = rules self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} {0.line}:{0.column}' ' {0.media}>'.format(self)) class ImportRule: """A parsed @import rule. .. attribute:: at_keyword Always ``'@import'`` .. attribute:: uri The URI to be imported, as read from the stylesheet. (URIs are not made absolute.) .. attribute:: media For CSS 2.1 without media queries: the media types as a list of strings. This attribute is explicitly ``['all']`` if the media was omitted in the source. """ at_keyword = '@import' __slots__ = 'uri', 'media', 'line', 'column' def __init__(self, uri, media, line, column): self.uri = uri self.media = media self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} {0.line}:{0.column}' ' {0.uri}>'.format(self)) def _remove_at_charset(tokens): """Remove any valid @charset at the beginning of a token stream. :param tokens: An iterable of tokens :returns: A possibly truncated iterable of tokens """ tokens = iter(tokens) header = list(islice(tokens, 4)) if [t.type for t in header] == ['ATKEYWORD', 'S', 'STRING', ';']: atkw, space, string, semicolon = header if ((atkw.value, space.value) == ('@charset', ' ') and string.as_css()[0] == '"'): # Found a valid @charset rule, only keep what’s after it. return tokens return chain(header, tokens) class CSS21Parser: """Parser for CSS 2.1 This parser supports the core CSS syntax as well as @import, @media, @page and !important. Note that property values are still not parsed, as UAs using this parser may only support some properties or some values. Currently the parser holds no state. It being a class only allows subclassing and overriding its methods. """ def __init__(self): self.at_parsers = { '@' + x:getattr(self, 'parse_%s_rule' % x) for x in ('media', 'page', 'import', 'charset')} # User API: def parse_stylesheet_file(self, css_file, protocol_encoding=None, linking_encoding=None, document_encoding=None): """Parse a stylesheet from a file or filename. Character encoding-related parameters and behavior are the same as in :meth:`parse_stylesheet_bytes`. :param css_file: Either a file (any object with a :meth:`~file.read` method) or a filename. :return: A :class:`Stylesheet`. """ if hasattr(css_file, 'read'): css_bytes = css_file.read() else: with open(css_file, 'rb') as fd: css_bytes = fd.read() return self.parse_stylesheet_bytes(css_bytes, protocol_encoding, linking_encoding, document_encoding) def parse_stylesheet_bytes(self, css_bytes, protocol_encoding=None, linking_encoding=None, document_encoding=None): """Parse a stylesheet from a byte string. The character encoding is determined from the passed metadata and the ``@charset`` rule in the stylesheet (if any). If no encoding information is available or decoding fails, decoding defaults to UTF-8 and then fall back on ISO-8859-1. :param css_bytes: A CSS stylesheet as a byte string. :param protocol_encoding: The "charset" parameter of a "Content-Type" HTTP header (if any), or similar metadata for other protocols. :param linking_encoding: ``<link charset="">`` or other metadata from the linking mechanism (if any) :param document_encoding: Encoding of the referring style sheet or document (if any) :return: A :class:`Stylesheet`. """ css_unicode, encoding = decode(css_bytes, protocol_encoding, linking_encoding, document_encoding) return self.parse_stylesheet(css_unicode, encoding=encoding) def parse_stylesheet(self, css_unicode, encoding=None): """Parse a stylesheet from an Unicode string. :param css_unicode: A CSS stylesheet as an unicode string. :param encoding: The character encoding used to decode the stylesheet from bytes, if any. :return: A :class:`Stylesheet`. """ tokens = tokenize_grouped(css_unicode) if encoding: tokens = _remove_at_charset(tokens) rules, errors = self.parse_rules(tokens, context='stylesheet') return Stylesheet(rules, errors, encoding) def parse_style_attr(self, css_source): """Parse a "style" attribute (eg. of an HTML element). This method only accepts Unicode as the source (HTML) document is supposed to handle the character encoding. :param css_source: The attribute value, as an unicode string. :return: A tuple of the list of valid :class:`Declaration` and a list of :class:`~.parsing.ParseError`. """ return self.parse_declaration_list(tokenize_grouped(css_source)) # API for subclasses: def parse_rules(self, tokens, context): """Parse a sequence of rules (rulesets and at-rules). :param tokens: An iterable of tokens. :param context: Either ``'stylesheet'`` or an at-keyword such as ``'@media'``. (Most at-rules are only allowed in some contexts.) :return: A tuple of a list of parsed rules and a list of :class:`~.parsing.ParseError`. """ rules = [] errors = [] tokens = iter(tokens) for token in tokens: if token.type not in ('S', 'CDO', 'CDC'): try: if token.type == 'ATKEYWORD': rule = self.read_at_rule(token, tokens) result = self.parse_at_rule( rule, rules, errors, context) rules.append(result) else: rule, rule_errors = self.parse_ruleset(token, tokens) rules.append(rule) errors.extend(rule_errors) except ParseError as exc: errors.append(exc) # Skip the entire rule return rules, errors def read_at_rule(self, at_keyword_token, tokens): """Read an at-rule from a token stream. :param at_keyword_token: The ATKEYWORD token that starts this at-rule You may have read it already to distinguish the rule from a ruleset. :param tokens: An iterator of subsequent tokens. Will be consumed just enough for one at-rule. :return: An unparsed :class:`AtRule`. :raises: :class:`~.parsing.ParseError` if the head is invalid for the core grammar. The body is **not** validated. See :class:`AtRule`. """ # CSS syntax is case-insensitive at_keyword = at_keyword_token.value.lower() head = [] # For the ParseError in case `tokens` is empty: token = at_keyword_token for token in tokens: if token.type in '{;': break # Ignore white space just after the at-keyword. else: head.append(token) # On unexpected end of stylesheet, pretend that a ';' was there head = strip_whitespace(head) for head_token in head: validate_any(head_token, 'at-rule head') body = token.content if token.type == '{' else None return AtRule(at_keyword, head, body, at_keyword_token.line, at_keyword_token.column) def parse_at_rule(self, rule, previous_rules, errors, context): """Parse an at-rule. Subclasses that override this method must use ``super()`` and pass its return value for at-rules they do not know. In CSS 2.1, this method handles @charset, @import, @media and @page rules. :param rule: An unparsed :class:`AtRule`. :param previous_rules: The list of at-rules and rulesets that have been parsed so far in this context. This list can be used to decide if the current rule is valid. (For example, @import rules are only allowed before anything but a @charset rule.) :param context: Either ``'stylesheet'`` or an at-keyword such as ``'@media'``. (Most at-rules are only allowed in some contexts.) :raises: :class:`~.parsing.ParseError` if the rule is invalid. :return: A parsed at-rule """ try: parser = self.at_parsers[rule.at_keyword] except KeyError: raise ParseError(rule, 'unknown at-rule in {0} context: {1}' .format(context, rule.at_keyword)) else: return parser(rule, previous_rules, errors, context) def parse_page_rule(self, rule, previous_rules, errors, context): if context != 'stylesheet': raise ParseError(rule, '@page rule not allowed in ' + context) selector, specificity = self.parse_page_selector(rule.head) if rule.body is None: raise ParseError(rule, 'invalid {0} rule: missing block'.format(rule.at_keyword)) declarations, at_rules, rule_errors = \ self.parse_declarations_and_at_rules(rule.body, '@page') errors.extend(rule_errors) return PageRule(selector, specificity, declarations, at_rules, rule.line, rule.column) def parse_media_rule(self, rule, previous_rules, errors, context): if context != 'stylesheet': raise ParseError(rule, '@media rule not allowed in ' + context) media = self.parse_media(rule.head, errors) if rule.body is None: raise ParseError(rule, 'invalid {0} rule: missing block'.format(rule.at_keyword)) rules, rule_errors = self.parse_rules(rule.body, '@media') errors.extend(rule_errors) return MediaRule(media, rules, rule.line, rule.column) def parse_import_rule(self, rule, previous_rules, errors, context): if context != 'stylesheet': raise ParseError(rule, '@import rule not allowed in ' + context) for previous_rule in previous_rules: if previous_rule.at_keyword not in ('@charset', '@import'): if previous_rule.at_keyword: type_ = 'an {0} rule'.format(previous_rule.at_keyword) else: type_ = 'a ruleset' raise ParseError(previous_rule, '@import rule not allowed after ' + type_) head = rule.head if not head: raise ParseError(rule, 'expected URI or STRING for @import rule') if head[0].type not in ('URI', 'STRING'): raise ParseError(rule, 'expected URI or STRING for @import rule, got ' + head[0].type) uri = head[0].value media = self.parse_media(strip_whitespace(head[1:]), errors) if rule.body is not None: # The position of the ';' token would be best, but we don’t # have it anymore here. raise ParseError(head[-1], "expected ';', got a block") return ImportRule(uri, media, rule.line, rule.column) def parse_charset_rule(self, rule, previous_rules, errors, context): raise ParseError(rule, 'mis-placed or malformed @charset rule') def parse_media(self, tokens, errors): """For CSS 2.1, parse a list of media types. Media Queries are expected to override this. :param tokens: A list of tokens :raises: :class:`~.parsing.ParseError` on invalid media types/queries :returns: For CSS 2.1, a list of media types as strings """ if not tokens: return ['all'] media_types = [] for part in split_on_comma(remove_whitespace(tokens)): types = [token.type for token in part] if types == ['IDENT']: media_types.append(part[0].value) else: raise ParseError(tokens[0], 'expected a media type' + ((', got ' + ', '.join(types)) if types else '')) return media_types def parse_page_selector(self, tokens): """Parse an @page selector. :param tokens: An iterable of token, typically from the ``head`` attribute of an unparsed :class:`AtRule`. :returns: A page selector. For CSS 2.1, this is ``'first'``, ``'left'``, ``'right'`` or ``None``. :raises: :class:`~.parsing.ParseError` on invalid selectors """ if not tokens: return None, (0, 0) if (len(tokens) == 2 and tokens[0].type == ':' and tokens[1].type == 'IDENT'): pseudo_class = tokens[1].value specificity = { 'first': (1, 0), 'left': (0, 1), 'right': (0, 1), }.get(pseudo_class) if specificity: return pseudo_class, specificity raise ParseError(tokens[0], 'invalid @page selector') def parse_declarations_and_at_rules(self, tokens, context): """Parse a mixed list of declarations and at rules, as found eg. in the body of an @page rule. Note that to add supported at-rules inside @page, :class:`~.page3.CSSPage3Parser` extends :meth:`parse_at_rule`, not this method. :param tokens: An iterable of token, typically from the ``body`` attribute of an unparsed :class:`AtRule`. :param context: An at-keyword such as ``'@page'``. (Most at-rules are only allowed in some contexts.) :returns: A tuple of: * A list of :class:`Declaration` * A list of parsed at-rules (empty for CSS 2.1) * A list of :class:`~.parsing.ParseError` """ at_rules = [] declarations = [] errors = [] tokens = iter(tokens) for token in tokens: if token.type == 'ATKEYWORD': try: rule = self.read_at_rule(token, tokens) result = self.parse_at_rule( rule, at_rules, errors, context) at_rules.append(result) except ParseError as err: errors.append(err) elif token.type != 'S': declaration_tokens = [] while token and token.type != ';': declaration_tokens.append(token) token = next(tokens, None) if declaration_tokens: try: declarations.append( self.parse_declaration(declaration_tokens)) except ParseError as err: errors.append(err) return declarations, at_rules, errors def parse_ruleset(self, first_token, tokens): """Parse a ruleset: a selector followed by declaration block. :param first_token: The first token of the ruleset (probably of the selector). You may have read it already to distinguish the rule from an at-rule. :param tokens: an iterator of subsequent tokens. Will be consumed just enough for one ruleset. :return: a tuple of a :class:`RuleSet` and an error list. The errors are recovered :class:`~.parsing.ParseError` in declarations. (Parsing continues from the next declaration on such errors.) :raises: :class:`~.parsing.ParseError` if the selector is invalid for the core grammar. Note a that a selector can be valid for the core grammar but not for CSS 2.1 or another level. """ selector = [] for token in chain([first_token], tokens): if token.type == '{': # Parse/validate once we’ve read the whole rule selector = strip_whitespace(selector) if not selector: raise ParseError(first_token, 'empty selector') for selector_token in selector: validate_any(selector_token, 'selector') declarations, errors = self.parse_declaration_list( token.content) ruleset = RuleSet(selector, declarations, first_token.line, first_token.column) return ruleset, errors else: selector.append(token) raise ParseError(token, 'no declaration block found for ruleset') def parse_declaration_list(self, tokens): """Parse a ``;`` separated declaration list. You may want to use :meth:`parse_declarations_and_at_rules` (or some other method that uses :func:`parse_declaration` directly) instead if you have not just declarations in the same context. :param tokens: an iterable of tokens. Should stop at (before) the end of the block, as marked by ``}``. :return: a tuple of the list of valid :class:`Declaration` and a list of :class:`~.parsing.ParseError` """ # split at ';' parts = [] this_part = [] for token in tokens: if token.type == ';': parts.append(this_part) this_part = [] else: this_part.append(token) parts.append(this_part) declarations = [] errors = [] for tokens in parts: tokens = strip_whitespace(tokens) if tokens: try: declarations.append(self.parse_declaration(tokens)) except ParseError as exc: errors.append(exc) # Skip the entire declaration return declarations, errors def parse_declaration(self, tokens): """Parse a single declaration. :param tokens: an iterable of at least one token. Should stop at (before) the end of the declaration, as marked by a ``;`` or ``}``. Empty declarations (ie. consecutive ``;`` with only white space in-between) should be skipped earlier and not passed to this method. :returns: a :class:`Declaration` :raises: :class:`~.parsing.ParseError` if the tokens do not match the 'declaration' production of the core grammar. """ tokens = iter(tokens) name_token = next(tokens) # assume there is at least one if name_token.type == 'IDENT': # CSS syntax is case-insensitive property_name = name_token.value.lower() else: raise ParseError(name_token, 'expected a property name, got {0}'.format(name_token.type)) token = name_token # In case ``tokens`` is now empty for token in tokens: if token.type == ':': break elif token.type != 'S': raise ParseError( token, "expected ':', got {0}".format(token.type)) else: raise ParseError(token, "expected ':'") value = strip_whitespace(list(tokens)) if not value: raise ParseError(token, 'expected a property value') validate_value(value) value, priority = self.parse_value_priority(value) return Declaration( property_name, value, priority, name_token.line, name_token.column) def parse_value_priority(self, tokens): """Separate any ``!important`` marker at the end of a property value. :param tokens: A list of tokens for the property value. :returns: A tuple of the actual property value (a list of tokens) and the :attr:`~Declaration.priority`. """ value = list(tokens) # Walk the token list from the end token = value.pop() if token.type == 'IDENT' and token.value.lower() == 'important': while value: token = value.pop() if token.type == 'DELIM' and token.value == '!': # Skip any white space before the '!' while value and value[-1].type == 'S': value.pop() if not value: raise ParseError( token, 'expected a value before !important') return value, 'important' # Skip white space between '!' and 'important' elif token.type != 'S': break return tokens, None
29,393
Python
.py
665
33.403008
121
0.57423
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,844
__init__.py
kovidgoyal_calibre/src/tinycss/__init__.py
# coding: utf8 """ tinycss ------- A CSS parser, and nothing else. :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ from .version import VERSION __version__ = VERSION from tinycss.css21 import CSS21Parser from tinycss.fonts3 import CSSFonts3Parser from tinycss.media3 import CSSMedia3Parser from tinycss.page3 import CSSPage3Parser PARSER_MODULES = { 'page3': CSSPage3Parser, 'fonts3': CSSFonts3Parser, 'media3': CSSMedia3Parser, } def make_parser(*features, **kwargs): """Make a parser object with the chosen features. :param features: Positional arguments are base classes the new parser class will extend. The string ``'page3'`` is accepted as short for :class:`~page3.CSSPage3Parser`. :param kwargs: Keyword arguments are passed to the parser’s constructor. :returns: An instance of a new subclass of :class:`CSS21Parser` """ if features: bases = tuple(PARSER_MODULES.get(f, f) for f in features) parser_class = type('CustomCSSParser', bases + (CSS21Parser,), {}) else: parser_class = CSS21Parser return parser_class(**kwargs) def make_full_parser(**kwargs): ''' A parser that parses all supported CSS 3 modules in addition to CSS 2.1 ''' features = tuple(PARSER_MODULES) return make_parser(*features, **kwargs)
1,407
Python
.py
40
30.375
83
0.698155
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,845
fonts3.py
kovidgoyal_calibre/src/tinycss/fonts3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' import re from tinycss.css21 import CSS21Parser, ParseError from .tokenizer import tokenize_grouped def parse_font_family_tokens(tokens): families = [] current_family = '' def commit(): val = current_family.strip() if val: families.append(val) for token in tokens: if token.type == 'STRING': if current_family: commit() current_family = token.value elif token.type == 'DELIM': if token.value == ',': if current_family: commit() current_family = '' elif token.type == 'IDENT': current_family += ' ' + token.value if current_family: commit() return families def parse_font_family(css_string): return parse_font_family_tokens(tokenize_grouped(type('')(css_string).strip())) def serialize_single_font_family(x): xl = x.lower() if xl in GENERIC_FAMILIES: if xl == 'sansserif': xl = 'sans-serif' return xl if SIMPLE_NAME_PAT.match(x) is not None and not x.lower().startswith('and'): # css_parser dies if a font name starts with and return x return '"%s"' % x.replace('"', r'\"') def serialize_font_family(families): return ', '.join(map(serialize_single_font_family, families)) GLOBAL_IDENTS = frozenset('inherit initial unset normal'.split()) STYLE_IDENTS = frozenset('italic oblique'.split()) VARIANT_IDENTS = frozenset(('small-caps',)) WEIGHT_IDENTS = frozenset('bold bolder lighter'.split()) STRETCH_IDENTS = frozenset('ultra-condensed extra-condensed condensed semi-condensed semi-expanded expanded extra-expanded ultra-expanded'.split()) BEFORE_SIZE_IDENTS = STYLE_IDENTS | VARIANT_IDENTS | WEIGHT_IDENTS | STRETCH_IDENTS SIZE_IDENTS = frozenset('xx-small x-small small medium large x-large xx-large larger smaller'.split()) WEIGHT_SIZES = frozenset(map(int, '100 200 300 400 500 600 700 800 900'.split())) LEGACY_FONT_SPEC = frozenset('caption icon menu message-box small-caption status-bar'.split()) GENERIC_FAMILIES = frozenset('serif sans-serif sansserif cursive fantasy monospace'.split()) SIMPLE_NAME_PAT = re.compile(r'[a-zA-Z][a-zA-Z0-9_-]*$') def serialize_font(font_dict): ans = [] for x in 'style variant weight stretch'.split(): val = font_dict.get('font-' + x) if val is not None: ans.append(val) val = font_dict.get('font-size') if val is not None: fs = val val = font_dict.get('line-height') if val is not None: fs += '/' + val ans.append(fs) val = font_dict.get('font-family') if val: ans.append(serialize_font_family(val)) return ' '.join(ans) def parse_font(css_string): # See https://www.w3.org/TR/css-fonts-3/#font-prop style = variant = weight = stretch = size = height = None tokens = list(reversed(tuple(tokenize_grouped(type('')(css_string).strip())))) if tokens and tokens[-1].value in LEGACY_FONT_SPEC: return {'font-family':['sans-serif']} while tokens: tok = tokens.pop() if tok.type == 'STRING': tokens.append(tok) break if tok.type == 'INTEGER': if size is None: if weight is None and tok.value in WEIGHT_SIZES: weight = tok.as_css() continue break if height is None: height = tok.as_css() break break if tok.type == 'NUMBER': if size is not None and height is None: height = tok.as_css() break if tok.type == 'DELIM': if tok.value == '/' and size is not None and height is None: continue break if tok.type in ('DIMENSION', 'PERCENTAGE'): if size is None: size = tok.as_css() continue if height is None: height = tok.as_css() break if tok.type == 'IDENT': if tok.value in GLOBAL_IDENTS: if size is not None: if height is None: height = tok.value else: tokens.append(tok) break if style is None: style = tok.value elif variant is None: variant = tok.value elif weight is None: weight = tok.value elif stretch is None: stretch = tok.value elif size is None: size = tok.value elif height is None: height = tok.value break else: tokens.append(tok) break continue if tok.value in BEFORE_SIZE_IDENTS: if size is not None: break if tok.value in STYLE_IDENTS: style = tok.value elif tok.value in VARIANT_IDENTS: variant = tok.value elif tok.value in WEIGHT_IDENTS: weight = tok.value elif tok.value in STRETCH_IDENTS: stretch = tok.value elif tok.value in SIZE_IDENTS: size = tok.value else: tokens.append(tok) break families = parse_font_family_tokens(reversed(tokens)) ans = {} if style is not None: ans['font-style'] = style if variant is not None: ans['font-variant'] = variant if weight is not None: ans['font-weight'] = weight if stretch is not None: ans['font-stretch'] = stretch if size is not None: ans['font-size'] = size if height is not None: ans['line-height'] = height if families: ans['font-family'] = families return ans class FontFaceRule: at_keyword = '@font-face' __slots__ = 'declarations', 'line', 'column' def __init__(self, declarations, line, column): self.declarations = declarations self.line = line self.column = column def __repr__(self): return ('<{0.__class__.__name__} at {0.line}:{0.column}>' .format(self)) class CSSFonts3Parser(CSS21Parser): ''' Parse @font-face rules from the CSS 3 fonts module ''' ALLOWED_CONTEXTS_FOR_FONT_FACE = {'stylesheet', '@media', '@page'} def __init__(self): super(CSSFonts3Parser, self).__init__() self.at_parsers['@font-face'] = self.parse_font_face_rule def parse_font_face_rule(self, rule, previous_rules, errors, context): if context not in self.ALLOWED_CONTEXTS_FOR_FONT_FACE: raise ParseError(rule, '@font-face rule not allowed in ' + context) if rule.body is None: raise ParseError(rule, 'invalid {0} rule: missing block'.format(rule.at_keyword)) if rule.head: raise ParseError(rule, '{0} rule is not allowed to have content before the descriptor declaration'.format(rule.at_keyword)) declarations, decerrors = self.parse_declaration_list(rule.body) errors.extend(decerrors) return FontFaceRule(declarations, rule.line, rule.column)
7,595
Python
.py
193
28.834197
147
0.56697
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,846
page3.py
kovidgoyal_calibre/src/tinycss/page3.py
# coding: utf8 """ tinycss.page3 ------------------ Support for CSS 3 Paged Media syntax: http://dev.w3.org/csswg/css3-page/ Adds support for named page selectors and margin rules. :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ from .css21 import CSS21Parser, ParseError class MarginRule: """A parsed at-rule for margin box. .. attribute:: at_keyword One of the 16 following strings: * ``@top-left-corner`` * ``@top-left`` * ``@top-center`` * ``@top-right`` * ``@top-right-corner`` * ``@bottom-left-corner`` * ``@bottom-left`` * ``@bottom-center`` * ``@bottom-right`` * ``@bottom-right-corner`` * ``@left-top`` * ``@left-middle`` * ``@left-bottom`` * ``@right-top`` * ``@right-middle`` * ``@right-bottom`` .. attribute:: declarations A list of :class:`~.css21.Declaration` objects. .. attribute:: line Source line where this was read. .. attribute:: column Source column where this was read. """ __slots__ = 'at_keyword', 'declarations', 'line', 'column' def __init__(self, at_keyword, declarations, line, column): self.at_keyword = at_keyword self.declarations = declarations self.line = line self.column = column class CSSPage3Parser(CSS21Parser): """Extend :class:`~.css21.CSS21Parser` for `CSS 3 Paged Media`_ syntax. .. _CSS 3 Paged Media: http://dev.w3.org/csswg/css3-page/ Compared to CSS 2.1, the ``at_rules`` and ``selector`` attributes of :class:`~.css21.PageRule` objects are modified: * ``at_rules`` is not always empty, it is a list of :class:`MarginRule` objects. * ``selector``, instead of a single string, is a tuple of the page name and the pseudo class. Each of these may be a ``None`` or a string. +--------------------------+------------------------+ | CSS | Parsed selectors | +==========================+========================+ | .. code-block:: css | .. code-block:: python | | | | | @page {} | (None, None) | | @page :first {} | (None, 'first') | | @page chapter {} | ('chapter', None) | | @page table:right {} | ('table', 'right') | +--------------------------+------------------------+ """ PAGE_MARGIN_AT_KEYWORDS = ( '@top-left-corner', '@top-left', '@top-center', '@top-right', '@top-right-corner', '@bottom-left-corner', '@bottom-left', '@bottom-center', '@bottom-right', '@bottom-right-corner', '@left-top', '@left-middle', '@left-bottom', '@right-top', '@right-middle', '@right-bottom', ) def __init__(self): super(CSSPage3Parser, self).__init__() for x in self.PAGE_MARGIN_AT_KEYWORDS: self.at_parsers[x] = self.parse_page_margin_rule def parse_page_margin_rule(self, rule, previous_rules, errors, context): if context != '@page': raise ParseError(rule, '%s rule not allowed in %s' % (rule.at_keyword, context)) if rule.head: raise ParseError(rule.head[0], 'unexpected %s token in %s rule header' % (rule.head[0].type, rule.at_keyword)) declarations, body_errors = self.parse_declaration_list(rule.body) errors.extend(body_errors) return MarginRule(rule.at_keyword, declarations, rule.line, rule.column) def parse_page_selector(self, head): """Parse an @page selector. :param head: The ``head`` attribute of an unparsed :class:`AtRule`. :returns: A page selector. For CSS 2.1, this is 'first', 'left', 'right' or None. 'blank' is added by GCPM. :raises: :class`~parsing.ParseError` on invalid selectors """ if not head: return (None, None), (0, 0, 0) if head[0].type == 'IDENT': name = head.pop(0).value while head and head[0].type == 'S': head.pop(0) if not head: return (name, None), (1, 0, 0) name_specificity = (1,) else: name = None name_specificity = (0,) if (len(head) == 2 and head[0].type == ':' and head[1].type == 'IDENT'): pseudo_class = head[1].value specificity = { 'first': (1, 0), 'blank': (1, 0), 'left': (0, 1), 'right': (0, 1), }.get(pseudo_class) if specificity: return (name, pseudo_class), (name_specificity + specificity) raise ParseError(head[0], 'invalid @page selector')
5,050
Python
.py
130
29.869231
77
0.507673
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,847
tokenizer.py
kovidgoyal_calibre/src/tinycss/tokenizer.py
# coding: utf8 """ tinycss.tokenizer ----------------- Tokenizer for the CSS core syntax: http://www.w3.org/TR/CSS21/syndata.html#tokenization This is the pure-python implementation. See also speedups.pyx :copyright: (c) 2012 by Simon Sapin. :license: BSD, see LICENSE for more details. """ from tinycss import token_data def tokenize_flat(css_source, ignore_comments=True, # Make these local variable to avoid global lookups in the loop tokens_dispatch=token_data.TOKEN_DISPATCH, unicode_unescape=token_data.UNICODE_UNESCAPE, newline_unescape=token_data.NEWLINE_UNESCAPE, simple_unescape=token_data.SIMPLE_UNESCAPE, find_newlines=token_data.FIND_NEWLINES, Token=token_data.Token, len=len, int=int, float=float, list=list, _None=None, ): """ :param css_source: CSS as an unicode string :param ignore_comments: if true (the default) comments will not be included in the return value :return: An iterator of :class:`Token` """ pos = 0 line = 1 column = 1 source_len = len(css_source) tokens = [] while pos < source_len: char = css_source[pos] if char in ':;{}()[]': type_ = char css_value = char else: codepoint = min(ord(char), 160) for _index, type_, regexp in tokens_dispatch[codepoint]: match = regexp(css_source, pos) if match is not None: # First match is the longest. See comments on TOKENS above. css_value = match.group() break else: # No match. # "Any other character not matched by the above rules, # and neither a single nor a double quote." # ... but quotes at the start of a token are always matched # by STRING or BAD_STRING. So DELIM is any single character. type_ = 'DELIM' css_value = char length = len(css_value) next_pos = pos + length # A BAD_COMMENT is a comment at EOF. Ignore it too. if not (ignore_comments and type_ in ('COMMENT', 'BAD_COMMENT')): # Parse numbers, extract strings and URIs, unescape unit = _None if type_ == 'DIMENSION': value = match.group(1) value = float(value) if '.' in value else int(value) unit = match.group(2) unit = simple_unescape(unit) unit = unicode_unescape(unit) unit = unit.lower() # normalize elif type_ == 'PERCENTAGE': value = css_value[:-1] value = float(value) if '.' in value else int(value) unit = '%' elif type_ == 'NUMBER': value = css_value if '.' in value: value = float(value) else: value = int(value) type_ = 'INTEGER' elif type_ in ('IDENT', 'ATKEYWORD', 'HASH', 'FUNCTION'): value = simple_unescape(css_value) value = unicode_unescape(value) elif type_ == 'URI': value = match.group(1) if value and value[0] in '"\'': value = value[1:-1] # Remove quotes value = newline_unescape(value) value = simple_unescape(value) value = unicode_unescape(value) elif type_ == 'STRING': value = css_value[1:-1] # Remove quotes value = newline_unescape(value) value = simple_unescape(value) value = unicode_unescape(value) # BAD_STRING can only be one of: # * Unclosed string at the end of the stylesheet: # Close the string, but this is not an error. # Make it a "good" STRING token. # * Unclosed string at the (unescaped) end of the line: # Close the string, but this is an error. # Leave it as a BAD_STRING, don’t bother parsing it. # See http://www.w3.org/TR/CSS21/syndata.html#parsing-errors elif type_ == 'BAD_STRING' and next_pos == source_len: type_ = 'STRING' value = css_value[1:] # Remove quote value = newline_unescape(value) value = simple_unescape(value) value = unicode_unescape(value) else: value = css_value tokens.append(Token(type_, css_value, value, unit, line, column)) pos = next_pos newlines = find_newlines(css_value) if newlines: line += len(newlines) # Add 1 to have lines start at column 1, not 0 column = length - newlines[-1].end() + 1 else: column += length return tokens def regroup(tokens): """ Match pairs of tokens: () [] {} function() (Strings in "" or '' are taken care of by the tokenizer.) Opening tokens are replaced by a :class:`ContainerToken`. Closing tokens are removed. Unmatched closing tokens are invalid but left as-is. All nested structures that are still open at the end of the stylesheet are implicitly closed. :param tokens: a *flat* iterable of tokens, as returned by :func:`tokenize_flat`. :return: A tree of tokens. """ # "global" objects for the inner recursion pairs = {'FUNCTION': ')', '(': ')', '[': ']', '{': '}'} tokens = iter(tokens) eof = [False] def _regroup_inner(stop_at=None, tokens=tokens, pairs=pairs, eof=eof, ContainerToken=token_data.ContainerToken, FunctionToken=token_data.FunctionToken): for token in tokens: type_ = token.type if type_ == stop_at: return end = pairs.get(type_) if end is None: yield token # Not a grouping token else: assert not isinstance(token, ContainerToken), ( 'Token looks already grouped: {0}'.format(token)) content = list(_regroup_inner(end)) if eof[0]: end = '' # Implicit end of structure at EOF. if type_ == 'FUNCTION': yield FunctionToken(token.type, token.as_css(), end, token.value, content, token.line, token.column) else: yield ContainerToken(token.type, token.as_css(), end, content, token.line, token.column) else: eof[0] = True # end of file/stylesheet return _regroup_inner() def tokenize_grouped(css_source, ignore_comments=True): """ :param css_source: CSS as an unicode string :param ignore_comments: if true (the default) comments will not be included in the return value :return: An iterator of :class:`Token` """ return regroup(tokenize_flat(css_source, ignore_comments)) # Optional Cython version of tokenize_flat # Make both versions available with explicit names for tests. python_tokenize_flat = tokenize_flat try: tok = token_data.load_c_tokenizer() except (ImportError, RuntimeError): c_tokenize_flat = None else: # Use the c tokenizer by default c_tokenize_flat = tokenize_flat = lambda s, ignore_comments=False:tok.tokenize_flat(s, ignore_comments)
7,737
Python
.py
191
29.109948
107
0.551316
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,848
color3.py
kovidgoyal_calibre/src/tinycss/tests/color3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from tinycss.color3 import hsl_to_rgb, parse_color_string from tinycss.tests import BaseTest class TestColor3(BaseTest): def test_color_parsing(self): for css_source, expected_result in [ ('', None), (' /* hey */\n', None), ('4', None), ('top', None), ('/**/transparent', (0, 0, 0, 0)), ('transparent', (0, 0, 0, 0)), (' transparent\n', (0, 0, 0, 0)), ('TransParent', (0, 0, 0, 0)), ('currentColor', 'currentColor'), ('CURRENTcolor', 'currentColor'), ('current_Color', None), ('black', (0, 0, 0, 1)), ('white', (1, 1, 1, 1)), ('fuchsia', (1, 0, 1, 1)), ('cyan', (0, 1, 1, 1)), ('CyAn', (0, 1, 1, 1)), ('darkkhaki', (189 / 255., 183 / 255., 107 / 255., 1)), ('#', None), ('#f', None), ('#ff', None), ('#fff', (1, 1, 1, 1)), ('#ffg', None), ('#ffff', None), ('#fffff', None), ('#ffffff', (1, 1, 1, 1)), ('#fffffg', None), ('#fffffff', None), ('#ffffffff', None), ('#fffffffff', None), ('#cba987', (203 / 255., 169 / 255., 135 / 255., 1)), ('#CbA987', (203 / 255., 169 / 255., 135 / 255., 1)), ('#1122aA', (17 / 255., 34 / 255., 170 / 255., 1)), ('#12a', (17 / 255., 34 / 255., 170 / 255., 1)), ('rgb(203, 169, 135)', (203 / 255., 169 / 255., 135 / 255., 1)), ('RGB(255, 255, 255)', (1, 1, 1, 1)), ('rgB(0, 0, 0)', (0, 0, 0, 1)), ('rgB(0, 51, 255)', (0, .2, 1, 1)), ('rgb(0,51,255)', (0, .2, 1, 1)), ('rgb(0\t, 51 ,255)', (0, .2, 1, 1)), ('rgb(/* R */0, /* G */51, /* B */255)', (0, .2, 1, 1)), ('rgb(-51, 306, 0)', (-.2, 1.2, 0, 1)), # out of 0..1 is allowed ('rgb(42%, 3%, 50%)', (.42, .03, .5, 1)), ('RGB(100%, 100%, 100%)', (1, 1, 1, 1)), ('rgB(0%, 0%, 0%)', (0, 0, 0, 1)), ('rgB(10%, 20%, 30%)', (.1, .2, .3, 1)), ('rgb(10%,20%,30%)', (.1, .2, .3, 1)), ('rgb(10%\t, 20% ,30%)', (.1, .2, .3, 1)), ('rgb(/* R */10%, /* G */20%, /* B */30%)', (.1, .2, .3, 1)), ('rgb(-12%, 110%, 1400%)', (-.12, 1.1, 14, 1)), # out of 0..1 is allowed ('rgb(10%, 50%, 0)', None), ('rgb(255, 50%, 0%)', None), ('rgb(0, 0 0)', None), ('rgb(0, 0, 0deg)', None), ('rgb(0, 0, light)', None), ('rgb()', None), ('rgb(0)', None), ('rgb(0, 0)', None), ('rgb(0, 0, 0, 0)', None), ('rgb(0%)', None), ('rgb(0%, 0%)', None), ('rgb(0%, 0%, 0%, 0%)', None), ('rgb(0%, 0%, 0%, 0)', None), ('rgba(0, 0, 0, 0)', (0, 0, 0, 0)), ('rgba(203, 169, 135, 0.3)', (203 / 255., 169 / 255., 135 / 255., 0.3)), ('RGBA(255, 255, 255, 0)', (1, 1, 1, 0)), ('rgBA(0, 51, 255, 1)', (0, 0.2, 1, 1)), ('rgba(0, 51, 255, 1.1)', (0, 0.2, 1, 1)), ('rgba(0, 51, 255, 37)', (0, 0.2, 1, 1)), ('rgba(0, 51, 255, 0.42)', (0, 0.2, 1, 0.42)), ('rgba(0, 51, 255, 0)', (0, 0.2, 1, 0)), ('rgba(0, 51, 255, -0.1)', (0, 0.2, 1, 0)), ('rgba(0, 51, 255, -139)', (0, 0.2, 1, 0)), ('rgba(42%, 3%, 50%, 0.3)', (.42, .03, .5, 0.3)), ('RGBA(100%, 100%, 100%, 0)', (1, 1, 1, 0)), ('rgBA(0%, 20%, 100%, 1)', (0, 0.2, 1, 1)), ('rgba(0%, 20%, 100%, 1.1)', (0, 0.2, 1, 1)), ('rgba(0%, 20%, 100%, 37)', (0, 0.2, 1, 1)), ('rgba(0%, 20%, 100%, 0.42)', (0, 0.2, 1, 0.42)), ('rgba(0%, 20%, 100%, 0)', (0, 0.2, 1, 0)), ('rgba(0%, 20%, 100%, -0.1)', (0, 0.2, 1, 0)), ('rgba(0%, 20%, 100%, -139)', (0, 0.2, 1, 0)), ('rgba(255, 255, 255, 0%)', None), ('rgba(10%, 50%, 0, 1)', None), ('rgba(255, 50%, 0%, 1)', None), ('rgba(0, 0, 0 0)', None), ('rgba(0, 0, 0, 0deg)', None), ('rgba(0, 0, 0, light)', None), ('rgba()', None), ('rgba(0)', None), ('rgba(0, 0, 0)', None), ('rgba(0, 0, 0, 0, 0)', None), ('rgba(0%)', None), ('rgba(0%, 0%)', None), ('rgba(0%, 0%, 0%)', None), ('rgba(0%, 0%, 0%, 0%)', None), ('rgba(0%, 0%, 0%, 0%, 0%)', None), ('HSL(0, 0%, 0%)', (0, 0, 0, 1)), ('hsL(0, 100%, 50%)', (1, 0, 0, 1)), ('hsl(60, 100%, 37.5%)', (0.75, 0.75, 0, 1)), ('hsl(780, 100%, 37.5%)', (0.75, 0.75, 0, 1)), ('hsl(-300, 100%, 37.5%)', (0.75, 0.75, 0, 1)), ('hsl(300, 50%, 50%)', (0.75, 0.25, 0.75, 1)), ('hsl(10, 50%, 0)', None), ('hsl(50%, 50%, 0%)', None), ('hsl(0, 0% 0%)', None), ('hsl(30deg, 100%, 100%)', None), ('hsl(0, 0%, light)', None), ('hsl()', None), ('hsl(0)', None), ('hsl(0, 0%)', None), ('hsl(0, 0%, 0%, 0%)', None), ('HSLA(-300, 100%, 37.5%, 1)', (0.75, 0.75, 0, 1)), ('hsLA(-300, 100%, 37.5%, 12)', (0.75, 0.75, 0, 1)), ('hsla(-300, 100%, 37.5%, 0.2)', (0.75, 0.75, 0, .2)), ('hsla(-300, 100%, 37.5%, 0)', (0.75, 0.75, 0, 0)), ('hsla(-300, 100%, 37.5%, -3)', (0.75, 0.75, 0, 0)), ('hsla(10, 50%, 0, 1)', None), ('hsla(50%, 50%, 0%, 1)', None), ('hsla(0, 0% 0%, 1)', None), ('hsla(30deg, 100%, 100%, 1)', None), ('hsla(0, 0%, light, 1)', None), ('hsla()', None), ('hsla(0)', None), ('hsla(0, 0%)', None), ('hsla(0, 0%, 0%, 50%)', None), ('hsla(0, 0%, 0%, 1, 0%)', None), ('cmyk(0, 0, 0, 0)', None), ]: result = parse_color_string(css_source) if isinstance(result, tuple): for got, expected in zip(result, expected_result): # Compensate for floating point errors: self.assertLess(abs(got - expected), 1e-10) for i, attr in enumerate(['red', 'green', 'blue', 'alpha']): self.ae(getattr(result, attr), result[i]) else: self.ae(result, expected_result) def test_hsl(self): for hsl, expected_rgb in [ # http://en.wikipedia.org/wiki/HSL_and_HSV#Examples ((0, 0, 100), (1, 1, 1)), ((127, 0, 100), (1, 1, 1)), ((0, 0, 50), (0.5, 0.5, 0.5)), ((127, 0, 50), (0.5, 0.5, 0.5)), ((0, 0, 0), (0, 0, 0)), ((127, 0, 0), (0, 0, 0)), ((0, 100, 50), (1, 0, 0)), ((60, 100, 37.5), (0.75, 0.75, 0)), ((780, 100, 37.5), (0.75, 0.75, 0)), ((-300, 100, 37.5), (0.75, 0.75, 0)), ((120, 100, 25), (0, 0.5, 0)), ((180, 100, 75), (0.5, 1, 1)), ((240, 100, 75), (0.5, 0.5, 1)), ((300, 50, 50), (0.75, 0.25, 0.75)), ((61.8, 63.8, 39.3), (0.628, 0.643, 0.142)), ((251.1, 83.2, 51.1), (0.255, 0.104, 0.918)), ((134.9, 70.7, 39.6), (0.116, 0.675, 0.255)), ((49.5, 89.3, 49.7), (0.941, 0.785, 0.053)), ((283.7, 77.5, 54.2), (0.704, 0.187, 0.897)), ((14.3, 81.7, 62.4), (0.931, 0.463, 0.316)), ((56.9, 99.1, 76.5), (0.998, 0.974, 0.532)), ((162.4, 77.9, 44.7), (0.099, 0.795, 0.591)), ((248.3, 60.1, 37.3), (0.211, 0.149, 0.597)), ((240.5, 29, 60.7), (0.495, 0.493, 0.721)), ]: for got, expected in zip(hsl_to_rgb(*hsl), expected_rgb): # Compensate for floating point errors and Wikipedia’s rounding: self.assertLess(abs(got - expected), 0.001)
8,384
Python
.py
177
34.655367
85
0.352065
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,849
media3.py
kovidgoyal_calibre/src/tinycss/tests/media3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from tinycss.media3 import CSSMedia3Parser from tinycss.media3 import MediaQuery as MQ from tinycss.tests import BaseTest, jsonify def jsonify_expr(e): if e is None: return None return next(jsonify([e])) def jsonify_expressions(mqlist): for mq in mqlist: mq.expressions = tuple( (k, jsonify_expr(e)) for k, e in mq.expressions) return mqlist class TestFonts3(BaseTest): def test_media_queries(self): 'Test parsing of media queries from the CSS 3 media module' for css, media_query_list, expected_errors in [ # CSS 2.1 (simple media queries) ('@media {}', [MQ()], []), ('@media all {}', [MQ()], []), ('@media screen {}', [MQ('screen')], []), ('@media , screen {}', [MQ(), MQ('screen')], []), ('@media screen, {}', [MQ('screen'), MQ()], []), # Examples from the CSS 3 specs ('@media screen and (color) {}', [MQ('screen', (('color', None),))], []), ('@media all and (min-width:500px) {}', [ MQ('all', (('min-width', ('DIMENSION', 500)),))], []), ('@media (min-width:500px) {}', [ MQ('all', (('min-width', ('DIMENSION', 500)),))], []), ('@media (orientation: portrait) {}', [ MQ('all', (('orientation', ('IDENT', 'portrait')),))], []), ('@media screen and (color), projection and (color) {}', [ MQ('screen', (('color', None),)), MQ('projection', (('color', None),)),], []), ('@media not screen and (color) {}', [ MQ('screen', (('color', None),), True)], []), ('@media only screen and (color) {}', [ MQ('screen', (('color', None),))], []), ('@media aural and (device-aspect-ratio: 16/9) {}', [ MQ('aural', (('device-aspect-ratio', ('RATIO', (16, 9))),))], []), ('@media (resolution: 166dpi) {}', [ MQ('all', (('resolution', ('DIMENSION', 166)),))], []), ('@media (min-resolution: 166DPCM) {}', [ MQ('all', (('min-resolution', ('DIMENSION', 166)),))], []), # Malformed media queries ('@media (example, all,), speech {}', [MQ(negated=True), MQ('speech')], ['expected a :']), ('@media &test, screen {}', [MQ(negated=True), MQ('screen')], ['expected a media expression not a DELIM']), ]: stylesheet = CSSMedia3Parser().parse_stylesheet(css) self.assert_errors(stylesheet.errors, expected_errors) self.ae(len(stylesheet.rules), 1) rule = stylesheet.rules[0] self.ae(jsonify_expressions(rule.media), media_query_list)
2,983
Python
.py
55
41.527273
123
0.493484
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,850
decoding.py
kovidgoyal_calibre/src/tinycss/tests/decoding.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from tinycss.decoding import decode from tinycss.tests import BaseTest def params(css, encoding, use_bom=False, expect_error=False, **kwargs): """Nicer syntax to make a tuple.""" return css, encoding, use_bom, expect_error, kwargs class TestDecoding(BaseTest): def test_decoding(self): for (css, encoding, use_bom, expect_error, kwargs) in [ params('', 'utf8'), # default to utf8 params('�', 'utf8'), params('é', 'latin1'), # utf8 fails, fall back on ShiftJIS params('£', 'ShiftJIS', expect_error=True), params('£', 'ShiftJIS', protocol_encoding='Shift-JIS'), params('£', 'ShiftJIS', linking_encoding='Shift-JIS'), params('£', 'ShiftJIS', document_encoding='Shift-JIS'), params('£', 'ShiftJIS', protocol_encoding='utf8', document_encoding='ShiftJIS'), params('@charset "utf8"; £', 'ShiftJIS', expect_error=True), params('@charset "utf£8"; £', 'ShiftJIS', expect_error=True), params('@charset "unknown-encoding"; £', 'ShiftJIS', expect_error=True), params('@charset "utf8"; £', 'ShiftJIS', document_encoding='ShiftJIS'), params('£', 'ShiftJIS', linking_encoding='utf8', document_encoding='ShiftJIS'), params('@charset "utf-32"; �', 'utf-32-be'), params('@charset "Shift-JIS"; £', 'ShiftJIS'), params('@charset "ISO-8859-8"; £', 'ShiftJIS', expect_error=True), params('�', 'utf-16-le', expect_error=True), # no BOM params('�', 'utf-16-le', use_bom=True), params('�', 'utf-32-be', expect_error=True), params('�', 'utf-32-be', use_bom=True), params('�', 'utf-32-be', document_encoding='utf-32-be'), params('�', 'utf-32-be', linking_encoding='utf-32-be'), params('@charset "utf-32-le"; �', 'utf-32-be', use_bom=True, expect_error=True), # protocol_encoding takes precedence over @charset params('@charset "ISO-8859-8"; £', 'ShiftJIS', protocol_encoding='Shift-JIS'), params('@charset "unknown-encoding"; £', 'ShiftJIS', protocol_encoding='Shift-JIS'), params('@charset "Shift-JIS"; £', 'ShiftJIS', protocol_encoding='utf8'), # @charset takes precedence over document_encoding params('@charset "Shift-JIS"; £', 'ShiftJIS', document_encoding='ISO-8859-8'), # @charset takes precedence over linking_encoding params('@charset "Shift-JIS"; £', 'ShiftJIS', linking_encoding='ISO-8859-8'), # linking_encoding takes precedence over document_encoding params('£', 'ShiftJIS', linking_encoding='Shift-JIS', document_encoding='ISO-8859-8'), ]: if use_bom: source = '\ufeff' + css else: source = css css_bytes = source.encode(encoding) result, result_encoding = decode(css_bytes, **kwargs) if expect_error: self.assertNotEqual(result, css) else: self.ae(result, css)
3,474
Python
.py
65
41.015385
85
0.555262
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,851
css21.py
kovidgoyal_calibre/src/tinycss/tests/css21.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' import io import os import tempfile from tinycss.css21 import CSS21Parser from tinycss.tests import BaseTest from tinycss.tests.tokenizing import jsonify class CoreParser(CSS21Parser): """A parser that always accepts unparsed at-rules.""" def parse_at_rule(self, rule, stylesheet_rules, errors, context): return rule def parse_bytes(css_bytes, kwargs): return CSS21Parser().parse_stylesheet_bytes(css_bytes, **kwargs) def parse_bytesio_file(css_bytes, kwargs): css_file = io.BytesIO(css_bytes) return CSS21Parser().parse_stylesheet_file(css_file, **kwargs) def parse_filename(css_bytes, kwargs): css_file = tempfile.NamedTemporaryFile(delete=False) try: css_file.write(css_bytes) # Windows can not open the filename a second time while # it is still open for writing. css_file.close() return CSS21Parser().parse_stylesheet_file(css_file.name, **kwargs) finally: os.remove(css_file.name) class TestCSS21(BaseTest): def test_bytes(self): for (css_bytes, kwargs, expected_result, parse) in [ params + (parse,) for parse in [parse_bytes, parse_bytesio_file, parse_filename] for params in [ ('@import "é";'.encode('utf8'), {}, 'é'), ('@import "é";'.encode('utf16'), {}, 'é'), # with a BOM ('@import "é";'.encode('latin1'), {}, 'é'), ('@import "£";'.encode('Shift-JIS'), {}, '\x81\x92'), # latin1 mojibake ('@charset "Shift-JIS";@import "£";'.encode('Shift-JIS'), {}, '£'), (' @charset "Shift-JIS";@import "£";'.encode('Shift-JIS'), {}, '\x81\x92'), ('@import "£";'.encode('Shift-JIS'), {'document_encoding': 'Shift-JIS'}, '£'), ('@import "£";'.encode('Shift-JIS'), {'document_encoding': 'utf8'}, '\x81\x92'), ('@charset "utf8"; @import "£";'.encode('utf8'), {'document_encoding': 'latin1'}, '£'), # Mojibake yay! (' @charset "utf8"; @import "é";'.encode('utf8'), {'document_encoding': 'latin1'}, 'é'), ('@import "é";'.encode('utf8'), {'document_encoding': 'latin1'}, 'é'), ] ]: stylesheet = parse(css_bytes, kwargs) self.ae(stylesheet.rules[0].at_keyword, '@import') self.ae(stylesheet.rules[0].uri, expected_result) def test_at_rules(self): for (css_source, expected_rules, expected_errors) in [ (' /* hey */\n', 0, []), ('foo {}', 1, []), ('foo{} @lipsum{} bar{}', 2, ['unknown at-rule in stylesheet context: @lipsum']), ('@charset "ascii"; foo {}', 1, []), (' @charset "ascii"; foo {}', 1, ['mis-placed or malformed @charset rule']), ('@charset ascii; foo {}', 1, ['mis-placed or malformed @charset rule']), ('foo {} @charset "ascii";', 1, ['mis-placed or malformed @charset rule']), ]: # Pass 'encoding' to allow @charset stylesheet = CSS21Parser().parse_stylesheet(css_source, encoding='utf8') self.assert_errors(stylesheet.errors, expected_errors) self.ae(len(stylesheet.rules), expected_rules) def test_core_parser(self): for (css_source, expected_rules, expected_errors) in [ (' /* hey */\n', [], []), ('foo{} /* hey */\n@bar;@baz{}', [('foo', []), ('@bar', [], None), ('@baz', [], [])], []), ('@import "foo.css"/**/;', [ ('@import', [('STRING', 'foo.css')], None)], []), ('@import "foo.css"/**/', [ ('@import', [('STRING', 'foo.css')], None)], []), ('@import "foo.css', [ ('@import', [('STRING', 'foo.css')], None)], []), ('{}', [], ['empty selector']), ('a{b:4}', [('a', [('b', [('INTEGER', 4)])])], []), ('@page {\t b: 4; @margin}', [('@page', [], [ ('S', '\t '), ('IDENT', 'b'), (':', ':'), ('S', ' '), ('INTEGER', 4), (';', ';'), ('S', ' '), ('ATKEYWORD', '@margin'), ])], []), ('foo', [], ['no declaration block found']), ('foo @page {} bar {}', [('bar', [])], ['unexpected ATKEYWORD token in selector']), ('foo { content: "unclosed string;\n color:red; ; margin/**/\n: 2cm; }', [('foo', [('margin', [('DIMENSION', 2)])])], ['unexpected BAD_STRING token in property value']), ('foo { 4px; bar: 12% }', [('foo', [('bar', [('PERCENTAGE', 12)])])], ['expected a property name, got DIMENSION']), ('foo { bar! 3cm auto ; baz: 7px }', [('foo', [('baz', [('DIMENSION', 7)])])], ["expected ':', got DELIM"]), ('foo { bar ; baz: {("}"/* comment */) {0@fizz}} }', [('foo', [('baz', [('{', [ ('(', [('STRING', '}')]), ('S', ' '), ('{', [('INTEGER', 0), ('ATKEYWORD', '@fizz')]) ])])])], ["expected ':'"]), ('foo { bar: ; baz: not(z) }', [('foo', [('baz', [('FUNCTION', 'not', [('IDENT', 'z')])])])], ['expected a property value']), ('foo { bar: (]) ; baz: U+20 }', [('foo', [('baz', [('UNICODE-RANGE', 'U+20')])])], ['unmatched ] token in (']), ]: stylesheet = CoreParser().parse_stylesheet(css_source) self.assert_errors(stylesheet.errors, expected_errors) result = [ (rule.at_keyword, list(jsonify(rule.head)), list(jsonify(rule.body)) if rule.body is not None else None) if rule.at_keyword else (rule.selector.as_css(), [ (decl.name, list(jsonify(decl.value))) for decl in rule.declarations]) for rule in stylesheet.rules ] self.ae(result, expected_rules) def test_parse_style_attr(self): for (css_source, expected_declarations, expected_errors) in [ (' /* hey */\n', [], []), ('b:4', [('b', [('INTEGER', 4)])], []), ('{b:4}', [], ['expected a property name, got {']), ('b:4} c:3', [], ['unmatched } token in property value']), (' 4px; bar: 12% ', [('bar', [('PERCENTAGE', 12)])], ['expected a property name, got DIMENSION']), ('bar! 3cm auto ; baz: 7px', [('baz', [('DIMENSION', 7)])], ["expected ':', got DELIM"]), ('foo; bar ; baz: {("}"/* comment */) {0@fizz}}', [('baz', [('{', [ ('(', [('STRING', '}')]), ('S', ' '), ('{', [('INTEGER', 0), ('ATKEYWORD', '@fizz')]) ])])], ["expected ':'", "expected ':'"]), ('bar: ; baz: not(z)', [('baz', [('FUNCTION', 'not', [('IDENT', 'z')])])], ['expected a property value']), ('bar: (]) ; baz: U+20', [('baz', [('UNICODE-RANGE', 'U+20')])], ['unmatched ] token in (']), ]: declarations, errors = CSS21Parser().parse_style_attr(css_source) self.assert_errors(errors, expected_errors) result = [(decl.name, list(jsonify(decl.value))) for decl in declarations] self.ae(result, expected_declarations) def test_important(self): for (css_source, expected_declarations, expected_errors) in [ (' /* hey */\n', [], []), ('a:1; b:2', [('a', [('INTEGER', 1)], None), ('b', [('INTEGER', 2)], None)], []), ('a:1 important; b: important', [('a', [('INTEGER', 1), ('S', ' '), ('IDENT', 'important')], None), ('b', [('IDENT', 'important')], None)], []), ('a:1 !important; b:2', [('a', [('INTEGER', 1)], 'important'), ('b', [('INTEGER', 2)], None)], []), ('a:1!\t Im\\50 O\\RTant; b:2', [('a', [('INTEGER', 1)], 'important'), ('b', [('INTEGER', 2)], None)], []), ('a: !important; b:2', [('b', [('INTEGER', 2)], None)], ['expected a value before !important']), ]: declarations, errors = CSS21Parser().parse_style_attr(css_source) self.assert_errors(errors, expected_errors) result = [(decl.name, list(jsonify(decl.value)), decl.priority) for decl in declarations] self.ae(result, expected_declarations) def test_at_import(self): for (css_source, expected_rules, expected_errors) in [ (' /* hey */\n', [], []), ('@import "foo.css";', [('foo.css', ['all'])], []), ('@import url(foo.css);', [('foo.css', ['all'])], []), ('@import "foo.css" screen, print;', [('foo.css', ['screen', 'print'])], []), ('@charset "ascii"; @import "foo.css"; @import "bar.css";', [('foo.css', ['all']), ('bar.css', ['all'])], []), ('foo {} @import "foo.css";', [], ['@import rule not allowed after a ruleset']), ('@page {} @import "foo.css";', [], ['@import rule not allowed after an @page rule']), ('@import ;', [], ['expected URI or STRING for @import rule']), ('@import foo.css;', [], ['expected URI or STRING for @import rule, got IDENT']), ('@import "foo.css" {}', [], ["expected ';', got a block"]), ]: # Pass 'encoding' to allow @charset stylesheet = CSS21Parser().parse_stylesheet(css_source, encoding='utf8') self.assert_errors(stylesheet.errors, expected_errors) result = [ (rule.uri, rule.media) for rule in stylesheet.rules if rule.at_keyword == '@import' ] self.ae(result, expected_rules) def test_at_page(self): for (css, expected_result, expected_errors) in [ ('@page {}', (None, (0, 0), []), []), ('@page:first {}', ('first', (1, 0), []), []), ('@page :left{}', ('left', (0, 1), []), []), ('@page\t\n:right {}', ('right', (0, 1), []), []), ('@page :last {}', None, ['invalid @page selector']), ('@page : right {}', None, ['invalid @page selector']), ('@page table:left {}', None, ['invalid @page selector']), ('@page;', None, ['invalid @page rule: missing block']), ('@page { a:1; ; b: 2 }', (None, (0, 0), [('a', [('INTEGER', 1)]), ('b', [('INTEGER', 2)])]), []), ('@page { a:1; c: ; b: 2 }', (None, (0, 0), [('a', [('INTEGER', 1)]), ('b', [('INTEGER', 2)])]), ['expected a property value']), ('@page { a:1; @top-left {} b: 2 }', (None, (0, 0), [('a', [('INTEGER', 1)]), ('b', [('INTEGER', 2)])]), ['unknown at-rule in @page context: @top-left']), ('@page { a:1; @top-left {}; b: 2 }', (None, (0, 0), [('a', [('INTEGER', 1)]), ('b', [('INTEGER', 2)])]), ['unknown at-rule in @page context: @top-left']), ]: stylesheet = CSS21Parser().parse_stylesheet(css) self.assert_errors(stylesheet.errors, expected_errors) if expected_result is None: self.assertFalse(stylesheet.rules) else: self.ae(len(stylesheet.rules), 1) rule = stylesheet.rules[0] self.ae(rule.at_keyword, '@page') self.ae(rule.at_rules, []) # in CSS 2.1 result = ( rule.selector, rule.specificity, [(decl.name, list(jsonify(decl.value))) for decl in rule.declarations], ) self.ae(result, expected_result) def test_at_media(self): for (css_source, expected_rules, expected_errors) in [ (' /* hey */\n', [], []), ('@media {}', [(['all'], [])], []), ('@media all {}', [(['all'], [])], []), ('@media screen, print {}', [(['screen', 'print'], [])], []), ('@media all;', [], ['invalid @media rule: missing block']), ('@media 4 {}', [], ['expected a media type, got INTEGER']), ('@media , screen {}', [], ['expected a media type']), ('@media screen, {}', [], ['expected a media type']), ('@media screen print {}', [], ['expected a media type, got IDENT, IDENT']), ('@media all { @page { a: 1 } @media; @import; foo { a: 1 } }', [(['all'], [('foo', [('a', [('INTEGER', 1)])])])], ['@page rule not allowed in @media', '@media rule not allowed in @media', '@import rule not allowed in @media']), ]: stylesheet = CSS21Parser().parse_stylesheet(css_source) self.assert_errors(stylesheet.errors, expected_errors) for rule in stylesheet.rules: self.ae(rule.at_keyword, '@media') result = [ (rule.media, [ (sub_rule.selector.as_css(), [ (decl.name, list(jsonify(decl.value))) for decl in sub_rule.declarations]) for sub_rule in rule.rules ]) for rule in stylesheet.rules ] self.ae(result, expected_rules)
14,361
Python
.py
281
36.75089
92
0.442548
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,852
__init__.py
kovidgoyal_calibre/src/tinycss/tests/__init__.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' import unittest def jsonify(tokens): """Turn tokens into "JSON-compatible" data structures.""" for token in tokens: if token.type == 'FUNCTION': yield (token.type, token.function_name, list(jsonify(token.content))) elif token.is_container: yield token.type, list(jsonify(token.content)) else: yield token.type, token.value class BaseTest(unittest.TestCase): longMessage = True maxDiff = None ae = unittest.TestCase.assertEqual def assert_errors(self, errors, expected_errors): """Test not complete error messages but only substrings.""" self.ae(len(errors), len(expected_errors)) for error, expected in zip(errors, expected_errors): self.assertIn(expected, type(u'')(error)) def jsonify_declarations(self, rule): return [(decl.name, list(jsonify(decl.value))) for decl in rule.declarations]
1,100
Python
.py
27
33.185185
67
0.65381
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,853
fonts3.py
kovidgoyal_calibre/src/tinycss/tests/fonts3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from polyglot.builtins import iteritems from tinycss.fonts3 import CSSFonts3Parser, parse_font, parse_font_family, serialize_font from tinycss.tests import BaseTest class TestFonts3(BaseTest): def test_font_face(self): 'Test parsing of font face rules' for css, expected_declarations, expected_errors in [ ('@font-face {}', [], []), ('@font-face { font-family: Moose; src: url(font1.ttf) }', [('font-family', [('IDENT', 'Moose')]), ('src', [('URI', 'font1.ttf')])], []), ]: stylesheet = CSSFonts3Parser().parse_stylesheet(css) self.assert_errors(stylesheet.errors, expected_errors) self.ae(len(stylesheet.rules), 1) rule = stylesheet.rules[0] self.ae(self.jsonify_declarations(rule), expected_declarations) stylesheet = CSSFonts3Parser().parse_stylesheet('@font-face;') self.assert_errors(stylesheet.errors, ['missing block']) def test_parse_font_family(self): ' Test parsing of font-family values ' for raw, q in iteritems({ '"1as"': ['1as'], 'A B C, serif': ['A B C', 'serif'], r'Red\/Black': ['Red/Black'], 'A B': ['A B'], r'Ahem\!': ['Ahem!'], r'"Ahem!"': ['Ahem!'], '€42': ['€42'], r'Hawaii\ 5-0': ['Hawaii 5-0'], r'"X \"Y"': ['X "Y'], 'A B, C D, "E", serif': ['A B', 'C D', 'E', 'serif'], '': [], '"", a': ['a'], }): self.ae(q, parse_font_family(raw)) for single in ('serif', 'sans-serif', 'A B C'): self.ae([single], parse_font_family(single)) def test_parse_font(self): def t(raw, **kw): q = {('line' if k == 'height' else 'font') + '-' + k:v for k, v in iteritems(kw)} self.ae(q, parse_font(raw)) self.ae(q, parse_font(serialize_font(q))) t('caption', family=['sans-serif']) t('serif', family=['serif']) t('12pt/14pt sans-serif', size='12pt', height='14pt', family=['sans-serif']) t('80% sans-serif', size='80%', family=['sans-serif']) t('x-large/110% "new century schoolbook", serif', size='x-large', height='110%', family=['new century schoolbook', 'serif']) t('bold italic large Palatino, serif', weight='bold', style='italic', size='large', family=['Palatino', 'serif']) t('normal small-caps 120%/120% fantasy', style='normal', variant='small-caps', size='120%', height='120%', family=['fantasy']) t('condensed oblique 12pt Helvetica Neue, serif', stretch='condensed', style='oblique', size='12pt', family=['Helvetica Neue', 'serif']) t('300 italic 1.3em/1.7em FB Armada, sans-serif', weight='300', style='italic', size='1.3em', height='1.7em', family=['FB Armada', 'sans-serif'])
3,066
Python
.py
55
44.945455
153
0.551218
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,854
page3.py
kovidgoyal_calibre/src/tinycss/tests/page3.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from tinycss.page3 import CSSPage3Parser from tinycss.tests import BaseTest class TestPage3(BaseTest): def test_selectors(self): for css, expected_selector, expected_specificity, expected_errors in [ ('@page {}', (None, None), (0, 0, 0), []), ('@page :first {}', (None, 'first'), (0, 1, 0), []), ('@page:left{}', (None, 'left'), (0, 0, 1), []), ('@page :right {}', (None, 'right'), (0, 0, 1), []), ('@page :blank{}', (None, 'blank'), (0, 1, 0), []), ('@page :last {}', None, None, ['invalid @page selector']), ('@page : first {}', None, None, ['invalid @page selector']), ('@page foo:first {}', ('foo', 'first'), (1, 1, 0), []), ('@page bar :left {}', ('bar', 'left'), (1, 0, 1), []), (r'@page \26:right {}', ('&', 'right'), (1, 0, 1), []), ('@page foo {}', ('foo', None), (1, 0, 0), []), (r'@page \26 {}', ('&', None), (1, 0, 0), []), ('@page foo fist {}', None, None, ['invalid @page selector']), ('@page foo, bar {}', None, None, ['invalid @page selector']), ('@page foo&first {}', None, None, ['invalid @page selector']), ]: stylesheet = CSSPage3Parser().parse_stylesheet(css) self.assert_errors(stylesheet.errors, expected_errors) if stylesheet.rules: self.ae(len(stylesheet.rules), 1) rule = stylesheet.rules[0] self.ae(rule.at_keyword, '@page') selector = rule.selector self.ae(rule.specificity, expected_specificity) else: selector = None self.ae(selector, expected_selector) def test_content(self): for css, expected_declarations, expected_rules, expected_errors in [ ('@page {}', [], [], []), ('@page { foo: 4; bar: z }', [('foo', [('INTEGER', 4)]), ('bar', [('IDENT', 'z')])], [], []), ('''@page { foo: 4; @top-center { content: "Awesome Title" } @bottom-left { content: counter(page) } bar: z }''', [('foo', [('INTEGER', 4)]), ('bar', [('IDENT', 'z')])], [('@top-center', [('content', [('STRING', 'Awesome Title')])]), ('@bottom-left', [('content', [ ('FUNCTION', 'counter', [('IDENT', 'page')])])])], []), ('''@page { foo: 4; @bottom-top { content: counter(page) } bar: z }''', [('foo', [('INTEGER', 4)]), ('bar', [('IDENT', 'z')])], [], ['unknown at-rule in @page context: @bottom-top']), ('@page{} @top-right{}', [], [], [ '@top-right rule not allowed in stylesheet']), ('@page{ @top-right 4 {} }', [], [], [ 'unexpected INTEGER token in @top-right rule header']), # Not much error recovery tests here. This should be covered in test_css21 ]: stylesheet = CSSPage3Parser().parse_stylesheet(css) self.assert_errors(stylesheet.errors, expected_errors) self.ae(len(stylesheet.rules), 1) rule = stylesheet.rules[0] self.ae(rule.at_keyword, '@page') self.ae(self.jsonify_declarations(rule), expected_declarations) rules = [(margin_rule.at_keyword, self.jsonify_declarations(margin_rule)) for margin_rule in rule.at_rules] self.ae(rules, expected_rules)
3,818
Python
.py
73
38.671233
86
0.469579
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,855
main.py
kovidgoyal_calibre/src/tinycss/tests/main.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' import argparse import unittest def find_tests(): from calibre.utils.run_tests import find_tests_in_package return find_tests_in_package('tinycss.tests') def run_tests(find_tests=find_tests, for_build=False): if not for_build: parser = argparse.ArgumentParser() parser.add_argument('name', nargs='?', default=None, help='The name of the test to run') args = parser.parse_args() if not for_build and args.name and args.name.startswith('.'): tests = find_tests() q = args.name[1:] if not q.startswith('test_'): q = 'test_' + q ans = None try: for suite in tests: for test in suite._tests: if test.__class__.__name__ == 'ModuleImportFailure': raise Exception('Failed to import a test module: %s' % test) for s in test: if s._testMethodName == q: ans = s raise StopIteration() except StopIteration: pass if ans is None: print('No test named %s found' % args.name) raise SystemExit(1) tests = ans else: tests = unittest.defaultTestLoader.loadTestsFromName(args.name) if not for_build and args.name else find_tests() r = unittest.TextTestRunner if for_build: r = r(verbosity=0, buffer=True, failfast=True) else: r = r(verbosity=4) result = r.run(tests) if for_build and result.errors or result.failures: raise SystemExit(1) if __name__ == '__main__': run_tests()
1,810
Python
.py
48
27.9375
120
0.571021
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,856
tokenizing.py
kovidgoyal_calibre/src/tinycss/tests/tokenizing.py
#!/usr/bin/env python # vim:fileencoding=utf-8 __license__ = 'GPL v3' __copyright__ = '2014, Kovid Goyal <kovid at kovidgoyal.net>' from tinycss.tests import BaseTest, jsonify from tinycss.tokenizer import c_tokenize_flat, python_tokenize_flat, regroup if c_tokenize_flat is None: tokenizers = (python_tokenize_flat,) else: tokenizers = (python_tokenize_flat, c_tokenize_flat) def token_api(self, tokenize): for css_source in [ '(8, foo, [z])', '[8, foo, (z)]', '{8, foo, [z]}', 'func(8, foo, [z])' ]: tokens = list(regroup(tokenize(css_source))) self.ae(len(tokens), 1) self.ae(len(tokens[0].content), 7) def token_serialize_css(self, tokenize): for tokenize in tokenizers: for css_source in [ r'''p[example="\ foo(int x) {\ this.x = x;\ }\ "]''', '"Lorem\\26Ipsum\ndolor" sit', '/* Lorem\nipsum */\fa {\n color: red;\tcontent: "dolor\\\fsit" }', 'not([[lorem]]{ipsum (42)})', 'a[b{d]e}', 'a[b{"d', ]: for _regroup in (regroup, lambda x: x): tokens = _regroup(tokenize(css_source, ignore_comments=False)) result = ''.join(token.as_css() for token in tokens) self.ae(result, css_source) def comments(self, tokenize): for ignore_comments, expected_tokens in [ (False, [ ('COMMENT', '/* lorem */'), ('S', ' '), ('IDENT', 'ipsum'), ('[', [ ('IDENT', 'dolor'), ('COMMENT', '/* sit */'), ]), ('BAD_COMMENT', '/* amet') ]), (True, [ ('S', ' '), ('IDENT', 'ipsum'), ('[', [ ('IDENT', 'dolor'), ]), ]), ]: css_source = '/* lorem */ ipsum[dolor/* sit */]/* amet' tokens = regroup(tokenize(css_source, ignore_comments)) result = list(jsonify(tokens)) self.ae(result, expected_tokens) def token_grouping(self, tokenize): for css_source, expected_tokens in [ ('', []), (r'Lorem\26 "i\psum"4px', [ ('IDENT', 'Lorem&'), ('STRING', 'ipsum'), ('DIMENSION', 4)]), ('not([[lorem]]{ipsum (42)})', [ ('FUNCTION', 'not', [ ('[', [ ('[', [ ('IDENT', 'lorem'), ]), ]), ('{', [ ('IDENT', 'ipsum'), ('S', ' '), ('(', [ ('INTEGER', 42), ]) ]) ])]), # Close everything at EOF, no error ('a[b{"d', [ ('IDENT', 'a'), ('[', [ ('IDENT', 'b'), ('{', [ ('STRING', 'd'), ]), ]), ]), # Any remaining ), ] or } token is a nesting error ('a[b{d]e}', [ ('IDENT', 'a'), ('[', [ ('IDENT', 'b'), ('{', [ ('IDENT', 'd'), (']', ']'), # The error is visible here ('IDENT', 'e'), ]), ]), ]), # ref: ('a[b{d}e]', [ ('IDENT', 'a'), ('[', [ ('IDENT', 'b'), ('{', [ ('IDENT', 'd'), ]), ('IDENT', 'e'), ]), ]), ]: tokens = regroup(tokenize(css_source, ignore_comments=False)) result = list(jsonify(tokens)) self.ae(result, expected_tokens) def positions(self, tokenize): css = '/* Lorem\nipsum */\fa {\n color: red;\tcontent: "dolor\\\fsit" }' tokens = tokenize(css, ignore_comments=False) result = [(token.type, token.line, token.column) for token in tokens] self.ae(result, [ ('COMMENT', 1, 1), ('S', 2, 9), ('IDENT', 3, 1), ('S', 3, 2), ('{', 3, 3), ('S', 3, 4), ('IDENT', 4, 5), (':', 4, 10), ('S', 4, 11), ('IDENT', 4, 12), (';', 4, 15), ('S', 4, 16), ('IDENT', 4, 17), (':', 4, 24), ('S', 4, 25), ('STRING', 4, 26), ('S', 5, 5), ('}', 5, 6)]) def tokens(self, tokenize): for css_source, expected_tokens in [ ('', []), ('red -->', [('IDENT', 'red'), ('S', ' '), ('CDC', '-->')]), # Longest match rule: no CDC ('red-->', [('IDENT', 'red--'), ('DELIM', '>')]), (r'''p[example="\ foo(int x) {\ this.x = x;\ }\ "]''', [ ('IDENT', 'p'), ('[', '['), ('IDENT', 'example'), ('DELIM', '='), ('STRING', 'foo(int x) { this.x = x;}'), (']', ']')]), # Numbers are parsed ('42 .5 -4pX 1.25em 30%', [('INTEGER', 42), ('S', ' '), ('NUMBER', .5), ('S', ' '), # units are normalized to lower-case: ('DIMENSION', -4, 'px'), ('S', ' '), ('DIMENSION', 1.25, 'em'), ('S', ' '), ('PERCENTAGE', 30, '%')]), # URLs are extracted ('url(foo.png)', [('URI', 'foo.png')]), ('url("foo.png")', [('URI', 'foo.png')]), # Escaping (r'/* Comment with a \ backslash */', [('COMMENT', r'/* Comment with a \ backslash */')]), # Unchanged # backslash followed by a newline in a string: ignored ('"Lorem\\\nIpsum"', [('STRING', 'LoremIpsum')]), # backslash followed by a newline outside a string: stands for itself ('Lorem\\\nIpsum', [ ('IDENT', 'Lorem'), ('DELIM', '\\'), ('S', '\n'), ('IDENT', 'Ipsum')]), # Cancel the meaning of special characters (r'"Lore\m Ipsum"', [('STRING', 'Lorem Ipsum')]), # or not special (r'"Lorem \49psum"', [('STRING', 'Lorem Ipsum')]), (r'"Lorem \49 psum"', [('STRING', 'Lorem Ipsum')]), (r'"Lorem\"Ipsum"', [('STRING', 'Lorem"Ipsum')]), (r'"Lorem\\Ipsum"', [('STRING', r'Lorem\Ipsum')]), (r'"Lorem\5c Ipsum"', [('STRING', r'Lorem\Ipsum')]), (r'Lorem\+Ipsum', [('IDENT', 'Lorem+Ipsum')]), (r'Lorem+Ipsum', [('IDENT', 'Lorem'), ('DELIM', '+'), ('IDENT', 'Ipsum')]), (r'url(foo\).png)', [('URI', 'foo).png')]), # Unicode and backslash escaping ('\\26 B', [('IDENT', '&B')]), ('\\&B', [('IDENT', '&B')]), ('@\\26\tB', [('ATKEYWORD', '@&B')]), ('@\\&B', [('ATKEYWORD', '@&B')]), ('#\\26\nB', [('HASH', '#&B')]), ('#\\&B', [('HASH', '#&B')]), ('\\26\r\nB(', [('FUNCTION', '&B(')]), ('\\&B(', [('FUNCTION', '&B(')]), (r'12.5\000026B', [('DIMENSION', 12.5, '&b')]), (r'12.5\0000263B', [('DIMENSION', 12.5, '&3b')]), # max 6 digits (r'12.5\&B', [('DIMENSION', 12.5, '&b')]), (r'"\26 B"', [('STRING', '&B')]), (r"'\000026B'", [('STRING', '&B')]), (r'"\&B"', [('STRING', '&B')]), (r'url("\26 B")', [('URI', '&B')]), (r'url(\26 B)', [('URI', '&B')]), (r'url("\&B")', [('URI', '&B')]), (r'url(\&B)', [('URI', '&B')]), (r'Lorem\110000Ipsum', [('IDENT', 'Lorem\uFFFDIpsum')]), # Bad strings # String ends at EOF without closing: no error, parsed ('"Lorem\\26Ipsum', [('STRING', 'Lorem&Ipsum')]), # Unescaped newline: ends the string, error, unparsed ('"Lorem\\26Ipsum\n', [ ('BAD_STRING', r'"Lorem\26Ipsum'), ('S', '\n')]), # Tokenization restarts after the newline, so the second " starts # a new string (which ends at EOF without errors, as above.) ('"Lorem\\26Ipsum\ndolor" sit', [ ('BAD_STRING', r'"Lorem\26Ipsum'), ('S', '\n'), ('IDENT', 'dolor'), ('STRING', ' sit')]), ]: sources = [css_source] for css_source in sources: tokens = tokenize(css_source, ignore_comments=False) result = [ (token.type, token.value) + ( () if token.unit is None else (token.unit,)) for token in tokens ] self.ae(result, expected_tokens) class TestTokenizer(BaseTest): def run_test(self, func): for tokenize in tokenizers: func(self, tokenize) def test_token_api(self): self.run_test(token_api) def test_token_serialize_css(self): self.run_test(token_serialize_css) def test_comments(self): self.run_test(comments) def test_token_grouping(self): self.run_test(token_grouping) def test_positions(self): """Test the reported line/column position of each token.""" self.run_test(positions) def test_tokens(self): self.run_test(tokens)
8,863
Python
.py
234
27.474359
83
0.430318
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,857
planet_python.recipe
kovidgoyal_calibre/recipes/planet_python.recipe
from calibre.web.feeds.news import AutomaticNewsRecipe class BasicUserRecipe(AutomaticNewsRecipe): title = u'Planet Python' language = 'en' __author__ = 'Jelle van der Waa' oldest_article = 10 max_articles_per_feed = 100 feeds = [(u'Planet Python', u'http://planetpython.org/rss20.xml')]
314
Python
.py
8
35
70
0.713816
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,858
default_tweaks.py
kovidgoyal_calibre/resources/default_tweaks.py
#!/usr/bin/env python # vim:fileencoding=UTF-8:ts=4:sw=4:sta:et:sts=4:ai # License: GPLv3 Copyright: 2010, Kovid Goyal <kovid at kovidgoyal.net> # Contains various tweaks that affect calibre behavior. Only edit this file if # you know what you are doing. If you delete this file, it will be recreated from # defaults. #: Auto increment series index # The algorithm used to assign a book added to an existing series a series number. # New series numbers assigned using this tweak are always integer values, except # if a constant non-integer is specified. # Possible values are: # next - First available integer larger than the largest existing number # first_free - First available integer larger than 0 # next_free - First available integer larger than the smallest existing number # last_free - First available integer smaller than the largest existing number. Return largest existing + 1 if no free number is found # const - Assign the number 1 always # no_change - Do not change the series index # a number - Assign that number always. The number is not in quotes. Note that 0.0 can be used here. # Examples: # series_index_auto_increment = 'next' # series_index_auto_increment = 'next_free' # series_index_auto_increment = 16.5 # # Set the use_series_auto_increment_tweak_when_importing tweak to True to # use the above values when importing/adding books. If this tweak is set to # False (the default) then the series number will be set to 1 if it is not # explicitly set during the import. If set to True, then the # series index will be set according to the series_index_auto_increment setting. # Note that the use_series_auto_increment_tweak_when_importing tweak is used # only when a value is not provided during import. If the importing regular # expression produces a value for series_index, or if you are reading metadata # from books and the import plugin produces a value, then that value will # be used irrespective of the setting of the tweak. series_index_auto_increment = 'next' use_series_auto_increment_tweak_when_importing = False #: Add separator after completing an author name # Set this if the completion separator should be appended to the end of the # completed text to automatically begin a new completion operation for authors. # It can be either True or False authors_completer_append_separator = False #: Author sort name algorithm # The algorithm used to copy author to author_sort. # Possible values are: # invert: use "fn ln" -> "ln, fn" # copy : copy author to author_sort without modification # comma : use 'copy' if there is a ',' in the name, otherwise use 'invert' # nocomma : "fn ln" -> "ln fn" (without the comma) # When this tweak is changed, the author_sort values stored with each author # must be recomputed by right-clicking on an author in the left-hand tags # panel, selecting 'Manage authors', and pressing # 'Recalculate all author sort values'. # # The author_name_suffixes are words that are ignored when they occur at the # end of an author name. The case of the suffix is ignored and trailing # periods are automatically handled. # # The same is true for author_name_prefixes. # # The author_name_copywords are a set of words which, if they occur in an # author name, cause the automatically generated author sort string to be # identical to the author's name. This means that the sort for a string like # "Acme Inc." will be "Acme Inc." instead of "Inc., Acme". # # If author_use_surname_prefixes is enabled, any of the words in # author_surname_prefixes will be treated as a prefix to the surname, if they # occur before the surname. So for example, "John von Neumann" would be sorted # as "von Neumann, John" and not "Neumann, John von". author_sort_copy_method = 'comma' author_name_suffixes = ('Jr', 'Sr', 'Inc', 'Ph.D', 'Phd', 'MD', 'M.D', 'I', 'II', 'III', 'IV', 'Junior', 'Senior') author_name_prefixes = ('Mr', 'Mrs', 'Ms', 'Dr', 'Prof') author_name_copywords = ( 'Agency', 'Corporation', 'Company', 'Co.', 'Council', 'Committee', 'Inc.', 'Institute', 'National', 'Society', 'Club', 'Team', 'Software', 'Games', 'Entertainment', 'Media', 'Studios', ) author_use_surname_prefixes = False author_surname_prefixes = ('da', 'de', 'di', 'la', 'le', 'van', 'von') #: Splitting multiple author names # By default, calibre splits a string containing multiple author names on # ampersands and the words "and" and "with". You can customize the splitting # by changing the regular expression below. Strings are split on whatever the # specified regular expression matches, in addition to ampersands. # Default: r'(?i),?\s+(and|with)\s+' authors_split_regex = r'(?i),?\s+(and|with)\s+' #: Use author sort in Tag browser # Set which author field to display in the Tag browser (the list of authors, # series, publishers etc on the left hand side). The choices are author and # author_sort. This tweak affects only what is displayed under the authors # category in the Tag browser and Content server. Please note that if you set this # to author_sort, it is very possible to see duplicate names in the list because # although it is guaranteed that author names are unique, there is no such # guarantee for author_sort values. Showing duplicates won't break anything, but # it could lead to some confusion. When using 'author_sort', the tooltip will # show the author's name. # Examples: # categories_use_field_for_author_name = 'author' # categories_use_field_for_author_name = 'author_sort' categories_use_field_for_author_name = 'author' #: Control partitioning of Tag browser # When partitioning the Tag browser, the format of the subcategory label is # controlled by a template: categories_collapsed_name_template if sorting by # name, categories_collapsed_rating_template if sorting by average rating, and # categories_collapsed_popularity_template if sorting by popularity. There are # two variables available to the template: first and last. The variable 'first' # is the initial item in the subcategory, and the variable 'last' is the final # item in the subcategory. Both variables are 'objects'; they each have multiple # values that are obtained by using a suffix. For example, first.name for an # author category will be the name of the author. The sub-values available are: # name: the printable name of the item # count: the number of books that references this item # avg_rating: the average rating of all the books referencing this item # sort: the sort value. For authors, this is the author_sort for that author # category: the category (e.g., authors, series) that the item is in. # Note that the "r'" in front of the { is necessary if there are backslashes # (\ characters) in the template. It doesn't hurt anything to leave it there # even if there aren't any backslashes. categories_collapsed_name_template = r'{first.sort:shorten(4,,0)} - {last.sort:shorten(4,,0)}' categories_collapsed_rating_template = r'{first.avg_rating:4.2f:ifempty(0)} - {last.avg_rating:4.2f:ifempty(0)}' categories_collapsed_popularity_template = r'{first.count:d} - {last.count:d}' #: Specify columns to sort the booklist by on startup # Provide a set of columns to be sorted on when calibre starts. # The argument is None if saved sort history is to be used # otherwise it is a list of column,order pairs. Column is the # lookup/search name, found using the tooltip for the column # Order is 0 for ascending, 1 for descending. # For example, set it to [('authors',0),('title',0)] to sort by # title within authors. sort_columns_at_startup = None #: Control how dates are displayed # Format to be used for publication date and the timestamp (date). # A string controlling how the publication date is displayed in the GUI # d the day as number without a leading zero (1 to 31) # dd the day as number with a leading zero (01 to 31) # ddd the abbreviated localized day name (e.g. 'Mon' to 'Sun'). # dddd the long localized day name (e.g. 'Monday' to 'Sunday'). # M the month as number without a leading zero (1-12) # MM the month as number with a leading zero (01-12) # MMM the abbreviated localized month name (e.g. 'Jan' to 'Dec'). # MMMM the long localized month name (e.g. 'January' to 'December'). # yy the year as two digit number (00-99) # yyyy the year as four digit number # h the hours without a leading 0 (0 to 11 or 0 to 23, depending on am/pm) ' # hh the hours with a leading 0 (00 to 11 or 00 to 23, depending on am/pm) ' # m the minutes without a leading 0 (0 to 59) ' # mm the minutes with a leading 0 (00 to 59) ' # s the seconds without a leading 0 (0 to 59) ' # ss the seconds with a leading 0 (00 to 59) ' # ap use a 12-hour clock instead of a 24-hour clock, with "ap" replaced by the localized string for am or pm # AP use a 12-hour clock instead of a 24-hour clock, with "AP" replaced by the localized string for AM or PM # iso the date with time and timezone. Must be the only format present # For example, given the date of 9 Jan 2010, the following formats show # MMM yyyy ==> Jan 2010 yyyy ==> 2010 dd MMM yyyy ==> 09 Jan 2010 # MM/yyyy ==> 01/2010 d/M/yy ==> 9/1/10 yy ==> 10 # # publication default if not set: MMM yyyy # timestamp default if not set: dd MMM yyyy # last_modified_display_format if not set: dd MMM yyyy gui_pubdate_display_format = 'MMM yyyy' gui_timestamp_display_format = 'dd MMM yyyy' gui_last_modified_display_format = 'dd MMM yyyy' #: Control sorting of titles and series in the library display # Control title and series sorting in the library view. If set to # 'library_order', the title sort field will be used instead of the title. # Unless you have manually edited the title sort field, leading articles such as # The and A will be ignored. If set to 'strictly_alphabetic', the titles will be # sorted as-is (sort by title instead of title sort). For example, with # library_order, The Client will sort under 'C'. With strictly_alphabetic, the # book will sort under 'T'. # This flag affects calibre's library display. It has no effect on devices. In # addition, titles for books added before changing the flag will retain their # order until the title is edited. Editing a title and hitting Enter # without changing anything is sufficient to change the sort. Or you can use # the 'Update title sort' action in the Bulk metadata edit dialog to update # it for many books at once. title_series_sorting = 'library_order' #: Control formatting of title and series when used in templates # Control how title and series names are formatted when saving to disk/sending # to device. The behavior depends on the field being processed. If processing # title, then if this tweak is set to 'library_order', the title will be # replaced with title_sort. If it is set to 'strictly_alphabetic', then the # title will not be changed. If processing series, then if set to # 'library_order', articles such as 'The' and 'An' will be moved to the end. If # set to 'strictly_alphabetic', the series will be sent without change. # For example, if the tweak is set to library_order, "The Lord of the Rings" # will become "Lord of the Rings, The". If the tweak is set to # strictly_alphabetic, it would remain "The Lord of the Rings". Note that the # formatter function raw_field will return the base value for title and # series regardless of the setting of this tweak. save_template_title_series_sorting = 'library_order' #: Set the list of words considered to be "articles" for sort strings # Set the list of words that are to be considered 'articles' when computing the # title sort strings. The articles differ by language. By default, calibre uses # a combination of articles from English and whatever language the calibre user # interface is set to. In addition, in some contexts where the book language is # available, the language of the book is used. You can change the list of # articles for a given language or add a new language by editing # per_language_title_sort_articles. To tell calibre to use a language other # than the user interface language, set, default_language_for_title_sort. For # example, to use German, set it to 'deu'. A value of None means the user # interface language is used. The setting title_sort_articles is ignored # (present only for legacy reasons). per_language_title_sort_articles = { # English 'eng' : (r'A\s+', r'The\s+', r'An\s+'), # Esperanto 'epo': (r'La\s+', r"L'", 'L´'), # Spanish 'spa' : (r'El\s+', r'La\s+', r'Lo\s+', r'Los\s+', r'Las\s+', r'Un\s+', r'Una\s+', r'Unos\s+', r'Unas\s+'), # French 'fra' : (r'Le\s+', r'La\s+', r"L'", u'L´', u'L’', r'Les\s+', r'Un\s+', r'Une\s+', r'Des\s+', r'De\s+La\s+', r'De\s+', r"D'", r'D´', r'D’'), # Polish 'pol': (), # Italian 'ita': ('Lo\\s+', 'Il\\s+', "L'", 'L´', 'La\\s+', 'Gli\\s+', 'I\\s+', 'Le\\s+', 'Uno\\s+', 'Un\\s+', 'Una\\s+', "Un'", 'Un´', 'Dei\\s+', 'Degli\\s+', 'Delle\\s+', 'Del\\s+', 'Della\\s+', 'Dello\\s+', "Dell'", 'Dell´'), # Portuguese 'por' : (r'A\s+', r'O\s+', r'Os\s+', r'As\s+', r'Um\s+', r'Uns\s+', r'Uma\s+', r'Umas\s+', ), # Romanian 'ron' : (r'Un\s+', r'O\s+', r'Nişte\s+', ), # German 'deu' : (r'Der\s+', r'Die\s+', r'Das\s+', r'Den\s+', r'Ein\s+', r'Eine\s+', r'Einen\s+', r'Dem\s+', r'Des\s+', r'Einem\s+', r'Eines\s+'), # Dutch 'nld' : (r'De\s+', r'Het\s+', r'Een\s+', r"'n\s+", r"'s\s+", r'Ene\s+', r'Ener\s+', r'Enes\s+', r'Den\s+', r'Der\s+', r'Des\s+', r"'t\s+"), # Swedish 'swe' : (r'En\s+', r'Ett\s+', r'Det\s+', r'Den\s+', r'De\s+', ), # Turkish 'tur' : (r'Bir\s+', ), # Afrikaans 'afr' : (r"'n\s+", r'Die\s+', ), # Greek 'ell' : (r'O\s+', r'I\s+', r'To\s+', r'Ta\s+', r'Tus\s+', r'Tis\s+', r"'Enas\s+", r"'Mia\s+", r"'Ena\s+", r"'Enan\s+", ), # Hungarian 'hun' : (r'A\s+', r'Az\s+', r'Egy\s+',), } default_language_for_title_sort = None title_sort_articles=r'^(A|The|An)\s+' #: Specify a folder calibre should connect to at startup # Specify a folder that calibre should connect to at startup using # connect_to_folder. This must be a full path to the folder. If the folder does # not exist when calibre starts, it is ignored. # Example for Windows: # auto_connect_to_folder = 'C:/Users/someone/Desktop/testlib' # Example for other operating systems: # auto_connect_to_folder = '/home/dropbox/My Dropbox/someone/library' auto_connect_to_folder = '' #: Specify renaming rules for SONY collections # Specify renaming rules for SONY collections. This tweak is only applicable if # metadata management is set to automatic. Collections on SONYs are named # depending upon whether the field is standard or custom. A collection derived # from a standard field is named for the value in that field. # # For example, if the standard 'series' column contains the value 'Darkover', then the # collection name is 'Darkover'. A collection derived from a custom field will # have the name of the field added to the value. For example, if a custom series # column named 'My Series' contains the name 'Darkover', then the collection # will by default be named 'Darkover (My Series)'. For purposes of this # documentation, 'Darkover' is called the value and 'My Series' is called the # category. If two books have fields that generate the same collection name, # then both books will be in that collection. # # This set of tweaks lets you specify for a standard or custom field how # the collections are to be named. You can use it to add a description to a # standard field, for example 'Foo (Tag)' instead of the 'Foo'. You can also use # it to force multiple fields to end up in the same collection. # # For example, you could force the values in 'series', '#my_series_1', and # '#my_series_2' to appear in collections named 'some_value (Series)', thereby # merging all of the fields into one set of collections. # # There are two related tweaks. The first determines the category name to use # for a metadata field. The second is a template, used to determines how the # value and category are combined to create the collection name. # The syntax of the first tweak, sony_collection_renaming_rules, is: # {'field_lookup_name':'category_name_to_use', 'lookup_name':'name', ...} # # The second tweak, sony_collection_name_template, is a template. It uses the # same template language as plugboards and save templates. This tweak controls # how the value and category are combined together to make the collection name. # The only two fields available are {category} and {value}. The {value} field is # never empty. The {category} field can be empty. The default is to put the # value first, then the category enclosed in parentheses, it isn't empty: # '{value} {category:|(|)}' # # Examples: The first three examples assume that the second tweak # has not been changed. # # 1) I want three series columns to be merged into one set of collections. The # column lookup names are 'series', '#series_1' and '#series_2'. I want nothing # in the parenthesis. The value to use in the tweak value would be: # sony_collection_renaming_rules={'series':'', '#series_1':'', '#series_2':''} # # 2) I want the word '(Series)' to appear on collections made from series, and # the word '(Tag)' to appear on collections made from tags. Use: # sony_collection_renaming_rules={'series':'Series', 'tags':'Tag'} # # 3) I want 'series' and '#myseries' to be merged, and for the collection name # to have '(Series)' appended. The renaming rule is: # sony_collection_renaming_rules={'series':'Series', '#myseries':'Series'} # # 4) Same as example 2, but instead of having the category name in parentheses # and appended to the value, I want it prepended and separated by a colon, such # as in Series: Darkover. I must change the template used to format the category name # # The resulting two tweaks are: # sony_collection_renaming_rules={'series':'Series', 'tags':'Tag'} # sony_collection_name_template='{category:||: }{value}' sony_collection_renaming_rules={} sony_collection_name_template='{value}{category:| (|)}' #: Specify how SONY collections are sorted # Specify how SONY collections are sorted. This tweak is only applicable if # metadata management is set to automatic. You can indicate which metadata is to # be used to sort on a collection-by-collection basis. The format of the tweak # is a list of metadata fields from which collections are made, followed by the # name of the metadata field containing the sort value. # Example: The following indicates that collections built from pubdate and tags # are to be sorted by the value in the custom column '#mydate', that collections # built from 'series' are to be sorted by 'series_index', and that all other # collections are to be sorted by title. If a collection metadata field is not # named, then if it is a series- based collection it is sorted by series order, # otherwise it is sorted by title order. # [(['pubdate', 'tags'],'#mydate'), (['series'],'series_index'), (['*'], 'title')] # Note that the bracketing and parentheses are required. The syntax is # [ ( [list of fields], sort field ) , ( [ list of fields ] , sort field ) ] # Default: empty (no rules), so no collection attributes are named. sony_collection_sorting_rules = [] #: Control how tags are applied when copying books to another library # Set this to True to ensure that tags in 'Tags to add when adding # a book' are added when copying books to another library add_new_book_tags_when_importing_books = False #: Set the maximum number of sort 'levels' # Set the maximum number of sort 'levels' that calibre will use to resort the # library after certain operations such as searches or device insertion. Each # sort level adds a performance penalty. If the database is large (thousands of # books) the penalty might be noticeable. If you are not concerned about multi- # level sorts, and if you are seeing a slowdown, reduce the value of this tweak. maximum_resort_levels = 5 #: Choose whether dates are sorted using visible fields # Date values contain both a date and a time. When sorted, all the fields are # used, regardless of what is displayed. Set this tweak to True to use only # the fields that are being displayed. sort_dates_using_visible_fields = False #: Fuzz value for trimming covers # The value used for the fuzz distance when trimming a cover. # Colors within this distance are considered equal. # The distance is in absolute intensity units. cover_trim_fuzz_value = 10 #: Control behavior of the book list # You can control the behavior of double clicks and pressing Enter on the books # list. Choices: open_viewer, do_nothing, show_book_details, # show_locked_book_details, edit_cell, edit_metadata. Selecting anything other # than open_viewer, show_book_details, or show_locked_book_details has the side # effect of disabling editing a field using a single click. # Default: open_viewer. # Example: doubleclick_on_library_view = 'do_nothing' # You can also control whether the book list scrolls per item or # per pixel. Default is per item. doubleclick_on_library_view = 'open_viewer' enter_key_behavior = 'do_nothing' horizontal_scrolling_per_column = False vertical_scrolling_per_row = False #: Language to use when sorting # Setting this tweak will force sorting to use the # collating order for the specified language. This might be useful if you run # calibre in English but want sorting to work in the language where you live. # Set the tweak to the desired ISO 639-1 language code, in lower case. # You can find the list of supported locales at # https://en.wikipedia.org/wiki/List_of_ISO_639-1_codes # Default: locale_for_sorting = '' -- use the language calibre displays in # Example: locale_for_sorting = 'fr' -- sort using French rules. # Example: locale_for_sorting = 'nb' -- sort using Norwegian rules. locale_for_sorting = '' #: The number of seconds to wait before sending emails # The number of seconds to wait before sending emails when using a # public email server like GMX/Hotmail/Gmail. Default is: 5 minutes # Setting it to lower may cause the server's SPAM controls to kick in, # making email sending fail. Changes will take effect only after a restart of # calibre. You can also change the list of hosts that calibre considers # to be public relays here. Any relay host ending with one of the suffixes # in the list below will be considered a public email server. public_smtp_relay_delay = 301 public_smtp_relay_host_suffixes = ['gmail.com', 'live.com', 'gmx.com', 'outlook.com'] #: The maximum width and height for covers saved in the calibre library # All covers in the calibre library will be resized, preserving aspect ratio, # to fit within this size. This is to prevent slowdowns caused by extremely # large covers maximum_cover_size = (1650, 2200) #: Where to send downloaded news # When automatically sending downloaded news to a connected device, calibre # will by default send it to the main memory. By changing this tweak, you can # control where it is sent. Valid values are "main", "carda", "cardb". Note # that if there isn't enough free space available on the location you choose, # the files will be sent to the location with the most free space. send_news_to_device_location = "main" #: Unified toolbar on macOS # If you enable this option and restart calibre, the toolbar will be 'unified' # with the titlebar as is normal for macOS applications. However, doing this has # various bugs, for instance the minimum width of the toolbar becomes twice # what it should be and it causes other random bugs on some systems, so turn it # on at your own risk! unified_title_toolbar_on_osx = False #: Save original file when converting/polishing from same format to same format # When calibre does a conversion from the same format to the same format, for # example, from EPUB to EPUB, the original file is saved, so that in case the # conversion is poor, you can tweak the settings and run it again. By setting # this to False you can prevent calibre from saving the original file. # Similarly, by setting save_original_format_when_polishing to False you can # prevent calibre from saving the original file when polishing. save_original_format = True save_original_format_when_polishing = True #: Number of recently viewed books to show # Right-clicking the "View" button shows a list of recently viewed books. Control # how many should be shown, here. gui_view_history_size = 15 #: Change the font size of the Book details panel in the interface # Change the font size at which book details are rendered in the side panel and # comments are rendered in the metadata edit dialog. Set it to a positive or # negative number to increase or decrease the font size. change_book_details_font_size_by = 0 #: What format to default to when using the "Unpack book" feature # The "Unpack book" feature of calibre allows direct editing of a book format. # If multiple formats are available, calibre will offer you a choice # of formats, defaulting to your preferred output format if it is available. # Set this tweak to a specific value of 'EPUB' or 'AZW3' to always default # to that format rather than your output format preference. # Set to a value of 'remember' to use whichever format you chose last time you # used the "Unpack book" feature. # Examples: # default_tweak_format = None (Use output format) # default_tweak_format = 'EPUB' # default_tweak_format = 'remember' default_tweak_format = None #: Do not preselect a completion when editing authors/tags/series/etc. # This means that you can make changes and press Enter and your changes will # not be overwritten by a matching completion. However, if you wish to use the # completions you will now have to press Tab to select one before pressing # Enter. Which technique you prefer will depend on the state of metadata in # your library and your personal editing style. # # If preselect_first_completion is False and you want Tab to accept what you # typed instead of the first completion then set tab_accepts_uncompleted_text # to True. If you do this then to select from the completions you must press # the Down or Up arrow keys. The tweak tab_accepts_uncompleted_text is ignored # if preselect_first_completion is True preselect_first_completion = False tab_accepts_uncompleted_text = False #: Completion mode when editing authors/tags/series/etc. # By default, when completing items, calibre will show you all the candidates # that start with the text you have already typed. You can instead have it show # all candidates that contain the text you have already typed. To do this, set # completion_mode to 'contains'. For example, if you type asi it will match both # Asimov and Quasimodo, whereas the default behavior would match only Asimov. completion_mode = 'prefix' #: Sort the list of libraries alphabetically # The list of libraries in the Copy to library and Quick switch menus are # normally sorted by most used. However, if there are more than a certain # number of such libraries, the sorting becomes alphabetic. You can set that # number here. The default is ten libraries. many_libraries = 10 #: Choose available output formats for conversion # Restrict the list of available output formats in the conversion dialogs. # For example, if you only want to convert to EPUB and AZW3, change this to # restrict_output_formats = ['EPUB', 'AZW3']. The default value of None causes # all available output formats to be present. restrict_output_formats = None #: Set the thumbnail image quality used by the Content server # The quality of a thumbnail is largely controlled by the compression quality # used when creating it. Set this to a larger number to improve the quality. # Note that the thumbnails get much larger with larger compression quality # numbers. # The value can be between 50 and 99 content_server_thumbnail_compression_quality = 75 #: Image file types to treat as e-books when dropping onto the "Book details" panel # Normally, if you drop any image file in a format known to calibre onto the # "Book details" panel, it will be used to set the cover. If you want to store # some image types as e-books instead, you can set this tweak. # Examples: # cover_drop_exclude = {'tiff', 'webp'} cover_drop_exclude = () #: Exclude fields when copy/pasting metadata # You can ask calibre to not paste some metadata fields when using the # Edit metadata->Copy metadata/Paste metadata actions. For example, # exclude_fields_on_paste = ['cover', 'timestamp', '#mycolumn'] # to prevent pasting of the cover, Date and custom column, mycolumn. # You can also add a shortcut in Preferences->Shortcuts->Edit metadata # to paste metadata ignoring this tweak. exclude_fields_on_paste = [] #: Skip internet connected check # Skip checking whether the internet is available before downloading news. # Useful if for some reason your operating systems network checking # facilities are not reliable (for example NetworkManager on Linux). skip_network_check = False #: Tab stop width in the template editor # Sets the width of the tab stop in the template editor in "average characters". # For example, a value of 1 results in a space with the width of one average character. template_editor_tab_stop_width = 4 #: Value for undefined numbers when sorting # Sets the value to use for undefined numbers when sorting. # For example, the value -10 sorts undefined numbers as if they were set to -10. # Use 'maximum' for the largest possible number. Use 'minimum' for the smallest # possible number. Quotes are optional if entering a number. # Examples: # value_for_undefined_numbers_when_sorting = -100 # value_for_undefined_numbers_when_sorting = '2' # value_for_undefined_numbers_when_sorting = -0.01 # value_for_undefined_numbers_when_sorting = 'minimum' # value_for_undefined_numbers_when_sorting = 'maximum' value_for_undefined_numbers_when_sorting = 0 #: Allow template database functions in composite columns # If True then the template database functions book_values() and book_count() # can be used in composite custom columns. Note: setting this tweak to True and # using these functions in composites can be very slow. # Default: False allow_template_database_functions_in_composites = False #: Change the programs that are run when opening files/URLs # By default, calibre passes URLs to the operating system to open using # whatever default programs are configured there. Here you can override # that by specifying the program to use, per URL type. For local files, # the type is "file" and for web links it is "http*". For example: # openers_by_scheme = { "http*": "firefox %u" } will make calibre run Firefox # for https://whatever URLs. %u is replaced by the URL to be opened. The scheme # takes a glob pattern allowing a single entry to match multiple URL types. openers_by_scheme = {} #: Set the first day of the week for calendar popups # It must be one of the values Default, Sunday, Monday, Tuesday, Wednesday, # Thursday, Friday, or Saturday, all in English, spelled exactly as shown. calendar_start_day_of_week = 'Default'
31,485
Python
.py
538
56.505576
136
0.737375
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,859
generate.py
kovidgoyal_calibre/imgsrc/generate.py
#!/usr/bin/env python # vim:fileencoding=utf-8 # License: GPLv3 Copyright: 2016, Kovid Goyal <kovid at kovidgoyal.net> import argparse import glob import hashlib import json import os import subprocess duplicates = { 'character-set': ['languages'], 'calibre': ['library', 'lt'], 'format-text-color': ['lookfeel'], 'books_in_series': ['series'], 'plugins.svg': ['plugins/plugin_upgrade_ok'], } sizes = { 'lt': '256', 'library': '1024', 'default_cover': 'original', 'viewer': '256', 'tweak': '256', } skip = {'calibre'} j = os.path.join base = os.path.dirname(os.path.abspath(__file__)) output_base = j(os.path.dirname(base), 'resources', 'images') hash_path = j(os.path.dirname(base), '.build-cache', 'imgsrc-gen.json') if os.path.exists(hash_path): with open(hash_path, 'rb') as f: hashes = json.load(f) else: hashes = {} src_hashes = {} def iterfiles(only=()): for src in glob.glob(j(base, '*.svg')) + glob.glob(j(base, 'plugins/*.svg')): name = os.path.relpath(src, base).rpartition('.')[0] if only and name not in only: continue src_hashes[name] = h = hashlib.sha1(open(src, 'rb').read()).hexdigest() if not only and h == hashes.get(name): continue output_names = [n for n in [name] + duplicates.get(name, []) if n not in skip] output_files = [j(output_base, n) + '.png' for n in output_names] if output_files: yield src, output_files def rsvg(src, size, dest): cmd = ['rsvg-convert', '-d', '96', '-p', '96'] if size != 'original': cmd += ['--width', size, '--height', size] subprocess.check_call(cmd + ['-o', dest, src]) subprocess.check_call(['optipng', '-o7', '-quiet', '-strip', 'all', dest]) def render(src, output_files): for dest in output_files: oname = os.path.basename(dest).rpartition('.')[0] size = sizes.get(oname, '128') print('Rendering', oname, 'at size:', size) rsvg(src, size, dest) name = os.path.relpath(src, base).rpartition('.')[0] hashes[name] = src_hashes[name] def main(): p = argparse.ArgumentParser() p.add_argument('only', nargs='*', default=[], help='Only render the specified icons') args = p.parse_args() for src, ofiles in iterfiles(args.only): render(src, ofiles) with open(hash_path, 'w') as f: json.dump(hashes, f, indent=2, sort_keys=True) if __name__ == '__main__': main()
2,496
Python
.py
70
30.614286
89
0.607217
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,860
copy-to-library.svg
kovidgoyal_calibre/imgsrc/copy-to-library.svg
<svg xmlns="http://www.w3.org/2000/svg" xmlns:svg="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" id="svg2" width="128" height="128" version="1.1" viewBox="0 0 128 128"><defs id="defs4"><marker id="marker5884" orient="auto" refX="0" refY="0" style="overflow:visible"><path id="path5886" d="M 0,0 5,-5 -12.5,0 5,5 0,0 Z" transform="matrix(-0.8,0,0,-0.8,-10,0)" style="fill:#8eea00;fill-opacity:1;fill-rule:evenodd;stroke:none;stroke-width:1pt;stroke-opacity:1"/></marker></defs><metadata id="metadata7"/><image id="image7" width="100%" height="100%" xlink:href="calibre.svg"/><g id="layer1" transform="translate(0,-924.36216)"><path id="path4" d="m 127.35998,951.6386 q 0,1.2023 -0.87854,2.0808 l -23.67428,23.6743 q -0.87854,0.8785 -2.08075,0.8785 -1.202207,0 -2.080745,-0.8785 -0.878538,-0.8786 -0.878538,-2.0808 l 0,-11.8371 -10.357497,0 q -4.531401,0 -8.114911,0.2774 -3.58351,0.2774 -7.120781,0.9941 -3.537271,0.7168 -6.149764,1.9652 -2.612494,1.2485 -4.878195,3.2136 -2.265706,1.9651 -3.699107,4.6701 -1.433402,2.705 -2.242581,6.4041 -0.80918,3.6991 -0.80918,8.3692 0,2.5431 0.231191,5.6874 0,0.2774 0.115597,1.0866 0.115598,0.8092 0.115598,1.2253 0,0.6936 -0.393032,1.156 -0.393027,0.4624 -1.086609,0.4624 -0.73982,0 -1.29469,-0.7861 -0.323668,-0.4161 -0.601103,-1.0172 -0.277434,-0.6011 -0.624222,-1.3872 -0.346794,-0.786 -0.485511,-1.1097 Q 44.5,981.5089 44.5,973.8333 q 0,-9.2015 2.450657,-15.3976 7.490688,-18.6342 40.458973,-18.6342 l 10.357497,0 0,-11.8372 q 0,-1.2021 0.878538,-2.0807 0.878538,-0.8785 2.080745,-0.8785 1.20221,0 2.08075,0.8785 l 23.67428,23.6743 q 0.87854,0.8786 0.87854,2.0807 z" style="fill:#58ab21;fill-opacity:.75720167"/></g></svg>
1,690
Python
.py
1
1,690
1,690
0.689349
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,861
render-logo.py
kovidgoyal_calibre/imgsrc/render-logo.py
#!/usr/bin/env python # vim:fileencoding=utf-8 # License: GPLv3 Copyright: 2016, Kovid Goyal <kovid at kovidgoyal.net> import os import shutil import subprocess import sys j = os.path.join base = os.path.dirname(os.path.abspath(__file__)) resources = j(os.path.dirname(base), 'resources') icons = j(os.path.dirname(base), 'icons') srv = j(os.path.dirname(os.path.dirname(base)), 'srv') def render(outpath, sz, background=None): sz = str(sz) cmd = ['rsvg-convert', j(base, 'calibre.svg'), '-w', sz, '-h', sz, '-d', '96', '-p', '96', '-o', outpath] if background: cmd.insert(2, background), cmd.insert(2, '-b') subprocess.check_call(cmd) subprocess.check_call(['optipng', '-o7', '-strip', 'all', outpath]) render(j(resources, 'images', 'library.png'), 1024) render(j(resources, 'images', 'lt.png'), 256) render(j(resources, 'images', 'apple-touch-icon.png'), 256, 'white') render(j(resources, 'content-server', 'calibre.png'), 128) render(j(srv, 'main', 'calibre-paypal-logo.png'), 60) shutil.copy2(j(resources, 'content-server', 'calibre.png'), j(resources, 'content_server', 'calibre.png')) shutil.copy2(j(resources, 'images', 'lt.png'), j(srv, 'common', 'favicon.png')) shutil.copy2(j(resources, 'images', 'lt.png'), j(srv, 'common', 'favicon.png')) subprocess.check_call([sys.executable, j(icons, 'make_ico_files.py'), 'only-logo']) shutil.copy2(j(icons, 'library.ico'), j(srv, 'common', 'favicon.ico')) shutil.copy2(j(icons, 'library.ico'), j(srv, 'main/static/resources/img', 'favicon.ico')) shutil.copy2(j(icons, 'library.ico'), j(srv, 'open-books/drmfree/static/img', 'favicon.ico')) subprocess.check_call([sys.executable, j(icons, 'icns', 'make_iconsets.py'), 'only-logo']) os.chdir(srv) subprocess.check_call(['git', 'commit', '-am', 'Update calibre favicons']) for s in 'main code open-books dl1'.split(): subprocess.check_call(['./publish', s, 'update'])
1,906
Python
.py
36
50.805556
109
0.684041
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,862
edit-copy.svg
kovidgoyal_calibre/imgsrc/edit-copy.svg
<svg xmlns="http://www.w3.org/2000/svg" xmlns:svg="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" id="svg2" width="128" height="128" version="1.1" viewBox="0 0 128 128"><defs id="defs4"><linearGradient id="linearGradient4139"><stop style="stop-color:#0c7e1b;stop-opacity:1" id="stop4141" offset="0"/><stop style="stop-color:#30b241;stop-opacity:1" id="stop4143" offset="1"/></linearGradient><radialGradient id="radialGradient4145" cx="64" cy="988.362" r="63" fx="64" fy="988.362" gradientTransform="translate(0,-3.1376577e-4)" gradientUnits="userSpaceOnUse" xlink:href="#linearGradient4139"/></defs><metadata id="metadata7"/><g id="layer1" transform="translate(0,-924.36216)"><path id="path4" d="m 118,1042.3622 0,-81.00004 -27,0 0,29.25 q 0,2.8125 -1.96875,4.78125 -1.96875,1.96875 -4.78125,1.96875 l -29.25,0 0,45.00004 63,0 z M 50.78125,985.83091 73,963.61217 l 0,-29.25001 -27,0 0,29.25001 q 0,2.81249 -1.96875,4.78124 -1.96875,1.96875 -4.78125,1.96875 l -29.25,0 0,45.00004 36,0 0,-18.00004 q 0,-2.8125 1.40625,-6.1875 1.40625,-3.375 3.375,-5.34375 z M 37,940.33872 l -21.02344,21.02344 21.02344,0 0,-21.02344 z m 45,27 -21.02343,21.02344 21.02343,0 0,-21.02344 z m 38.25,-14.97656 q 2.8125,0 4.78125,1.96875 Q 127,956.29966 127,959.11216 l 0,85.50004 q 0,2.8125 -1.96875,4.7812 -1.96875,1.9688 -4.78125,1.9688 l -67.5,0 q -2.8125,0 -4.78125,-1.9688 Q 46,1047.4247 46,1044.6122 l 0,-20.25 -38.25,0 q -2.8125,0 -4.78125,-1.9688 Q 1,1020.4247 1,1017.6122 l 0,-47.25004 q 0,-2.81249 1.40625,-6.1875 1.40625,-3.375 3.375,-5.34375 l 28.6875,-28.6875 q 1.96875,-1.96875 5.34375,-3.375 3.375,-1.40625 6.1875,-1.40625 l 29.25,0 q 2.8125,0 4.78125,1.96875 Q 82,929.29966 82,932.11217 l 0,23.06249 q 4.78125,-2.8125 9,-2.8125 l 29.25,0 z" style="fill:url(#radialGradient4145);fill-opacity:1"/><path style="fill:#2149c1;fill-opacity:1" id="path4195" d="m 37,940.33872 -21.02344,21.02344 21.02344,0 z m 36,23.27345 0,-29.25001 -27,0 c -0.02042,9.97976 0.04091,19.96133 -0.03076,29.93995 -0.250252,4.07148 -4.435187,6.63772 -8.254207,6.06005 l -27.715031,0 0,45.00004 36,0 0,9 c -12.979754,-0.02 -25.96133,0.041 -38.9399414,-0.031 -4.0714417,-0.2501 -6.63773528,-4.4349 -6.0600586,-8.254 0.014586,-15.4755 -0.0292134,-30.95197 0.021973,-46.42695 0.3402367,-5.4141 3.5223417,-10.04233 7.4727445,-13.53281 9.0340545,-8.95433 17.9139465,-18.07489 27.0462985,-26.92325 4.100446,-3.32858 9.440696,-4.12321 14.552234,-3.83203 8.834578,0.0786 17.687411,-0.15949 26.510266,0.12305 3.91375,0.64488 5.909632,4.76166 5.396484,8.41776 l 0,21.27169 c 0,0 -16.708261,15.5236 -9.000002,8.43751 z"/></g></svg>
2,609
Python
.py
1
2,609
2,609
0.701418
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,863
search_copy_saved.svg
kovidgoyal_calibre/imgsrc/search_copy_saved.svg
<svg xmlns="http://www.w3.org/2000/svg" xmlns:svg="http://www.w3.org/2000/svg" xmlns:xlink="http://www.w3.org/1999/xlink" id="svg2" width="128" height="128" version="1.1" viewBox="0 0 128 128"><defs id="defs4"><linearGradient id="linearGradient4136"><stop style="stop-color:#0b59bd;stop-opacity:1" id="stop4138" offset="0"/><stop style="stop-color:#b2cffa;stop-opacity:1" id="stop4140" offset="1"/></linearGradient><linearGradient id="linearGradient4142" x1="126.5" x2=".929" y1="1051.166" y2="925.594" gradientUnits="userSpaceOnUse" xlink:href="#linearGradient4136"/><radialGradient id="radialGradient4145" cx="64" cy="988.362" r="63" fx="64" fy="988.362" gradientTransform="matrix(0.45956341,0,0,0.45956341,13.055023,512.49001)" gradientUnits="userSpaceOnUse" xlink:href="#linearGradient4139"/><linearGradient id="linearGradient4139"><stop style="stop-color:#0c7e1b;stop-opacity:1" id="stop4141" offset="0"/><stop style="stop-color:#30b241;stop-opacity:1" id="stop4143" offset="1"/></linearGradient></defs><metadata id="metadata7"/><g id="layer1" transform="translate(0,-924.36216)"><path id="path6" d="M 122.3068,1022.695 84.283489,991.81358 c -1.3702,-1.02937 -2.05531,-3.08813 -1.02765,-4.80378 7.53615,-16.12698 4.79573,-35.68523 -8.56381,-49.06719 -17.81272,-17.8426 -47.95733,-16.81324 -64.399844,3.77442 -12.3318854,15.44072 -12.3318854,37.74396 -0.3425504,53.52782 13.0169874,16.47005 34.9403344,20.58765 52.4105044,12.35255 1.71277,-0.6862 3.42553,-0.3431 4.79574,1.0294 l 30.82971,38.0871 c 4.795731,5.8331 13.359541,6.1763 18.497831,1.0294 l 7.1936,-7.2057 c 4.79573,-4.4606 4.45317,-13.0388 -1.37022,-17.8426 z M 67.155879,992.15673 c -12.67444,12.69567 -33.570133,12.69567 -46.587123,0 -12.6744374,-12.69574 -12.6744374,-33.62647 0,-46.66532 12.674438,-13.03881 33.570133,-12.69569 46.587123,0 13.01699,13.03885 13.01699,33.96958 0,46.66532 z" style="fill:url(#linearGradient4142);fill-opacity:1"/><g id="g4154" transform="translate(-2.1428571,-0.80357143)"><path style="fill:url(#radialGradient4145);fill-opacity:1" id="path4" d="m 67.283504,991.52167 0,-37.22465 -12.40821,0 0,13.44223 q 0,1.29253 -0.904766,2.19729 -0.904765,0.90477 -2.197287,0.90477 l -13.44223,0 0,20.68036 28.952493,0 z m -30.891276,-25.9797 10.210924,-10.21092 0,-13.44224 -12.408212,0 0,13.44224 q 0,1.29251 -0.904765,2.19728 -0.904766,0.90476 -2.197288,0.90476 l -13.44223,0 0,20.68037 16.544283,0 0,-8.27215 q 0,-1.29253 0.646261,-2.84355 0.646261,-1.55103 1.551027,-2.45579 z m -6.333359,-20.90655 -9.661603,9.6616 9.661603,0 0,-9.6616 z m 20.680354,12.40821 -9.661599,9.66161 9.661599,0 0,-9.66161 z m 17.578299,-6.88268 q 1.292522,0 2.197288,0.90477 0.904765,0.90476 0.904765,2.19729 l 0,39.29268 q 0,1.29252 -0.904765,2.19726 -0.904766,0.90479 -2.197288,0.90479 l -31.020529,0 q -1.292522,0 -2.197287,-0.90479 -0.904766,-0.90474 -0.904766,-2.19726 l 0,-9.30616 -17.5783,0 q -1.292522,0 -2.197288,-0.90479 -0.904765,-0.90474 -0.904765,-2.19727 l 0,-21.71438 q 0,-1.29251 0.646261,-2.84354 0.646261,-1.55103 1.551026,-2.4558 L 28.8956,939.95003 q 0.904765,-0.90477 2.455792,-1.55103 1.551026,-0.64626 2.843548,-0.64626 l 13.44223,0 q 1.292522,0 2.197288,0.90477 0.904765,0.90476 0.904765,2.19729 l 0,10.59868 q 2.197288,-1.29253 4.136071,-1.29253 l 13.442228,0 z"/><path id="path4195" d="m 30.058869,944.63542 -9.661603,9.6616 9.661603,0 z m 16.544283,10.69563 0,-13.44224 -12.408212,0 c -0.0094,4.58633 0.0188,9.1735 -0.01414,13.75931 -0.115007,1.8711 -2.03825,3.05045 -3.793332,2.78497 l -12.736814,0 0,20.68037 16.544283,0 0,4.13607 c -5.96502,-0.009 -11.930877,0.0188 -17.895372,-0.0142 -1.871086,-0.11494 -3.050461,-2.03812 -2.784981,-3.79323 0.0067,-7.11197 -0.01342,-14.22439 0.0101,-21.33612 0.15636,-2.48813 1.618739,-4.61509 3.4342,-6.21919 4.151721,-4.11508 8.232595,-8.30656 12.429489,-12.37294 1.884415,-1.52969 4.338599,-1.89487 6.687675,-1.76106 4.060048,0.0361 8.128487,-0.0733 12.183148,0.0566 1.798616,0.29636 2.715851,2.18829 2.480027,3.86849 l 0,9.7757 c 0,0 -7.678506,7.13407 -4.136072,3.87757 z" style="fill:#2149c1;fill-opacity:1"/></g></g></svg>
4,063
Python
.py
1
4,063
4,063
0.717696
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,864
copy.svg
kovidgoyal_calibre/imgsrc/srv/copy.svg
<svg width="1792" height="1792" viewBox="0 0 1792 1792" xmlns="http://www.w3.org/2000/svg"><path d="M1696 384q40 0 68 28t28 68v1216q0 40-28 68t-68 28h-960q-40 0-68-28t-28-68v-288h-544q-40 0-68-28t-28-68v-672q0-40 20-88t48-76l408-408q28-28 76-48t88-20h416q40 0 68 28t28 68v328q68-40 128-40h416zm-544 213l-299 299h299v-299zm-640-384l-299 299h299v-299zm196 647l316-316v-416h-384v416q0 40-28 68t-68 28h-416v640h512v-256q0-40 20-88t48-76zm956 804v-1152h-384v416q0 40-28 68t-68 28h-416v640h896z"/></svg>
497
Python
.py
1
497
497
0.770624
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,865
generate.py
kovidgoyal_calibre/imgsrc/srv/generate.py
#!/usr/bin/env python # vim:fileencoding=utf-8 # License: GPLv3 Copyright: 2016, Kovid Goyal <kovid at kovidgoyal.net> import os import re import sys from lxml import etree SVG_NS = 'http://www.w3.org/2000/svg' XLINK_NS = 'http://www.w3.org/1999/xlink' def clone_node(node, parent): ans = parent.makeelement(node.tag) for k in node.keys(): ans.set(k, node.get(k)) ans.text, ans.tail = node.text, node.tail for child in node.iterchildren('*'): clone_node(child, ans) parent.append(ans) return ans def merge(): base = os.path.dirname(os.path.abspath(__file__)) ans = etree.fromstring( '<svg xmlns="%s" xmlns:xlink="%s"/>' % (SVG_NS, XLINK_NS), parser=etree.XMLParser( recover=True, no_network=True, resolve_entities=False ) ) for f in os.listdir(base): if not f.endswith('.svg'): continue with open(os.path.join(base, f), 'rb') as ff: raw = ff.read() svg = etree.fromstring( raw, parser=etree.XMLParser( recover=True, no_network=True, resolve_entities=False ) ) symbol = ans.makeelement('{%s}symbol' % SVG_NS) symbol.set('viewBox', svg.get('viewBox')) symbol.set('id', 'icon-' + f.rpartition('.')[0]) for child in svg.iterchildren('*'): clone_node(child, symbol) ans.append(symbol) ans = etree.tostring(ans, encoding='unicode', pretty_print=True, with_tail=False) ans = re.sub('<svg[^>]+>', '<svg style="display:none">', ans, count=1) return ans if __name__ == '__main__': sys.stdout.write(merge().encode('utf-8'))
1,691
Python
.py
48
28.4375
85
0.600122
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,866
the_daily_news_egypt.recipe
kovidgoyal_calibre/recipes/the_daily_news_egypt.recipe
__license__ = 'GPL v3' __copyright__ = '2011, Pat Stapleton <pat.stapleton at gmail.com>' ''' abc.net.au/news ''' import re from calibre.web.feeds.recipes import BasicNewsRecipe class TheDailyNewsEG(BasicNewsRecipe): title = u'The Daily News Egypt' __author__ = 'Omm Mishmishah' description = 'News from Egypt' masthead_url = 'http://www.thedailynewsegypt.com/images/DailyNews-03_05.gif' cover_url = 'http://www.thedailynewsegypt.com/images/DailyNews-03_05.gif' auto_cleanup = True oldest_article = 7 max_articles_per_feed = 100 no_stylesheets = False use_embedded_content = False encoding = 'utf8' publisher = 'The Daily News Egypt' category = 'News, Egypt, World' language = 'en_EG' publication_type = 'newsportal' # preprocess_regexps = [(re.compile(r'<!--.*?-->', re.DOTALL), lambda m: '')] # Remove annoying map links (inline-caption class is also used for some # image captions! hence regex to match maps.google) preprocess_regexps = [(re.compile( r'<a class="inline-caption" href="http://maps\.google\.com.*?/a>', re.DOTALL), lambda m: '')] conversion_options = { 'comments': description, 'tags': category, 'language': language, 'publisher': publisher, 'linearize_tables': False } keep_only_tags = [dict(attrs={'class': ['article section']})] remove_tags = [dict(attrs={'class': ['related', 'tags', 'tools', 'attached-content ready', 'inline-content story left', 'inline-content map left contracted', 'published', 'story-map', 'statepromo', 'topics', ]})] remove_attributes = ['width', 'height'] feeds = [(u'The Daily News Egypt', u'http://www.thedailynewsegypt.com/rss.php?sectionid=all')]
1,813
Python
.gyp
38
41.236842
122
0.642898
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,867
ancient_egypt.recipe
kovidgoyal_calibre/recipes/ancient_egypt.recipe
#!/usr/bin/env python # vim:fileencoding=utf-8 ''' https://ancientegyptmagazine.com ''' from calibre import browser from calibre.web.feeds.news import BasicNewsRecipe class ancientegypt(BasicNewsRecipe): title = 'The Past: Ancient Egypt Magazine' language = 'en' __author__ = 'unkn0wn' description = ( 'Ancient Egypt is the world\'s leading Egyptology magazine, exploring the history, people and culture of the Nile Valley. ' 'Now in a larger format with a fresh new design, AE brings you the latest news and discoveries, and feature articles covering ' 'more than 5000 years of Egyptian history. Published bimonthly.' ) no_stylesheets = True use_embedded_content = False remove_attributes = ['style', 'height', 'width'] ignore_duplicate_articles = {'url'} resolve_internal_links = True masthead_url = 'https://ancientegyptmagazine.com/media/website/ae-logo-2.png' simultaneous_downloads = 1 extra_css = ''' [class^="meta"] { font-size:small; } .post-subtitle { font-style: italic; color:#202020; } .wp-block-image { font-size:small; text-align:center; } ''' keep_only_tags = [ dict(attrs={'class':lambda x: x and '__header' in x}), dict(attrs={'class':lambda x: x and '__background' in x}), dict(attrs={'class':lambda x: x and '__body_area' in x}), ] remove_tags = [ dict(attrs={'class':'ad-break'}), dict(attrs={'class':lambda x: x and 'avatar' in x.split()}), dict(attrs={'class':lambda x: x and '--share' in x}) ] def preprocess_html(self, soup): exp = soup.find(attrs={'class':lambda x: x and 'post-subtitle' in x.split()}) if exp: exp.name = 'p' return soup recipe_specific_options = { 'issue': { 'short': 'Enter the Issue Number you want to download ', 'long': 'For example, 136' } } def parse_index(self): soup = self.index_to_soup('https://the-past.com/category/magazines/ae/') art = soup.find('article', attrs={'class':lambda x: x and 'tag-magazines' in x.split()}) url = art.h2.a['href'] d = self.recipe_specific_options.get('issue') if d and isinstance(d, str): url = 'https://the-past.com/magazines/ae/ancient-egypt-magazine-' + d + '/' issue = self.index_to_soup(url) ti = issue.find('h1', attrs={'class':lambda x: x and 'post-title' in x.split()}) if ti: self.title = self.tag_to_string(ti).strip() dt = soup.find(attrs={'class':lambda x: x and '__date' in x}) if dt: self.timefmt = ' [' + self.tag_to_string(dt).strip() + ']' edit = issue.find('h2', attrs={'id':'from-the-editor'}) if edit and edit.findParent('div'): self.description = self.tag_to_string(edit.findParent('div')) cov = issue.find('figure', attrs={'class':lambda x: x and 'wp-block-image' in x.split()}) if cov: self.cover_url = cov.img['src'] div = issue.find('div', attrs={'class':lambda x: x and 'entry-content' in x.split()}) feeds = [] h2 = div.findAll('h2', attrs={'class':lambda x: x and 'wp-block-heading' in x.split()}) lt = div.findAll(attrs={'class':'display-posts-listing'}) for x, y in zip(h2, lt): section = self.tag_to_string(x).strip() self.log(section) articles = [] for a in y.findAll('a', href=True, attrs={'class':'title'}): url = a['href'] title = self.tag_to_string(a).strip() desc = '' exp = a.findNext(attrs={'class':'excerpt'}) if exp: desc = self.tag_to_string(exp).strip() self.log('\t', title, '\n\t', desc, '\n\t\t', url) articles.append({'title': title, 'description':desc, 'url': url}) if articles: feeds.append((section, articles)) return feeds def get_browser(self, *args, **kwargs): return self def clone_browser(self, *args, **kwargs): return self.get_browser() def open_novisit(self, *args, **kwargs): br = browser() return br.open_novisit(*args, **kwargs) open = open_novisit
4,347
Python
.gyp
97
36.061856
135
0.582428
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,868
run-python.h
kovidgoyal_calibre/bypy/run-python.h
/* * Copyright (C) 2019 Kovid Goyal <kovid at kovidgoyal.net> * * Distributed under terms of the GPL3 license. */ #pragma once #define PY_SSIZE_T_CLEAN #include <stdio.h> #include <stdbool.h> #include <time.h> #include <stdlib.h> #include <stdarg.h> #ifdef _WIN32 #include <string.h> #define PATH_MAX MAX_PATH #else #include <strings.h> #endif #include <errno.h> #include <Python.h> #ifdef __APPLE__ #include <os/log.h> #endif #include <bypy-freeze.h> static void pre_initialize_interpreter(bool is_gui_app) { bypy_pre_initialize_interpreter(is_gui_app); } #define decode_char_buf(src, dest) { \ size_t tsz; \ wchar_t* t__ = Py_DecodeLocale(src, &tsz); \ if (!t__) fatal("Failed to decode path: %s", src); \ if (tsz > sizeof(dest) - 1) tsz = sizeof(dest) - 1; \ memcpy(dest, t__, tsz * sizeof(wchar_t)); \ dest[tsz] = 0; \ PyMem_RawFree(t__); \ } #define MAX_SYS_PATHS 3 typedef struct { int argc; wchar_t exe_path[PATH_MAX], python_home_path[PATH_MAX], python_lib_path[PATH_MAX]; wchar_t extensions_path[PATH_MAX], resources_path[PATH_MAX], executables_path[PATH_MAX]; #ifdef __APPLE__ wchar_t bundle_resource_path[PATH_MAX], frameworks_path[PATH_MAX]; #elif defined(_WIN32) wchar_t app_dir[PATH_MAX]; #endif const wchar_t *basename, *module, *function; #ifdef _WIN32 wchar_t* const *argv; #else char* const *argv; #endif } InterpreterData; static InterpreterData interpreter_data = {0}; static void run_interpreter() { bypy_initialize_interpreter( interpreter_data.exe_path, interpreter_data.python_home_path, L"site", interpreter_data.extensions_path, interpreter_data.argc, interpreter_data.argv); set_sys_bool("gui_app", use_os_log); set_sys_bool("frozen", true); set_sys_string("calibre_basename", interpreter_data.basename); set_sys_string("calibre_module", interpreter_data.module); set_sys_string("calibre_function", interpreter_data.function); set_sys_string("extensions_location", interpreter_data.extensions_path); set_sys_string("resources_location", interpreter_data.resources_path); set_sys_string("executables_location", interpreter_data.executables_path); #ifdef __APPLE__ set_sys_string("resourcepath", interpreter_data.bundle_resource_path); set_sys_string("frameworks_dir", interpreter_data.frameworks_path); set_sys_bool("new_app_bundle", true); #elif defined(_WIN32) set_sys_string("app_dir", interpreter_data.app_dir); set_sys_bool("new_app_layout", true); #else set_sys_string("frozen_path", interpreter_data.executables_path); #endif int ret = bypy_run_interpreter(); exit(ret); }
2,641
Python
.pyt
81
29.962963
116
0.717366
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,869
ipython.py
kovidgoyal_calibre/src/calibre/utils/ipython.py
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2012, Kovid Goyal <kovid@kovidgoyal.net>' __docformat__ = 'restructuredtext en' import os import re import sys from calibre.constants import cache_dir, get_version, iswindows from polyglot.builtins import exec_path ipydir = os.path.join(cache_dir(), 'ipython') BANNER = ('Welcome to the interactive calibre shell!\n') def setup_pyreadline(): config = ''' #Bind keys for exit (keys only work on empty lines #disable_readline(True) #Disable pyreadline completely. debug_output("off") #"on" saves log info to./pyreadline_debug_log.txt #"on_nologfile" only enables print warning messages bind_exit_key("Control-d") bind_exit_key("Control-z") #Commands for moving bind_key("Home", "beginning_of_line") bind_key("End", "end_of_line") bind_key("Left", "backward_char") bind_key("Control-b", "backward_char") bind_key("Right", "forward_char") bind_key("Control-f", "forward_char") bind_key("Alt-f", "forward_word") bind_key("Alt-b", "backward_word") bind_key("Clear", "clear_screen") bind_key("Control-l", "clear_screen") bind_key("Control-a", "beginning_of_line") bind_key("Control-e", "end_of_line") #bind_key("Control-l", "redraw_current_line") #Commands for Manipulating the History bind_key("Return", "accept_line") bind_key("Control-p", "previous_history") bind_key("Control-n", "next_history") bind_key("Up", "history_search_backward") bind_key("Down", "history_search_forward") bind_key("Alt-<", "beginning_of_history") bind_key("Alt->", "end_of_history") bind_key("Control-r", "reverse_search_history") bind_key("Control-s", "forward_search_history") bind_key("Alt-p", "non_incremental_reverse_search_history") bind_key("Alt-n", "non_incremental_forward_search_history") bind_key("Control-z", "undo") bind_key("Control-_", "undo") #Commands for Changing Text bind_key("Delete", "delete_char") bind_key("Control-d", "delete_char") bind_key("BackSpace", "backward_delete_char") #bind_key("Control-Shift-v", "quoted_insert") bind_key("Control-space", "self_insert") bind_key("Control-BackSpace", "backward_delete_word") #Killing and Yanking bind_key("Control-k", "kill_line") bind_key("Control-shift-k", "kill_whole_line") bind_key("Escape", "kill_whole_line") bind_key("Meta-d", "kill_word") bind_key("Control-w", "unix_word_rubout") #bind_key("Control-Delete", "forward_kill_word") #Copy paste bind_key("Shift-Right", "forward_char_extend_selection") bind_key("Shift-Left", "backward_char_extend_selection") bind_key("Shift-Control-Right", "forward_word_extend_selection") bind_key("Shift-Control-Left", "backward_word_extend_selection") bind_key("Control-m", "set_mark") bind_key("Control-Shift-x", "copy_selection_to_clipboard") #bind_key("Control-c", "copy_selection_to_clipboard") #Needs allow_ctrl_c(True) below to be uncommented bind_key("Control-q", "copy_region_to_clipboard") bind_key('Control-Shift-v', "paste_mulitline_code") bind_key("Control-x", "cut_selection_to_clipboard") bind_key("Control-v", "paste") bind_key("Control-y", "yank") bind_key("Alt-v", "ipython_paste") #Unbinding keys: #un_bind_key("Home") #Other bell_style("none") #modes: none, audible, visible(not implemented) show_all_if_ambiguous("on") mark_directories("on") completer_delims(" \t\n\"\\'`@$><=;|&{(?") complete_filesystem("on") debug_output("off") #allow_ctrl_c(True) #(Allows use of ctrl-c as copy key, still propagate keyboardinterrupt when not waiting for input) history_filename(%r) history_length(2000) #value of -1 means no limit #set_mode("vi") #will cause following bind_keys to bind to vi mode as well as activate vi mode #ctrl_c_tap_time_interval(0.3) ''' try: import pyreadline.rlmain if not os.path.exists(ipydir): os.makedirs(ipydir) conf = os.path.join(ipydir, 'pyreadline.txt') hist = os.path.join(ipydir, 'history.txt') config = config % hist with open(conf, 'wb') as f: f.write(config.encode('utf-8')) pyreadline.rlmain.config_path = conf import atexit import readline import pyreadline.unicode_helper # noqa # Normally the codepage for pyreadline is set to be sys.stdout.encoding # if you need to change this uncomment the following line # pyreadline.unicode_helper.pyreadline_codepage="utf8" except ImportError: print("Module readline not available.") else: # import tab completion functionality import rlcompleter # Override completer from rlcompleter to disable automatic ( on callable completer_obj = rlcompleter.Completer() def nop(val, word): return word completer_obj._callable_postfix = nop readline.set_completer(completer_obj.complete) # activate tab completion readline.parse_and_bind("tab: complete") readline.read_history_file() atexit.register(readline.write_history_file) del readline, rlcompleter, atexit class Exit: def __repr__(self): raise SystemExit(0) __str__ = __repr__ def __call__(self): raise SystemExit(0) class Helper: def __repr__(self): return "Type help() for interactive help, " \ "or help(object) for help about object." def __call__(self, *args, **kwds): import pydoc return pydoc.help(*args, **kwds) def simple_repl(user_ns={}): if iswindows: setup_pyreadline() else: try: import rlcompleter # noqa import readline # noqa readline.parse_and_bind("tab: complete") except ImportError: pass user_ns = user_ns or {} import sys, re # noqa for x in ('os', 'sys', 're'): user_ns[x] = user_ns.get(x, globals().get(x, locals().get(x))) user_ns['exit'] = Exit() user_ns['help'] = Helper() from code import InteractiveConsole console = InteractiveConsole(user_ns) console.interact(BANNER + 'Use exit to quit') def ipython(user_ns=None): os.environ['IPYTHONDIR'] = ipydir have_ipython = True try: from IPython.terminal.embed import InteractiveShellEmbed from IPython.terminal.prompts import Prompts, Token from traitlets.config.loader import Config except ImportError: have_ipython = False if not have_ipython: return simple_repl(user_ns=user_ns) class CustomPrompt(Prompts): def in_prompt_tokens(self, cli=None): return [ (Token.Prompt, 'calibre['), (Token.PromptNum, get_version()), (Token.Prompt, ']> '), ] def out_prompt_tokens(self): return [] defns = {'os':os, 're':re, 'sys':sys} defns.update(user_ns or {}) c = Config() user_conf = os.path.expanduser('~/.ipython/profile_default/ipython_config.py') if os.path.exists(user_conf): exec_path(user_conf, {'get_config': lambda: c}) c.TerminalInteractiveShell.prompts_class = CustomPrompt c.InteractiveShellApp.exec_lines = [ 'from __future__ import division, absolute_import, unicode_literals, print_function', ] c.TerminalInteractiveShell.confirm_exit = False c.TerminalInteractiveShell.banner1 = BANNER c.BaseIPythonApplication.ipython_dir = ipydir c.InteractiveShell.separate_in = '' c.InteractiveShell.separate_out = '' c.InteractiveShell.separate_out2 = '' ipshell = InteractiveShellEmbed.instance(config=c, user_ns=user_ns) ipshell()
8,109
Python
.pyt
192
37.15625
118
0.631652
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,870
planet_python.recipe
kovidgoyal_calibre/recipes/planet_python.recipe
from calibre.web.feeds.news import AutomaticNewsRecipe class BasicUserRecipe(AutomaticNewsRecipe): title = u'Planet Python' language = 'en' __author__ = 'Jelle van der Waa' oldest_article = 10 max_articles_per_feed = 100 feeds = [(u'Planet Python', u'http://planetpython.org/rss20.xml')]
314
Python
.pyt
8
35
70
0.713816
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,871
substack.recipe
kovidgoyal_calibre/recipes/substack.recipe
#!/usr/bin/env python # vim:fileencoding=utf-8 # # Title: Substack # License: GNU General Public License v3 – https://www.gnu.org/licenses/gpl-3.0.html # Copyright: Nathan Cook (nathan.cook@gmail.com) ## # Written: 2020-12-18 ## __license__ = 'GNU General Public License v3 – https://www.gnu.org/licenses/gpl-3.0.html' __copyright__ = 'Nathan Cook – 2020-12-19' __version__ = 'v0.1.1' __date__ = '2020-12-19' __author__ = 'topynate' import json from calibre.web.feeds.news import BasicNewsRecipe from mechanize import Request class Substack(BasicNewsRecipe): title = 'Substack' __author__ = 'topynate' oldest_article = 7 language = 'en' max_articles_per_feed = 100 auto_cleanup = True needs_subscription = 'optional' use_embedded_content = False recipe_specific_options = { 'days': { 'short': 'Oldest article to download from this news source. In days ', 'long': 'For example, 0.5, gives you articles from the past 12 hours', 'default': str(oldest_article) } } def __init__(self, *args, **kwargs): BasicNewsRecipe.__init__(self, *args, **kwargs) d = self.recipe_specific_options.get('days') if d and isinstance(d, str): self.oldest_article = float(d) # Every Substack publication has an RSS feed at https://{name}.substack.com/feed. # The same URL provides either all posts, or all free posts + previews of paid posts, # depending on whether you're logged in. feeds = [ ('Novum Lumen', 'https://novumlumen.substack.com/feed'), # gratuitously self-promotional example ] def get_browser(self): br = BasicNewsRecipe.get_browser(self) if self.username is not None and self.password is not None: br.open('https://substack.com/account/login?redirect=%2F&email=&with_password=') data = json.dumps({'email': self.username, 'password': self.password, 'captcha_response':None}) req = Request( url='https://substack.com/api/v1/email-login', headers={ 'Accept': '*/*', 'Content-Type': 'application/json', 'Origin': 'https://substack.com', 'Referer': 'https://substack.com/account/login?redirect=%2F&email=&with_password=', }, data=data, method='POST') res = br.open(req) if res.getcode() != 200: raise ValueError('Login failed, check username and password') return br
2,630
Python
.tac
63
33.968254
107
0.605016
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,872
attac_es.recipe
kovidgoyal_calibre/recipes/attac_es.recipe
# vim:fileencoding=utf-8 from __future__ import unicode_literals from calibre.web.feeds.news import BasicNewsRecipe class AttacEspanaRecipe (BasicNewsRecipe): __author__ = 'Marc Busqué <marc@lamarciana.com>' __url__ = 'http://www.lamarciana.com' __version__ = '1.0.2' __license__ = 'GPL v3' __copyright__ = '2012, Marc Busqué <marc@lamarciana.com>' title = u'attac.es' description = u'La Asociación por la Tasación de las Transacciones Financieras y por la Ayuda a los Ciudadanos (ATTAC) es un movimiento internacional altermundialista que promueve el control democrático de los mercados financieros y las instituciones encargadas de su control mediante la reflexión política y la movilización social.' # noqa url = 'http://www.attac.es' language = 'es' tags = 'contrainformación, información alternativa' oldest_article = 7 remove_empty_feeds = True no_stylesheets = True def get_extra_css(self): if not self.extra_css: br = self.get_browser() self.extra_css = br.open_novisit( 'https://raw.githubusercontent.com/laMarciana/gutenweb/master/dist/gutenweb.css').read().replace('@charset "UTF-8";', '') return self.extra_css cover_url = u'http://www.attac.es/wp-content/themes/attacweb/images/attaces.jpg' feeds = [ (u'Attac', u'http://www.attac.es/feed'), ]
1,406
Python
.tac
27
45.592593
345
0.691123
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,873
stackoverflow.recipe
kovidgoyal_calibre/recipes/stackoverflow.recipe
#!/usr/bin/env python __license__ = 'GPL v3' __copyright__ = '2009, Darko Miletic <darko.miletic at gmail.com>' ''' blog.stackoverflow.com ''' from calibre.web.feeds.news import BasicNewsRecipe class StackOverflowBlog(BasicNewsRecipe): title = 'Stack Overflow - Blog' __author__ = 'Darko Miletic' description = 'a programming community exploit' category = 'blog, programming' publisher = 'StackOverflow team' oldest_article = 30 language = 'en' max_articles_per_feed = 100 no_stylesheets = True use_embedded_content = True encoding = 'utf-8' html2lrf_options = [ '--comment', description, '--category', category, '--publisher', publisher ] html2epub_options = 'publisher="' + publisher + \ '"\ncomments="' + description + '"\ntags="' + category + '"' feeds = [(u'Articles', u'http://blog.stackoverflow.com/feed/')]
899
Python
.tac
25
31.6
82
0.666282
kovidgoyal/calibre
19,243
2,250
4
GPL-3.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,874
tldata.py
andikleen_pmu-tools/tldata.py
import os import csv import re from collections import defaultdict import gen_level class TLData: """Read a toplev output CSV file. Exported: times[n] All time stamps vals[n] All values, as dicts mapping (name, cpu)->float levels{name} All levels (includes metrics), name->list of fields units{name} All units, name->unit headers(set) All headers (including metrics) metrics(set) All metrics helptxt[col] All help texts. cpus(set) All CPUs """ def __init__(self, fn, verbose=False): self.times = [] self.vals = [] self.fn = fn self.levels = defaultdict(set) self.metrics = set() self.headers = set() self.mtime = None self.helptxt = {} self.cpus = set() self.verbose = verbose self.units = {} def update(self): mtime = os.path.getmtime(self.fn) if self.mtime == mtime: return self.mtime = mtime csvf = csv.reader(open(self.fn, 'r')) prevts = None val = {} for r in csvf: if r[0].strip().startswith("#"): continue if r[0] == "Timestamp" or r[0] == "CPUs": continue # 1.001088024,C1,Frontend_Bound,42.9,% Slots,,frontend_retired.latency_ge_4:pp,0.0,100.0,<==,Y if re.match(r'[CS]?\d+.*', r[1]): ts, cpu, name, pct, unit, helptxt = r[0], r[1], r[2], r[3], r[4], r[5] else: ts, name, pct, unit, helptxt = r[0], r[1], r[2], r[3], r[4] cpu = None key = (name, cpu) ts, pct = float(ts), float(pct.replace("%", "")) if name not in self.helptxt or self.helptxt[name] == "": self.helptxt[name] = helptxt if unit.endswith("<"): unit = unit[:-2] if not self.verbose: continue self.units[name] = unit if prevts and ts != prevts: self.times.append(prevts) self.vals.append(val) val = {} val[key] = pct n = gen_level.level_name(name) if cpu: self.cpus.add(cpu) self.headers.add(name) if gen_level.is_metric(name): self.metrics.add(n) self.levels[n].add(name) prevts = ts if len(val.keys()) > 0: self.times.append(prevts) self.vals.append(val) early_plots = ["TopLevel", "CPU utilization", "Power", "Frequency", "CPU-METRIC"] def sort_key(i, data): if i in early_plots: return early_plots.index(i) if i in data.metrics: return 30 return list(data.levels.keys()).index(i) def level_order(data): """Return plot order of all levels.""" return sorted(data.levels.keys(), key=lambda a: sort_key(a, data))
2,915
Python
.py
83
25.373494
106
0.529204
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,875
msr.py
andikleen_pmu-tools/msr.py
#!/usr/bin/env python3 # library and tool to access Intel MSRs (model specific registers) # Author: Andi Kleen from __future__ import print_function import glob import struct import os def writemsr(msr, val): n = glob.glob('/dev/cpu/[0-9]*/msr') for c in n: f = os.open(c, os.O_WRONLY) os.lseek(f, msr, os.SEEK_SET) os.write(f, struct.pack('Q', val)) os.close(f) if not n: raise OSError("msr module not loaded (run modprobe msr)") def readmsr(msr, cpu = 0): f = os.open('/dev/cpu/%d/msr' % (cpu,), os.O_RDONLY) os.lseek(f, msr, os.SEEK_SET) val = struct.unpack('Q', os.read(f, 8))[0] os.close(f) return val def changebit(msr, bit, val): n = glob.glob('/dev/cpu/[0-9]*/msr') for c in n: f = os.open(c, os.O_RDWR) os.lseek(f, msr, os.SEEK_SET) v = struct.unpack('Q', os.read(f, 8))[0] if val: v = v | (1 << bit) else: v = v & ~(1 << bit) os.lseek(f, msr, os.SEEK_SET) os.write(f, struct.pack('Q', v)) os.close(f) if not n: raise OSError("msr module not loaded (run modprobe msr)") if __name__ == '__main__': import argparse def parse_hex(s): try: return int(s, 16) except ValueError: raise argparse.ArgumentError("Bad hex number %s" % (s)) if not os.path.exists("/dev/cpu/0/msr"): os.system("/sbin/modprobe msr") p = argparse.ArgumentParser(description='Access x86 model specific registers.') p.add_argument('msr', type=parse_hex, help='number of the MSR to access') p.add_argument('value', nargs='?', type=parse_hex, help='value to write (if not specified read)') p.add_argument('--setbit', type=int, help='Bit number to set') p.add_argument('--clearbit', type=int, help='Bit number to clear') p.add_argument('--cpu', type=int, default=0, help='CPU to read on (writes always change all)') args = p.parse_args() if args.value is None and not args.setbit and not args.clearbit: print("%x" % (readmsr(args.msr, args.cpu))) elif args.setbit: changebit(args.msr, args.setbit, 1) elif args.clearbit: changebit(args.msr, args.clearbit, 0) else: writemsr(args.msr, args.value)
2,291
Python
.py
61
31.262295
101
0.603689
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,876
counterdiff.py
andikleen_pmu-tools/counterdiff.py
#!/usr/bin/env python3 # counterdiff.py < plog program .. (or general perf arguments) # verify plog.* output from toplev by running event one by one # this can be used to estimate multiplexing measurement errors from __future__ import print_function import sys, os def run(x): print(x) os.system(x) for l in sys.stdin: if l.find(",") < 0: continue n = l.strip().split(",") run("perf stat --output l -x, -e %s %s" % (n[1], " ".join(sys.argv[1:]))) f = open("l", "r") for i in f: if i.find(",") < 0: continue j = i.strip().split(",") break f.close() if float(n[0]) > 0: delta = (float(j[0]) - float(n[0])) / float(n[0]) else: delta = 0 print(n[1], j[0], n[0], "%.2f" % (delta * 100.0))
803
Python
.py
27
24.518519
67
0.546512
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,877
utilized.py
andikleen_pmu-tools/utilized.py
#!/usr/bin/python # extract utilized CPUs out of toplev CSV output # toplev ... -I 1000 --node +CPU_Utilization -x, -o x.csv ... # utilized.py < x.csv # note it duplicates the core output from __future__ import print_function import argparse import csv import sys import re import collections ap = argparse.ArgumentParser() ap.add_argument('--min-util', default=10., type=float) ap.add_argument('file', nargs='?', type=argparse.FileType('r'), default=sys.stdin) ap.add_argument('--output', '-o', type=argparse.FileType('w'), default=sys.stdout) args = ap.parse_args() key = None c = csv.reader(args.file) wr = csv.writer(args.output) fields = collections.OrderedDict() util = collections.defaultdict(list) for t in c: if len(t) < 3 or t[0].startswith("#"): continue if t[0] == "Timestamp": wr.writerow(t) key = t[1] # XXX handle no -I if key in fields: fields[key].append(t) else: fields[key] = [t] if t[2] == "CPU_Utilization": util[key].append(float(t[3])) final = [] skipped = [] for j in fields.keys(): if "-T" not in j and not j.startswith("CPU"): if "S" in j: final.append(j) continue core = re.sub(r'-T\d+', '', j) utilization = 100 if len(util[j]) > 0: utilization = (sum(util[j]) / len(util[j])) * 100. if utilization >= float(args.min_util): for k in fields[core] + fields[j]: wr.writerow(k) else: skipped.append(j) for j in final: for k in fields[j]: wr.writerow(k) print("skipped", " ".join(skipped), file=sys.stderr)
1,600
Python
.py
53
25.981132
82
0.630759
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,878
ivb_server_ratios.py
andikleen_pmu-tools/ivb_server_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel Xeon E5 v2 (code named IvyBridge EP) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 6 Mem_L2_Store_Cost = 9 Mem_L3_Weight = 7 Mem_STLB_Hit_Cost = 7 BAClear_Cost = 12 MS_Switches_Cost = 3 Avg_Assist_Cost = 66 Pipeline_Width = 4 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 15.6 EBS_Mode = 0 DS = 1 # Aux. formulas def Backend_Bound_Cycles(self, EV, level): return (STALLS_TOTAL(self, EV, level) + EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - Few_Uops_Executed_Threshold(self, EV, level) - Frontend_RS_Empty_Cycles(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) def Cycles_0_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:i1:c1", level)) / 2 if smt_enabled else(STALLS_TOTAL(self, EV, level) - Frontend_RS_Empty_Cycles(self, EV, level)) def Cycles_1_Port_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) - EV("UOPS_EXECUTED.CORE:c2", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level)) def Cycles_2_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c2", level) - EV("UOPS_EXECUTED.CORE:c3", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level)) def Cycles_3m_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c3", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) def Fetched_Uops(self, EV, level): return (EV("IDQ.DSB_UOPS", level) + EV("LSD.UOPS", level) + EV("IDQ.MITE_UOPS", level) + EV("IDQ.MS_UOPS", level)) def Few_Uops_Executed_Threshold(self, EV, level): EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) return EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) if (IPC(self, EV, level)> 1.8) else EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return (1 *(EV("FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE", level) + EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", level)) + 2 * EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", level) + 4 *(EV("FP_COMP_OPS_EXE.SSE_PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_DOUBLE", level)) + 8 * EV("SIMD_FP_256.PACKED_SINGLE", level)) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE", level) + EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", level) + EV("FP_COMP_OPS_EXE.SSE_PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_DOUBLE", level) def Frontend_RS_Empty_Cycles(self, EV, level): EV("RS_EVENTS.EMPTY_CYCLES", level) return EV("RS_EVENTS.EMPTY_CYCLES", level) if (self.Fetch_Latency.compute(EV)> 0.1) else 0 def Frontend_Latency_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE", level)) , level ) def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def ITLB_Miss_Cycles(self, EV, level): return (12 * EV("ITLB_MISSES.STLB_HIT", level) + EV("ITLB_MISSES.WALK_DURATION", level)) def LOAD_L1_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L2_HIT", level) + EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) + EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT", level) + EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM", level) + EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS", level) def LOAD_L1_MISS_NET(self, EV, level): return LOAD_L1_MISS(self, EV, level) + EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM", level) + EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM", level) + EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM", level) + EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD", level) def LOAD_L3_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_LCL_MEM(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.LOCAL_DRAM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_FWD(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_HITM(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_MEM(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HITM(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def Mem_L3_Hit_Fraction(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) / (EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) + Mem_L3_Weight * EV("MEM_LOAD_UOPS_RETIRED.LLC_MISS", level)) def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_UOPS_RETIRED.LOCK_LOADS", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) def Memory_Bound_Fraction(self, EV, level): return (STALLS_MEM_ANY(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) / Backend_Bound_Cycles(self, EV, level) def Mispred_Clears_Fraction(self, EV, level): return EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c6", level)) , level ) def SQ_Full_Cycles(self, EV, level): return (EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) / 2) if smt_enabled else EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) def STALLS_MEM_ANY(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("CYCLE_ACTIVITY.STALLS_LDM_PENDING", level)) , level ) def STALLS_TOTAL(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("CYCLE_ACTIVITY.CYCLES_NO_EXECUTE", level)) , level ) def Store_L2_Hit_Cycles(self, EV, level): return EV("L2_RQSTS.RFO_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 60 def Mem_XSNP_Hit_Cost(self, EV, level): return 43 def Mem_XSNP_None_Cost(self, EV, level): return 41 def Mem_Local_DRAM_Cost(self, EV, level): return 200 def Mem_Remote_DRAM_Cost(self, EV, level): return 310 def Mem_Remote_HitM_Cost(self, EV, level): return 200 def Mem_Remote_Fwd_Cost(self, EV, level): return 180 def Recovery_Cycles(self, EV, level): return (EV("INT_MISC.RECOVERY_CYCLES_ANY", level) / 2) if smt_enabled else EV("INT_MISC.RECOVERY_CYCLES", level) def Retire_Fraction(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return EV("UOPS_RETIRED.RETIRE_SLOTS", level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return 8 if smt_enabled else 4 # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return Pipeline_Width * CORE_CLKS(self, EV, level) # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return ((EV("CPU_CLK_UNHALTED.THREAD", level) / 2) * (1 + EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_XCLK", level))) if ebs_mode else(EV("CPU_CLK_UNHALTED.THREAD_ANY", level) / 2) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = 1 /(self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV)) self.thresh = (val < 10) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.RETIRE_SLOTS:c1", level) def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = Instructions(self, EV, level) / (Retire_Fraction(self, EV, level) * EV("BR_MISP_EXEC.INDIRECT", level)) self.thresh = (val < 1000) return val # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / (EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level)) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("OFFCORE_REQUESTS.DEMAND_RFO", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.LLC_MISS", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_DURATION", level) + EV("DTLB_LOAD_MISSES.WALK_DURATION", level) + EV("DTLB_STORE_MISSES.WALK_DURATION", level)) / CORE_CLKS(self, EV, level) self.thresh = (val > 0.5) return val # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Measured Average Uncore Frequency for the SoC [GHz] def Uncore_Frequency(self, EV, level): return Socket_CLKS(self, EV, level) / 1e9 / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / (EV("CPU_CLK_UNHALTED.REF_XCLK_ANY", level) / 2) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return (64 *(EV("UNC_M_CAS_COUNT.RD", level) + EV("UNC_M_CAS_COUNT.WR", level)) / OneBillion) / Time(self, EV, level) # Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches. def MEM_Read_Latency(self, EV, level): return OneBillion *(EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182", level) / EV("UNC_C_TOR_INSERTS.MISS_OPCODE:Match=0x182", level)) / (Socket_CLKS(self, EV, level) / Time(self, EV, level)) # Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches def MEM_Parallel_Reads(self, EV, level): return EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182", level) / EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182:c1", level) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_C_CLOCKTICKS:one_unit", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['RS_EVENTS.EMPTY_END'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Pipeline_Width * Frontend_Latency_Cycles(self, EV, 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE.IFETCH_STALL", 3) / CLKS(self, EV, 3) - self.ITLB_Misses.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['ITLB_MISSES.WALK_COMPLETED'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = ITLB_Miss_Cycles(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = BAClear_Cost *(EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) + EV("MACHINE_CLEARS.COUNT", 3) + EV("BACLEARS.ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['IDQ.MS_SWITCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("IDQ.MS_SWITCHES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("ILD_STALL.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_MITE_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_MITE_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_DSB_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_DSB_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_ISSUED.ANY", 1) - Retired_Slots(self, EV, 1) + Pipeline_Width * Recovery_Cycles(self, EV, 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 2) * self.Bad_Speculation.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = 1 -(self.Frontend_Bound.compute(EV) + self.Bad_Speculation.compute(EV) + self.Retiring.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Memory_Bound_Fraction(self, EV, 2) * self.Backend_Bound.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L1_HIT:pp', 'MEM_LOAD_UOPS_RETIRED.HIT_LFB:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((STALLS_MEM_ANY(self, EV, 3) - EV("CYCLE_ACTIVITY.STALLS_L1D_PENDING", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT", 4) + EV("DTLB_LOAD_MISSES.WALK_DURATION", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.LOCK_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Lock_St_Fraction(self, EV, 4) * ORO_Demand_RFO_C1(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.NO_SR", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class G4K_Aliasing: name = "4K_Aliasing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("LD_BLOCKS_PARTIAL.ADDRESS_ALIAS", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "G4K_Aliasing zero division") return self.val desc = """ This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).. Consider reducing independent loads/stores accesses with 4K offsets. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("L1D_PEND_MISS.FB_FULL:c1", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("CYCLE_ACTIVITY.STALLS_L1D_PENDING", 3) - EV("CYCLE_ACTIVITY.STALLS_L2_PENDING", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = Mem_L3_Hit_Fraction(self, EV, 3) * EV("CYCLE_ACTIVITY.STALLS_L2_PENDING", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_HitM_Cost(self, EV, 4) * LOAD_XSNP_HITM(self, EV, 4) + Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_MISS(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_None_Cost(self, EV, 4) * LOAD_L3_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = SQ_Full_Cycles(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = (1 - Mem_L3_Hit_Fraction(self, EV, 3)) * EV("CYCLE_ACTIVITY.STALLS_L2_PENDING", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Local_MEM: name = "Local_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Server']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Local_DRAM_Cost(self, EV, 5) * LOAD_LCL_MEM(self, EV, 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Local_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance.""" class Remote_MEM: name = "Remote_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_DRAM:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Remote_DRAM_Cost(self, EV, 5) * LOAD_RMT_MEM(self, EV, 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations.""" class Remote_Cache: name = "Remote_Cache" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM:pp', 'MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_FWD:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore', 'Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Remote_HitM_Cost(self, EV, 5) * LOAD_RMT_HITM(self, EV, 5) + Mem_Remote_Fwd_Cost(self, EV, 5) * LOAD_RMT_FWD(self, EV, 5)) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_Cache zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations.""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_UOPS_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SB", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_LLC_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_UOPS_LLC_MISS_RETIRED.REMOTE_HITM:pp', 'OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE', 'OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Remote_HitM_Cost(self, EV, 4) * EV("OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM", 4) + Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 2 * EV("MEM_UOPS_RETIRED.SPLIT_STORES", 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_STORE_MISSES.STLB_HIT", 4) + EV("DTLB_STORE_MISSES.WALK_DURATION", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.FPU_DIV_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.FPU_DIV_ACTIVE", 3) / CORE_CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = (Backend_Bound_Cycles(self, EV, 3) - EV("RESOURCE_STALLS.SB", 3) - STALLS_MEM_ANY(self, EV, 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_0_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_1_Port_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_2_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_3m_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_0", 5) + EV("UOPS_DISPATCHED_PORT.PORT_1", 5) + EV("UOPS_DISPATCHED_PORT.PORT_5", 5)) / (3 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_5: name = "Port_5" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_5'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_5", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_5 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 Branches and ALU. See section 'Handling Port 5 Pressure' in Optimization Manual:. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_2", 5) + EV("UOPS_DISPATCHED_PORT.PORT_3", 5) - EV("UOPS_DISPATCHED_PORT.PORT_4", 5)) / (2 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Port_2: name = "Port_2" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_2'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_2", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_2 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 Loads and Store-address""" class Port_3: name = "Port_3" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_3'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_3", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_3 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 Loads and Store-address""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Port_4: name = "Port_4" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_4'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_4 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data)""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.RETIRE_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. . Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 4) * EV("FP_COMP_OPS_EXE.X87", 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", 5) + EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", 5)) / EV("UOPS_EXECUTED.THREAD", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("SIMD_FP_256.PACKED_DOUBLE", 5) + EV("SIMD_FP_256.PACKED_SINGLE", 5)) / EV("UOPS_EXECUTED.THREAD", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Microcode_Sequencer.compute(EV) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['IDQ.MS_UOPS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = Retire_Fraction(self, EV, 3) * EV("IDQ.MS_UOPS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['OTHER_ASSISTS.ANY_WB_ASSIST'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("OTHER_ASSISTS.ANY_WB_ASSIST", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_Uncore_Frequency: name = "Uncore_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Uncore_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Uncore_Frequency zero division") desc = """ Measured Average Uncore Frequency for the SoC [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_MEM_Read_Latency: name = "MEM_Read_Latency" domain = "NanoSeconds" maxval = 1000 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryLat', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Read_Latency zero division") desc = """ Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches.""" class Metric_MEM_Parallel_Reads: name = "MEM_Parallel_Reads" domain = "SystemMetric" maxval = 100 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Parallel_Reads(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Parallel_Reads zero division") desc = """ Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = G4K_Aliasing() ; r.run(n) ; o["G4K_Aliasing"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Local_MEM() ; r.run(n) ; o["Local_MEM"] = n n = Remote_MEM() ; r.run(n) ; o["Remote_MEM"] = n n = Remote_Cache() ; r.run(n) ; o["Remote_Cache"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_5() ; r.run(n) ; o["Port_5"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Port_2() ; r.run(n) ; o["Port_2"] = n n = Port_3() ; r.run(n) ; o["Port_3"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Port_4() ; r.run(n) ; o["Port_4"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["DSB"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["G4K_Aliasing"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Local_MEM"].parent = o["MEM_Latency"] o["Remote_MEM"].parent = o["MEM_Latency"] o["Remote_Cache"].parent = o["MEM_Latency"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_5"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_2"].parent = o["Load_Op_Utilization"] o["Port_3"].parent = o["Load_Op_Utilization"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_4"].parent = o["Store_Op_Utilization"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["Heavy_Operations"].parent = o["Retiring"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_Uncore_Frequency() ; r.metric(n) ; o["Uncore_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_MEM_Read_Latency() ; r.metric(n) ; o["MEM_Read_Latency"] = n n = Metric_MEM_Parallel_Reads() ; r.metric(n) ; o["MEM_Parallel_Reads"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["ICache_Misses"].ITLB_Misses = o["ITLB_Misses"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Branch_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Backend_Bound"].Retiring = o["Retiring"] o["Backend_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Backend_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Retiring = o["Retiring"] o["Memory_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Backend_Bound = o["Backend_Bound"] o["Memory_Bound"].Fetch_Latency = o["Fetch_Latency"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Core_Bound"].Retiring = o["Retiring"] o["Core_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Core_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Core_Bound"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilization"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilized_0"].Fetch_Latency = o["Fetch_Latency"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["Heavy_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] o["IpArith"].FP_Vector = o["FP_Vector"] o["IpArith"].FP_Scalar = o["FP_Scalar"] # siblings cross-tree o["MS_Switches"].sibling = (o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Machine_Clears"].sibling = (o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Remote_Cache"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Remote_Cache"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"],) o["Split_Stores"].sibling = (o["Port_4"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_5"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_4"].sibling = (o["Split_Stores"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"],) o["Microcode_Sequencer"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
133,732
Python
.py
3,304
34.814165
306
0.656987
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,879
pci.py
andikleen_pmu-tools/pci.py
#!/usr/bin/env python3 # library and tool to access PCI config space import os import struct # no multiple domains, controllers so far def probe(bus, dev, func): fn = "/sys/devices/pci0000:%02x/0000:%02x:%02x.%01x/config" % (bus, bus, dev, func) return os.path.isfile(fn) def openpci(bus, dev, func, offset, mode): fn = "/sys/devices/pci0000:%02x/0000:%02x:%02x.%01x/config" % (bus, bus, dev, func) f = os.open(fn, mode) os.lseek(f, offset, os.SEEK_SET) return f sizes = {8: "Q", 4: "I", 2: "H", 1: "B"} def writepci(bus, device, func, offset, size, val): f = openpci(bus, device, func, offset, os.O_WRONLY) os.write(f, struct.pack(sizes[size], val)) os.close(f) def readpci(bus, device, func, offset, size): f = openpci(bus, device, func, offset, os.O_RDONLY) v = struct.unpack(sizes[size], os.read(f, size))[0] os.close(f) return v def changebit(bus, device, func, offset, bit, val): f = openpci(bus, device, func, offset, os.O_RDWR) v = struct.unpack("I", os.read(f, 4))[0] if val: v = v | (1 << bit) else: v = v & ~(1 << bit) os.lseek(f, offset, os.SEEK_SET) os.write(f, struct.pack('I', v)) os.close(f)
1,211
Python
.py
33
32.575758
87
0.625107
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,880
tl_stat.py
andikleen_pmu-tools/tl_stat.py
# Copyright (c) 2012-2020, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. # # Maintain error data on perf measurements from __future__ import print_function import math from collections import namedtuple from tl_io import warn, warn_test, inform ValStat = namedtuple('ValStat', ['stddev', 'multiplex']) def geoadd(l): return math.sqrt(sum([x**2 for x in l])) # use geomean of stddevs and minimum of multiplex ratios for combining # XXX better way to combine multiplex ratios? def combine_valstat(l): if not l: return [] return ValStat(geoadd([x.stddev for x in l]), min([x.multiplex for x in l])) class ComputeStat: """Maintain statistics on measurement data.""" def __init__(self, quiet): self.referenced = set() self.already_warned = set() self.errcount = 0 self.errors = set() self.prev_errors = set() self.mismeasured = set() self.prev_mismeasured = set() self.quiet = quiet def referenced_check(self, res, evnum): referenced = self.referenced referenced = referenced - self.already_warned if not referenced: return self.already_warned |= referenced # sanity check: did we reference all results? if len(res.keys()) > 0: r = res[list(res.keys())[0]] if len(r) != len(evnum): warn("results len %d does not match event len %d" % (len(r), len(evnum))) return if len(referenced) != len(r): dummies = {i for i, d in enumerate(evnum) if d == "dummy"} notr = set(range(len(r))) - referenced - dummies if notr: warn_test("%d results not referenced: " % (len(notr)) + " ".join(["%d" % x for x in sorted(notr)])) def compute_errors(self): if self.errcount > 0 and self.errors != self.prev_errors: inform(("%d nodes had zero counts: " % (self.errcount)) + " ".join(sorted(self.errors))) self.errcount = 0 self.prev_errors = self.errors self.errors = set() if self.mismeasured and self.mismeasured > self.prev_mismeasured: inform("Mismeasured (out of bound values):" + " ".join(sorted(self.mismeasured))) self.prev_mismeasured = self.mismeasured
2,835
Python
.py
66
34.924242
89
0.623823
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,881
icl_client_ratios.py
andikleen_pmu-tools/icl_client_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel 10th gen Core (code name Icelake) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 topdown_use_fixed = False def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 10 Mem_L2_Store_Cost = 10 Mem_STLB_Hit_Cost = 7 BAClear_Cost = 10 MS_Switches_Cost = 3 Avg_Assist_Cost = 34 Pipeline_Width = 5 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 61 PERF_METRICS_MSR = 1 DS = 0 # Aux. formulas def Backend_Bound_Cycles(self, EV, level): return EV("CYCLE_ACTIVITY.STALLS_TOTAL", level) + Few_Uops_Executed_Threshold(self, EV, level) + EV("EXE_ACTIVITY.BOUND_ON_STORES", level) def Br_DoI_Jumps(self, EV, level): return EV("BR_INST_RETIRED.NEAR_TAKEN", level) - EV("BR_INST_RETIRED.COND_TAKEN", level) - 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) def Branching_Retired(self, EV, level): return (EV("BR_INST_RETIRED.ALL_BRANCHES", level) + 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("INST_RETIRED.NOP", level)) / SLOTS(self, EV, level) def Serialize_Core(self, EV, level): return self.Core_Bound.compute(EV) * (self.Serializing_Operation.compute(EV) + self.Core_Bound.compute(EV) * EV("RS_EVENTS.EMPTY_CYCLES", level) / CLKS(self, EV, level) * self.Ports_Utilized_0.compute(EV)) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV)) def Umisp(self, EV, level): return 10 * self.Microcode_Sequencer.compute(EV) * self.Other_Mispredicts.compute(EV) / self.Branch_Mispredicts.compute(EV) def Assist(self, EV, level): return (self.Microcode_Sequencer.compute(EV) / (self.Microcode_Sequencer.compute(EV) + self.Few_Uops_Instructions.compute(EV))) * (self.Assists.compute(EV) / self.Microcode_Sequencer.compute(EV)) def Assist_Frontend(self, EV, level): return Assist(self, EV, level) * self.Fetch_Latency.compute(EV) * (self.MS_Switches.compute(EV) + self.Branch_Resteers.compute(EV) * (self.Clears_Resteers.compute(EV) + self.Mispredicts_Resteers.compute(EV) * Umisp(self, EV, level)) / (self.Clears_Resteers.compute(EV) + self.Unknown_Branches.compute(EV) + self.Mispredicts_Resteers.compute(EV))) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) def Assist_Retired(self, EV, level): return Assist(self, EV, level) * self.Heavy_Operations.compute(EV) def Core_Bound_Cycles(self, EV, level): return self.Ports_Utilized_0.compute(EV) * CLKS(self, EV, level) + Few_Uops_Executed_Threshold(self, EV, level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE_CYCLES_GE_1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.THREAD:c1", level) # factor used for metrics associating fixed costs for FB Hits - according to probability theory if all FB Hits come at a random rate in original L1_Miss cost interval then the average cost for each one is 0.5 of the fixed cost def FB_Factor(self, EV, level): return 1 + FBHit_per_L1Miss(self, EV, level) / 2 def FBHit_per_L1Miss(self, EV, level): return EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("MEM_LOAD_RETIRED.L1_MISS", level) def Fetched_Uops(self, EV, level): return EV("UOPS_ISSUED.ANY", level) def Few_Uops_Executed_Threshold(self, EV, level): return EV("EXE_ACTIVITY.1_PORTS_UTIL", level) + self.Retiring.compute(EV) * EV("EXE_ACTIVITY.2_PORTS_UTIL", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) + 2 * EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 * EV("FP_ARITH_INST_RETIRED.8_FLOPS", level) + 16 * EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_ARITH_INST_RETIRED.VECTOR", level) def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def L2_Bound_Ratio(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", level) - EV("CYCLE_ACTIVITY.STALLS_L2_MISS", level)) / CLKS(self, EV, level) def Light_Ops_Sum(self, EV, level): return self.FP_Arith.compute(EV) + self.Memory_Operations.compute(EV) + self.Branch_Instructions.compute(EV) def LOAD_L2_HIT(self, EV, level): return EV("MEM_LOAD_RETIRED.L2_HIT", level) * (1 + FBHit_per_L1Miss(self, EV, level)) def LOAD_L3_HIT(self, EV, level): return EV("MEM_LOAD_RETIRED.L3_HIT", level) * FB_Factor(self, EV, level) def LOAD_XSNP_HIT(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT", level) def LOAD_XSNP_HITM(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM", level) def LOAD_XSNP_MISS(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", level) def MEM_Bound_Ratio(self, EV, level): return EV("CYCLE_ACTIVITY.STALLS_L3_MISS", level) / CLKS(self, EV, level) + L2_Bound_Ratio(self, EV, level) - self.L2_Bound.compute(EV) def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_INST_RETIRED.LOCK_LOADS", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) def Memory_Bound_Fraction(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", level) + EV("EXE_ACTIVITY.BOUND_ON_STORES", level)) / Backend_Bound_Cycles(self, EV, level) def Mispred_Clears_Fraction(self, EV, level): return EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c4", level)) , level ) def Store_L2_Hit_Cycles(self, EV, level): return EV("L2_RQSTS.RFO_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 32.5 * Core_Frequency(self, EV, level) def Mem_XSNP_Hit_Cost(self, EV, level): return 27 * Core_Frequency(self, EV, level) def Mem_XSNP_None_Cost(self, EV, level): return 12.5 * Core_Frequency(self, EV, level) def Mem_L2_Hit_Cost(self, EV, level): return 3.5 * Core_Frequency(self, EV, level) def PERF_METRICS_SUM(self, EV, level): return (EV("PERF_METRICS.FRONTEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BAD_SPECULATION", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.RETIRING", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BACKEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) def Retire_Fraction(self, EV, level): return EV("UOPS_RETIRED.SLOTS", level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return self.Retiring.compute(EV) * SLOTS(self, EV, level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return 8 if smt_enabled else 4 # A system parameter for dependent-loads (pointer chasing like access pattern) of the workload. An integer fraction in range from 0 (no dependent loads) to 100 (all loads are dependent loads) def Dependent_Loads_Weight(self, EV, level): return 20 # Total pipeline cost of Branch Misprediction related bottlenecks def Mispredictions(self, EV, level): val = 100 *(1 - Umisp(self, EV, level)) * (self.Branch_Mispredicts.compute(EV) + self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses) def Big_Code(self, EV, level): val = 100 * self.Fetch_Latency.compute(EV) * (self.ITLB_Misses.compute(EV) + self.ICache_Misses.compute(EV) + self.Unknown_Branches.compute(EV)) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end) def Instruction_Fetch_BW(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) - (1 - Umisp(self, EV, level)) * self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) - Assist_Frontend(self, EV, level)) - Big_Code(self, EV, level) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks def Cache_Memory_Bandwidth(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.MEM_Bandwidth.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.SQ_Full.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.FB_Full.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.G4K_Aliasing.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Latency related bottlenecks def Cache_Memory_Latency(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.MEM_Latency.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.L3_Hit_Latency.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * self.L2_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.Store_Latency.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.L1_Hit_Latency.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.G4K_Aliasing.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs) def Memory_Data_TLBs(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / max(self.Memory_Bound.compute(EV) , (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV)))) * (self.DTLB_Load.compute(EV) / max(self.L1_Bound.compute(EV) , (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.G4K_Aliasing.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.DTLB_Store.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors) def Memory_Synchronization(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * ((self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.Contested_Accesses.compute(EV) + self.Data_Sharing.compute(EV)) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)) + (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * self.False_Sharing.compute(EV) / ((self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)) - self.Store_Latency.compute(EV))) + self.Machine_Clears.compute(EV) * (1 - self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV)))) self.thresh = (val > 10) return val # Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy. def Compute_Bound_Est(self, EV, level): val = 100 *((self.Core_Bound.compute(EV) * self.Divider.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV))) + (self.Core_Bound.compute(EV) * (self.Ports_Utilization.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV))) * (self.Ports_Utilized_3m.compute(EV) / (self.Ports_Utilized_0.compute(EV) + self.Ports_Utilized_1.compute(EV) + self.Ports_Utilized_2.compute(EV) + self.Ports_Utilized_3m.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments) def Irregular_Overhead(self, EV, level): val = 100 *(Assist_Frontend(self, EV, level) + Umisp(self, EV, level) * self.Branch_Mispredicts.compute(EV) + (self.Machine_Clears.compute(EV) * self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV))) + Serialize_Core(self, EV, level) + Assist_Retired(self, EV, level)) self.thresh = (val > 10) return val # Total pipeline cost of remaining bottlenecks in the back-end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls. def Other_Bottlenecks(self, EV, level): val = 100 -(Big_Code(self, EV, level) + Instruction_Fetch_BW(self, EV, level) + Mispredictions(self, EV, level) + Cache_Memory_Bandwidth(self, EV, level) + Cache_Memory_Latency(self, EV, level) + Memory_Data_TLBs(self, EV, level) + Memory_Synchronization(self, EV, level) + Compute_Bound_Est(self, EV, level) + Irregular_Overhead(self, EV, level) + Branching_Overhead(self, EV, level) + Useful_Work(self, EV, level)) self.thresh = (val > 20) return val # Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations def Branching_Overhead(self, EV, level): val = 100 * Branching_Retired(self, EV, level) self.thresh = (val > 5) return val # Total pipeline cost of "useful operations" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead. def Useful_Work(self, EV, level): val = 100 *(self.Retiring.compute(EV) - Branching_Retired(self, EV, level) - Assist_Retired(self, EV, level)) self.thresh = (val > 20) return val # Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled def Core_Bound_Likely(self, EV, level): val = 100 *(1 - self.Core_Bound.compute(EV) / self.Ports_Utilization.compute(EV) if self.Core_Bound.compute(EV)< self.Ports_Utilization.compute(EV) else 1) if SMT_2T_Utilization(self, EV, level)> 0.5 else 0 self.thresh = (val > 0.5) return val # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return EV("TOPDOWN.SLOTS", level) if topdown_use_fixed else EV("TOPDOWN.SLOTS", level) # Fraction of Physical Core issue-slots utilized by this Logical Processor def Slots_Utilization(self, EV, level): return SLOTS(self, EV, level) / (EV("TOPDOWN.SLOTS:percore", level) / 2) if smt_enabled else 1 # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common. def FP_Arith_Utilization(self, EV, level): return (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) / (2 * CORE_CLKS(self, EV, level)) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # uops Executed per Cycle def EPC(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / CLKS(self, EV, level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.DISTRIBUTED", level) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408 def IpFLOP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / FLOP_Count(self, EV, level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_SP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_SINGLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_DP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX128(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX256(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX512(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per PAUSE (lower number means higher occurrence rate) def IpPause(self, EV, level): return Instructions(self, EV, level) / EV("MISC_RETIRED.PAUSE_INST", level) # Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate) def IpSWPF(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("SW_PREFETCH_ACCESS.T0:u0xF", level) self.thresh = (val < 100) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.SLOTS:c1", level) # Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate) def IpAssist(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("ASSISTS.ANY", level) self.thresh = (val < 100000) return val def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Average number of uops fetched from LSD per cycle def Fetch_LSD(self, EV, level): return EV("LSD.UOPS", level) / EV("LSD.CYCLES_ACTIVE", level) # Average number of uops fetched from DSB per cycle def Fetch_DSB(self, EV, level): return EV("IDQ.DSB_UOPS", level) / EV("IDQ.DSB_CYCLES_ANY", level) # Average number of uops fetched from MITE per cycle def Fetch_MITE(self, EV, level): return EV("IDQ.MITE_UOPS", level) / EV("IDQ.MITE_CYCLES_ANY", level) # Average number of Uops issued by front-end when it issued something def Fetch_UpC(self, EV, level): return EV("UOPS_ISSUED.ANY", level) / EV("UOPS_ISSUED.ANY:c1", level) # Fraction of Uops delivered by the LSD (Loop Stream Detector; aka Loop Cache) def LSD_Coverage(self, EV, level): return EV("LSD.UOPS", level) / Fetched_Uops(self, EV, level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details. def DSB_Switch_Cost(self, EV, level): return EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", level) / EV("DSB2MITE_SWITCHES.PENALTY_CYCLES:c1:e1", level) # Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.DSB_Switches.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) + self.Fetch_Bandwidth.compute(EV) * self.MITE.compute(EV) / (self.LSD.compute(EV) + self.MITE.compute(EV) + self.DSB.compute(EV))) self.thresh = (val > 10) return val # Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Bandwidth(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) * (self.Fetch_Bandwidth.compute(EV) / (self.Fetch_Bandwidth.compute(EV) + self.Fetch_Latency.compute(EV))) * (self.DSB.compute(EV) / (self.LSD.compute(EV) + self.MITE.compute(EV) + self.DSB.compute(EV)))) self.thresh = (val > 10) return val # Average Latency for L1 instruction cache misses def ICache_Miss_Latency(self, EV, level): return EV("ICACHE_16B.IFDATA_STALL", level) / EV("ICACHE_16B.IFDATA_STALL:c1:e1", level) # Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck. def IC_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.ICache_Misses.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 5) return val # Instructions per non-speculative DSB miss (lower number means higher occurrence rate) def IpDSB_Miss_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FRONTEND_RETIRED.ANY_DSB_MISS", level) self.thresh = (val < 50) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # L2 cache true code cacheline misses per kilo instruction def L2MPKI_Code(self, EV, level): return 1000 * EV("FRONTEND_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache speculative code cacheline misses per kilo instruction def L2MPKI_Code_All(self, EV, level): return 1000 * EV("L2_RQSTS.CODE_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional non-taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Ntaken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_NTAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Taken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_TAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate). def IpMisp_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.RET", level) self.thresh = (val < 500) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.INDIRECT", level) self.thresh = (val < 1000) return val # Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear) def Branch_Misprediction_Cost(self, EV, level): return Mispredictions(self, EV, level) * SLOTS(self, EV, level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / 100 # Speculative to Retired ratio of all clears (covering Mispredicts and nukes) def Spec_Clears_Ratio(self, EV, level): return EV("INT_MISC.CLEARS_COUNT", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) # Fraction of branches that are non-taken conditionals def Cond_NT(self, EV, level): return EV("BR_INST_RETIRED.COND_NTAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are taken conditionals def Cond_TK(self, EV, level): return EV("BR_INST_RETIRED.COND_TAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are CALL or RET def CallRet(self, EV, level): return (EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("BR_INST_RETIRED.NEAR_RETURN", level)) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are unconditional (direct or indirect) jumps def Jump(self, EV, level): return Br_DoI_Jumps(self, EV, level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches of other types (not individually covered by other metrics in Info.Branches group) def Other_Branches(self, EV, level): return 1 -(Cond_NT(self, EV, level) + Cond_TK(self, EV, level) + CallRet(self, EV, level) + Jump(self, EV, level)) # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / (EV("MEM_LOAD_RETIRED.L1_MISS", level) + EV("MEM_LOAD_RETIRED.FB_HIT", level)) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L1 cache true misses per kilo instruction for all demand loads (including speculative) def L1MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.ALL_DEMAND_DATA_RD", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache misses per kilo instruction for all request types (including speculative) def L2MPKI_All(self, EV, level): return 1000 *((EV("OFFCORE_REQUESTS.ALL_DATA_RD", level) - EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level)) + EV("L2_RQSTS.ALL_DEMAND_MISS", level) + EV("L2_RQSTS.SWPF_MISS", level)) / Instructions(self, EV, level) # L2 cache misses per kilo instruction for all demand loads (including speculative) def L2MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("L2_RQSTS.RFO_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all demand loads (including speculative) def L2HPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_HIT", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L3_MISS", level) / EV("INST_RETIRED.ANY", level) # Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries) def FB_HPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) def L3_Cache_Access_BW(self, EV, level): return 64 * EV("OFFCORE_REQUESTS.ALL_REQUESTS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_PENDING", level) + EV("DTLB_LOAD_MISSES.WALK_PENDING", level) + EV("DTLB_STORE_MISSES.WALK_PENDING", level)) / (2 * CORE_CLKS(self, EV, level)) self.thresh = (val > 0.5) return val # STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Code_STLB_MPKI(self, EV, level): return 1000 * EV("ITLB_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Load_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_LOAD_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Store_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_STORE_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average per-core data access bandwidth to the L3 cache [GB / sec] def L3_Cache_Access_BW_2T(self, EV, level): return L3_Cache_Access_BW(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Latency for L3 cache miss demand Loads def Load_L3_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD:u0x10", level) / EV("OFFCORE_REQUESTS.L3_MISS_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD:c1", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Un-cacheable retired load per kilo instruction def UC_Load_PKI(self, EV, level): return 1000 * EV("MEM_LOAD_MISC_RETIRED.UC", level) / EV("INST_RETIRED.ANY", level) # "Bus lock" per kilo instruction def Bus_Lock_PKI(self, EV, level): return 1000 * EV("SQ_MISC.BUS_LOCK", level) / EV("INST_RETIRED.ANY", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes. def Power_License0_Utilization(self, EV, level): return EV("CORE_POWER.LVL0_TURBO_LICENSE", level) / CORE_CLKS(self, EV, level) # Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions. def Power_License1_Utilization(self, EV, level): val = EV("CORE_POWER.LVL1_TURBO_LICENSE", level) / CORE_CLKS(self, EV, level) self.thresh = (val > 0.5) return val # Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions. def Power_License2_Utilization(self, EV, level): val = EV("CORE_POWER.LVL2_TURBO_LICENSE", level) / CORE_CLKS(self, EV, level) self.thresh = (val > 0.5) return val # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_DISTRIBUTED", level) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return 64 *(EV("UNC_ARB_TRK_REQUESTS.ALL", level) + EV("UNC_ARB_COH_TRK_REQUESTS.ALL", level)) / OneMillion / Time(self, EV, level) / 1000 # Total package Power in Watts def Power(self, EV, level): return EV("UNC_PKG_ENERGY_STATUS", level) * Energy_Unit /(Time(self, EV, level) * OneMillion ) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_CLOCK.SOCKET", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_4:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.FRONTEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) - EV("INT_MISC.UOP_DROPPING", 1) / SLOTS(self, EV, 1) if topdown_use_fixed else(EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) - EV("INT_MISC.UOP_DROPPING", 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_16:pp', 'FRONTEND_RETIRED.LATENCY_GE_8:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (Pipeline_Width * EV("IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE", 2) - EV("INT_MISC.UOP_DROPPING", 2)) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.L2_MISS:pp', 'FRONTEND_RETIRED.L1I_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_DATA.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.STLB_MISS:pp', 'FRONTEND_RETIRED.ITLB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_TAG.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.CLEAR_RESTEER_CYCLES", 3) / CLKS(self, EV, 3) + self.Unknown_Branches.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class Mispredicts_Resteers: name = "Mispredicts_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 4) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Mispredicts_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage.""" class Clears_Resteers: name = "Clears_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'MachineClears']) maxval = None def compute(self, EV): try: self.val = (1 - Mispred_Clears_Fraction(self, EV, 4)) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Clears_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears.""" class Unknown_Branches: name = "Unknown_Branches" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['BACLEARS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = BAClear_Cost * EV("BACLEARS.ANY", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Unknown_Branches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (e.g. first time the branch is fetched or hitting BTB capacity limit) hence called Unknown Branches""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['IDQ.MS_SWITCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("IDQ.MS_SWITCHES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DECODE.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.DSB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_2:pp'] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.ANY_DSB_MISS'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.MITE_CYCLES_ANY", 3) - EV("IDQ.MITE_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Decoder0_Alone: name = "Decoder0_Alone" domain = "Slots_Estimated" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("INST_DECODED.DECODERS:c1", 4) - EV("INST_DECODED.DECODERS:c2", 4)) / CORE_CLKS(self, EV, 4) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Decoder0_Alone zero division") return self.val desc = """ This metric represents fraction of cycles where decoder-0 was the only active decoder""" class MITE_4wide: name = "MITE_4wide" domain = "Core_Clocks" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.MITE_UOPS:c4", 4) - EV("IDQ.MITE_UOPS:c5", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE_4wide zero division") return self.val desc = """ This metric represents fraction of cycles where (only) 4 uops were delivered by the MITE pipeline""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.DSB_CYCLES_ANY", 3) - EV("IDQ.DSB_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class LSD: name = "LSD" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'LSD']) maxval = None def compute(self, EV): try: self.val = (EV("LSD.CYCLES_ACTIVE", 3) - EV("LSD.CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LSD zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = max(1 -(self.Frontend_Bound.compute(EV) + self.Backend_Bound.compute(EV) + self.Retiring.compute(EV)) , 0 ) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 2) * self.Bad_Speculation.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Other_Mispredicts: name = "Other_Mispredicts" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'BrMispredicts']) maxval = None def compute(self, EV): try: self.val = max(self.Branch_Mispredicts.compute(EV) * (1 - EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) / (EV("INT_MISC.CLEARS_COUNT", 3) - EV("MACHINE_CLEARS.COUNT", 3))) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Mispredicts zero division") return self.val desc = """ This metric estimates fraction of slots the CPU was stalled due to other cases of misprediction (non-retired x86 branches or other types).""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Other_Nukes: name = "Other_Nukes" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'Machine_Clears']) maxval = None def compute(self, EV): try: self.val = max(self.Machine_Clears.compute(EV) * (1 - EV("MACHINE_CLEARS.MEMORY_ORDERING", 3) / EV("MACHINE_CLEARS.COUNT", 3)) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Nukes zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Nukes (Machine Clears) not related to memory ordering.""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = ['TOPDOWN.BACKEND_BOUND_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.BACKEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) + (Pipeline_Width * EV("INT_MISC.CLEARS_COUNT", 1)) / SLOTS(self, EV, 1) if topdown_use_fixed else(EV("TOPDOWN.BACKEND_BOUND_SLOTS", 1) + Pipeline_Width * EV("INT_MISC.CLEARS_COUNT", 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Memory_Bound_Fraction(self, EV, 2) * self.Backend_Bound.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT:pp', 'MEM_LOAD_RETIRED.FB_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3) - EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = min(Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT:c1", 4) + EV("DTLB_LOAD_MISSES.WALK_ACTIVE", 4) , max(EV("CYCLE_ACTIVITY.CYCLES_MEM_ANY", 4) - EV("CYCLE_ACTIVITY.CYCLES_L1D_MISS", 4) , 0)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Load_STLB_Hit: name = "Load_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = self.DTLB_Load.compute(EV) - self.Load_STLB_Miss.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)""" class Load_STLB_Miss: name = "Load_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_LOAD_MISSES.WALK_ACTIVE", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class L1_Hit_Latency: name = "L1_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = min(2 *(EV("MEM_INST_RETIRED.ALL_LOADS", 4) - EV("MEM_LOAD_RETIRED.FB_HIT", 4) - EV("MEM_LOAD_RETIRED.L1_MISS", 4)) * Dependent_Loads_Weight(self, EV, 4) / 100 , max(EV("CYCLE_ACTIVITY.CYCLES_MEM_ANY", 4) - EV("CYCLE_ACTIVITY.CYCLES_L1D_MISS", 4) , 0)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Hit_Latency zero division") return self.val desc = """ This metric roughly estimates fraction of cycles with demand load accesses that hit the L1 cache. The short latency of the L1 data cache may be exposed in pointer-chasing memory access patterns as an example.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.LOCK_LOADS'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (16 * max(0 , EV("MEM_INST_RETIRED.LOCK_LOADS", 4) - EV("L2_RQSTS.ALL_RFO", 4)) + Mem_Lock_St_Fraction(self, EV, 4) * (Mem_L2_Store_Cost * EV("L2_RQSTS.RFO_HIT", 4) + ORO_Demand_RFO_C1(self, EV, 4))) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("LD_BLOCKS.NO_SR", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class G4K_Aliasing: name = "4K_Aliasing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("LD_BLOCKS_PARTIAL.ADDRESS_ALIAS", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "G4K_Aliasing zero division") return self.val desc = """ This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).. Consider reducing independent loads/stores accesses with 4K offsets. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = EV("L1D_PEND_MISS.FB_FULL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (LOAD_L2_HIT(self, EV, 3) / (LOAD_L2_HIT(self, EV, 3) + EV("L1D_PEND_MISS.FB_FULL_PERIODS", 3))) * L2_Bound_Ratio(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3) - EV("CYCLE_ACTIVITY.STALLS_L3_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = ((Mem_XSNP_HitM_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_HITM(self, EV, 4) + (Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_MISS(self, EV, 4)) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_HIT(self, EV, 4) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_None_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_L3_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = EV("L1D_PEND_MISS.L2_STALL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = MEM_Bound_Ratio(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_INST_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.BOUND_ON_STORES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.SPLIT_STORES", 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class Streaming_Stores: name = "Streaming_Stores" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.STREAMING_WR.ANY_RESPONSE'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBW', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = 9 * EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) / CLKS(self, EV, 4) if DS else 0 EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Streaming_Stores zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_STORE_MISSES.STLB_HIT:c1", 4) + EV("DTLB_STORE_MISSES.WALK_ACTIVE", 4)) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Store_STLB_Hit: name = "Store_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = self.DTLB_Store.compute(EV) - self.Store_STLB_Miss.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second- level TLB (STLB)""" class Store_STLB_Miss: name = "Store_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_STORE_MISSES.WALK_ACTIVE", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.DIVIDER_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.DIVIDER_ACTIVE", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Serializing_Operation: name = "Serializing_Operation" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['RESOURCE_STALLS.SCOREBOARD'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SCOREBOARD", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Serializing_Operation zero division") return self.val desc = """ This metric represents fraction of cycles the CPU issue- pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out- of-order execution which may limit performance.""" class Slow_Pause: name = "Slow_Pause" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['MISC_RETIRED.PAUSE_INST'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 140 * EV("MISC_RETIRED.PAUSE_INST", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Slow_Pause zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Core_Bound_Cycles(self, EV, 3) / CLKS(self, EV, 3) if (EV("ARITH.DIVIDER_ACTIVE", 3)<(EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) - EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3))) else Few_Uops_Executed_Threshold(self, EV, 3) / CLKS(self, EV, 3) EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3) EV("ARITH.DIVIDER_ACTIVE", 3) EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.3_PORTS_UTIL:u0x80", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Mixing_Vectors: name = "Mixing_Vectors" domain = "Clocks" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = EV("UOPS_ISSUED.VECTOR_WIDTH_MISMATCH", 5) / EV("UOPS_ISSUED.ANY", 5) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error(self, "Mixing_Vectors zero division") return self.val desc = """ This metric estimates penalty in terms of percentage of injected blend uops out of all Uops Issued -- the Count Domain. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.1_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.1_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.2_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.2_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['UOPS_EXECUTED.CYCLES_GE_3'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_EXECUTED.CYCLES_GE_3", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_0", 5) + EV("UOPS_DISPATCHED.PORT_1", 5) + EV("UOPS_DISPATCHED.PORT_5", 5) + EV("UOPS_DISPATCHED.PORT_6", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU and 2nd branch""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_5: name = "Port_5" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_5'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_5", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_5 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ALU. See section 'Handling Port 5 Pressure' in Optimization Manual:. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Port_6: name = "Port_6" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_6'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_6", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_6 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 Primary Branch and simple ALU""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_2_3'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_2_3", 5) / (2 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_7_8'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_4_9", 5) + EV("UOPS_DISPATCHED.PORT_7_8", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.RETIRING", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) if topdown_use_fixed else EV("UOPS_RETIRED.SLOTS", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. Note this may undercount due to approximation using indirect events. Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) * EV("UOPS_EXECUTED.X87", 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_512b: name = "FP_Vector_512b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_512b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.""" class Memory_Operations: name = "Memory_Operations" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("MEM_INST_RETIRED.ANY", 3) / EV("INST_RETIRED.ANY", 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.""" class Branch_Instructions: name = "Branch_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Branches', 'BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("BR_INST_RETIRED.ALL_BRANCHES", 3) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring branch instructions.""" class Other_Light_Ops: name = "Other_Light_Ops" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Light_Operations.compute(EV) - Light_Ops_Sum(self, EV, 3)) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Light_Ops zero division") return self.val desc = """ This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting""" class Nop_Instructions: name = "Nop_Instructions" domain = "Slots" area = "RET" level = 4 htoff = False sample = ['INST_RETIRED.NOP'] errcount = 0 sibling = None metricgroup = frozenset(['BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INST_RETIRED.NOP", 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Nop_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.. Improve Codegen by correctly placing NOPs outside hot sections (e.g. outside loop body).""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Microcode_Sequencer.compute(EV) + self.Retiring.compute(EV) * (EV("UOPS_DECODED.DEC0", 2) - EV("UOPS_DECODED.DEC0:c1", 2)) / EV("IDQ.MITE_UOPS", 2) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences. Note this may overcount due to approximation using indirect events""" class Few_Uops_Instructions: name = "Few_Uops_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = self.Heavy_Operations.compute(EV) - self.Microcode_Sequencer.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Few_Uops_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to four uops. This highly-correlates with the number of uops in such instructions.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['IDQ.MS_UOPS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = Retire_Fraction(self, EV, 3) * EV("IDQ.MS_UOPS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['ASSISTS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("ASSISTS.ANY", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class FP_Assists: name = "FP_Assists" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = 34 * EV("ASSISTS.FP", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "FP_Assists zero division") return self.val desc = """ This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists. FP Assist may apply when working with very small floating point values (so-called Denormals).. Consider DAZ (Denormals Are Zero) and/or FTZ (Flush To Zero) options in your compiler; \"-ffast-math\" with -O2 in GCC for example. This option may improve performance if the denormal values are not critical in your application. Also note that the DAZ and FTZ modes are not compatible with the IEEE Standard 754.. https://www.intel.com/content/www/us/en/develop/docume ntation/vtune-help/top/reference/cpu-metrics-reference/bad- speculation-back-end-bound-pipeline-slots/fp-assists.html""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_Mispredictions: name = "Mispredictions" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts', 'BvMP']) sibling = None def compute(self, EV): try: self.val = Mispredictions(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Mispredictions zero division") desc = """ Total pipeline cost of Branch Misprediction related bottlenecks""" class Metric_Big_Code: name = "Big_Code" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBC', 'BigFootprint', 'Fed', 'Frontend', 'IcMiss', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Big_Code(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Big_Code zero division") desc = """ Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)""" class Metric_Instruction_Fetch_BW: name = "Instruction_Fetch_BW" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvFB', 'Fed', 'FetchBW', 'Frontend']) sibling = None def compute(self, EV): try: self.val = Instruction_Fetch_BW(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Instruction_Fetch_BW zero division") desc = """ Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end)""" class Metric_Cache_Memory_Bandwidth: name = "Cache_Memory_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMB', 'Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Bandwidth(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Bandwidth zero division") desc = """ Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks""" class Metric_Cache_Memory_Latency: name = "Cache_Memory_Latency" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvML', 'Mem', 'MemoryLat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Latency(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Latency zero division") desc = """ Total pipeline cost of external Memory- or Cache-Latency related bottlenecks""" class Metric_Memory_Data_TLBs: name = "Memory_Data_TLBs" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMT', 'Mem', 'MemoryTLB', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Data_TLBs(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Memory_Data_TLBs zero division") desc = """ Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)""" class Metric_Memory_Synchronization: name = "Memory_Synchronization" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMS', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Synchronization(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Memory_Synchronization zero division") desc = """ Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors)""" class Metric_Compute_Bound_Est: name = "Compute_Bound_Est" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvCB', 'Cor']) sibling = None def compute(self, EV): try: self.val = Compute_Bound_Est(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Compute_Bound_Est zero division") desc = """ Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy.""" class Metric_Irregular_Overhead: name = "Irregular_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BvIO', 'Cor', 'Ret']) sibling = None def compute(self, EV): try: self.val = Irregular_Overhead(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Irregular_Overhead zero division") desc = """ Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments)""" class Metric_Other_Bottlenecks: name = "Other_Bottlenecks" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvOB', 'Cor', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Other_Bottlenecks(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Other_Bottlenecks zero division") desc = """ Total pipeline cost of remaining bottlenecks in the back- end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls.""" class Metric_Branching_Overhead: name = "Branching_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBO', 'Ret']) sibling = None def compute(self, EV): try: self.val = Branching_Overhead(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "Branching_Overhead zero division") desc = """ Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations""" class Metric_Useful_Work: name = "Useful_Work" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvUW', 'Ret']) sibling = None def compute(self, EV): try: self.val = Useful_Work(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Useful_Work zero division") desc = """ Total pipeline cost of \"useful operations\" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead.""" class Metric_Core_Bound_Likely: name = "Core_Bound_Likely" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Botlnk.L0" metricgroup = frozenset(['Cor', 'SMT']) sibling = None def compute(self, EV): try: self.val = Core_Bound_Likely(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Core_Bound_Likely zero division") desc = """ Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Slots_Utilization: name = "Slots_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Slots_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Slots_Utilization zero division") desc = """ Fraction of Physical Core issue-slots utilized by this Logical Processor""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_FP_Arith_Utilization: name = "FP_Arith_Utilization" domain = "Core_Metric" maxval = 2.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = FP_Arith_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FP_Arith_Utilization zero division") desc = """ Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common.""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_EPC: name = "EPC" domain = "Metric" maxval = 20.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = EPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "EPC zero division") desc = """ uops Executed per Cycle""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpFLOP: name = "IpFLOP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpFLOP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpFLOP zero division") desc = """ Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_IpArith_Scalar_SP: name = "IpArith_Scalar_SP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_SP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_SP zero division") desc = """ Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_DP: name = "IpArith_Scalar_DP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_DP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_DP zero division") desc = """ Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX128: name = "IpArith_AVX128" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX128(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX128 zero division") desc = """ Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX256: name = "IpArith_AVX256" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX256(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX256 zero division") desc = """ Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX512: name = "IpArith_AVX512" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX512(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX512 zero division") desc = """ Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpPause: name = "IpPause" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpPause(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpPause zero division") desc = """ Instructions per PAUSE (lower number means higher occurrence rate)""" class Metric_IpSWPF: name = "IpSWPF" domain = "Inst_Metric" maxval = 1000 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Prefetches']) sibling = None def compute(self, EV): try: self.val = IpSWPF(self, EV, 0) self.thresh = (self.val < 100) except ZeroDivisionError: handle_error_metric(self, "IpSWPF zero division") desc = """ Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate)""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_IpAssist: name = "IpAssist" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['MicroSeq', 'Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = IpAssist(self, EV, 0) self.thresh = (self.val < 100000) except ZeroDivisionError: handle_error_metric(self, "IpAssist zero division") desc = """ Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate)""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_Fetch_LSD: name = "Fetch_LSD" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_LSD(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_LSD zero division") desc = """ Average number of uops fetched from LSD per cycle""" class Metric_Fetch_DSB: name = "Fetch_DSB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_DSB(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_DSB zero division") desc = """ Average number of uops fetched from DSB per cycle""" class Metric_Fetch_MITE: name = "Fetch_MITE" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_MITE(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_MITE zero division") desc = """ Average number of uops fetched from MITE per cycle""" class Metric_Fetch_UpC: name = "Fetch_UpC" domain = "Metric" maxval = 6.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_UpC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_UpC zero division") desc = """ Average number of Uops issued by front-end when it issued something""" class Metric_LSD_Coverage: name = "LSD_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'LSD']) sibling = None def compute(self, EV): try: self.val = LSD_Coverage(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "LSD_Coverage zero division") desc = """ Fraction of Uops delivered by the LSD (Loop Stream Detector; aka Loop Cache)""" class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_DSB_Switch_Cost: name = "DSB_Switch_Cost" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss']) sibling = None def compute(self, EV): try: self.val = DSB_Switch_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DSB_Switch_Cost zero division") desc = """ Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details.""" class Metric_DSB_Misses: name = "DSB_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = DSB_Misses(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Misses zero division") desc = """ Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_DSB_Bandwidth: name = "DSB_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSB', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Bandwidth(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Bandwidth zero division") desc = """ Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_ICache_Miss_Latency: name = "ICache_Miss_Latency" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = ICache_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ICache_Miss_Latency zero division") desc = """ Average Latency for L1 instruction cache misses""" class Metric_IC_Misses: name = "IC_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = IC_Misses(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "IC_Misses zero division") desc = """ Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck.""" class Metric_IpDSB_Miss_Ret: name = "IpDSB_Miss_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = IpDSB_Miss_Ret(self, EV, 0) self.thresh = (self.val < 50) except ZeroDivisionError: handle_error_metric(self, "IpDSB_Miss_Ret zero division") desc = """ Instructions per non-speculative DSB miss (lower number means higher occurrence rate)""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_L2MPKI_Code: name = "L2MPKI_Code" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code zero division") desc = """ L2 cache true code cacheline misses per kilo instruction""" class Metric_L2MPKI_Code_All: name = "L2MPKI_Code_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code_All zero division") desc = """ L2 cache speculative code cacheline misses per kilo instruction""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Cond_Ntaken: name = "IpMisp_Cond_Ntaken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Ntaken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Ntaken zero division") desc = """ Instructions per retired Mispredicts for conditional non- taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Cond_Taken: name = "IpMisp_Cond_Taken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Taken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Taken zero division") desc = """ Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Ret: name = "IpMisp_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Ret(self, EV, 0) self.thresh = (self.val < 500) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Ret zero division") desc = """ Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Branch_Misprediction_Cost: name = "Branch_Misprediction_Cost" domain = "Core_Metric" maxval = 300 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Branch_Misprediction_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Branch_Misprediction_Cost zero division") desc = """ Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)""" class Metric_Spec_Clears_Ratio: name = "Spec_Clears_Ratio" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Spec_Clears_Ratio(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Spec_Clears_Ratio zero division") desc = """ Speculative to Retired ratio of all clears (covering Mispredicts and nukes)""" class Metric_Cond_NT: name = "Cond_NT" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_NT(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_NT zero division") desc = """ Fraction of branches that are non-taken conditionals""" class Metric_Cond_TK: name = "Cond_TK" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_TK(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_TK zero division") desc = """ Fraction of branches that are taken conditionals""" class Metric_CallRet: name = "CallRet" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = CallRet(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CallRet zero division") desc = """ Fraction of branches that are CALL or RET""" class Metric_Jump: name = "Jump" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Jump(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Jump zero division") desc = """ Fraction of branches that are unconditional (direct or indirect) jumps""" class Metric_Other_Branches: name = "Other_Branches" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Other_Branches(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Other_Branches zero division") desc = """ Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L1MPKI_Load: name = "L1MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI_Load zero division") desc = """ L1 cache true misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_All: name = "L2MPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_All zero division") desc = """ L2 cache misses per kilo instruction for all request types (including speculative)""" class Metric_L2MPKI_Load: name = "L2MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Load zero division") desc = """ L2 cache misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L2HPKI_Load: name = "L2HPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_Load zero division") desc = """ L2 cache hits per kilo instruction for all demand loads (including speculative)""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_FB_HPKI: name = "FB_HPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = FB_HPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FB_HPKI zero division") desc = """ Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss- handling entries)""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Access_BW: name = "L3_Cache_Access_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_Code_STLB_MPKI: name = "Code_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Fed', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Code_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Code_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Load_STLB_MPKI: name = "Load_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Load_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Store_STLB_MPKI: name = "Store_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Store_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Store_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_L3_Cache_Access_BW_2T: name = "L3_Cache_Access_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW_2T zero division") desc = """ Average per-core data access bandwidth to the L3 cache [GB / sec]""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L3_Miss_Latency: name = "Load_L3_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L3_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L3_Miss_Latency zero division") desc = """ Average Latency for L3 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_UC_Load_PKI: name = "UC_Load_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = UC_Load_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "UC_Load_PKI zero division") desc = """ Un-cacheable retired load per kilo instruction""" class Metric_Bus_Lock_PKI: name = "Bus_Lock_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = Bus_Lock_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Bus_Lock_PKI zero division") desc = """ \"Bus lock\" per kilo instruction""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_Power_License0_Utilization: name = "Power_License0_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Power_License0_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Power_License0_Utilization zero division") desc = """ Fraction of Core cycles where the core was running with power-delivery for baseline license level 0. This includes non-AVX codes, SSE, AVX 128-bit, and low-current AVX 256-bit codes.""" class Metric_Power_License1_Utilization: name = "Power_License1_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Power_License1_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Power_License1_Utilization zero division") desc = """ Fraction of Core cycles where the core was running with power-delivery for license level 1. This includes high current AVX 256-bit instructions as well as low current AVX 512-bit instructions.""" class Metric_Power_License2_Utilization: name = "Power_License2_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Power_License2_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Power_License2_Utilization zero division") desc = """ Fraction of Core cycles where the core was running with power-delivery for license level 2 (introduced in SKX). This includes high current AVX 512-bit instructions.""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_Power: name = "Power" domain = "SystemMetric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power', 'SoC']) sibling = None def compute(self, EV): try: self.val = Power(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Power zero division") desc = """ Total package Power in Watts""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = Mispredicts_Resteers() ; r.run(n) ; o["Mispredicts_Resteers"] = n n = Clears_Resteers() ; r.run(n) ; o["Clears_Resteers"] = n n = Unknown_Branches() ; r.run(n) ; o["Unknown_Branches"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = Decoder0_Alone() ; r.run(n) ; o["Decoder0_Alone"] = n n = MITE_4wide() ; r.run(n) ; o["MITE_4wide"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = LSD() ; r.run(n) ; o["LSD"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Other_Mispredicts() ; r.run(n) ; o["Other_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Other_Nukes() ; r.run(n) ; o["Other_Nukes"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Load_STLB_Hit() ; r.run(n) ; o["Load_STLB_Hit"] = n n = Load_STLB_Miss() ; r.run(n) ; o["Load_STLB_Miss"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = L1_Hit_Latency() ; r.run(n) ; o["L1_Hit_Latency"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = G4K_Aliasing() ; r.run(n) ; o["G4K_Aliasing"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = Streaming_Stores() ; r.run(n) ; o["Streaming_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Store_STLB_Hit() ; r.run(n) ; o["Store_STLB_Hit"] = n n = Store_STLB_Miss() ; r.run(n) ; o["Store_STLB_Miss"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Serializing_Operation() ; r.run(n) ; o["Serializing_Operation"] = n n = Slow_Pause() ; r.run(n) ; o["Slow_Pause"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Mixing_Vectors() ; r.run(n) ; o["Mixing_Vectors"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_5() ; r.run(n) ; o["Port_5"] = n n = Port_6() ; r.run(n) ; o["Port_6"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = FP_Vector_512b() ; r.run(n) ; o["FP_Vector_512b"] = n n = Memory_Operations() ; r.run(n) ; o["Memory_Operations"] = n n = Branch_Instructions() ; r.run(n) ; o["Branch_Instructions"] = n n = Other_Light_Ops() ; r.run(n) ; o["Other_Light_Ops"] = n n = Nop_Instructions() ; r.run(n) ; o["Nop_Instructions"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Few_Uops_Instructions() ; r.run(n) ; o["Few_Uops_Instructions"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = FP_Assists() ; r.run(n) ; o["FP_Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["Mispredicts_Resteers"].parent = o["Branch_Resteers"] o["Clears_Resteers"].parent = o["Branch_Resteers"] o["Unknown_Branches"].parent = o["Branch_Resteers"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["Decoder0_Alone"].parent = o["MITE"] o["MITE_4wide"].parent = o["MITE"] o["DSB"].parent = o["Fetch_Bandwidth"] o["LSD"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Other_Mispredicts"].parent = o["Branch_Mispredicts"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Other_Nukes"].parent = o["Machine_Clears"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Load_STLB_Hit"].parent = o["DTLB_Load"] o["Load_STLB_Miss"].parent = o["DTLB_Load"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["L1_Hit_Latency"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["G4K_Aliasing"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["Streaming_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Store_STLB_Hit"].parent = o["DTLB_Store"] o["Store_STLB_Miss"].parent = o["DTLB_Store"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Serializing_Operation"].parent = o["Core_Bound"] o["Slow_Pause"].parent = o["Serializing_Operation"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Mixing_Vectors"].parent = o["Ports_Utilized_0"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_5"].parent = o["ALU_Op_Utilization"] o["Port_6"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["FP_Vector_512b"].parent = o["FP_Vector"] o["Memory_Operations"].parent = o["Light_Operations"] o["Branch_Instructions"].parent = o["Light_Operations"] o["Other_Light_Ops"].parent = o["Light_Operations"] o["Nop_Instructions"].parent = o["Other_Light_Ops"] o["Heavy_Operations"].parent = o["Retiring"] o["Few_Uops_Instructions"].parent = o["Heavy_Operations"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["FP_Assists"].parent = o["Assists"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_Mispredictions() ; r.metric(n) ; o["Mispredictions"] = n n = Metric_Big_Code() ; r.metric(n) ; o["Big_Code"] = n n = Metric_Instruction_Fetch_BW() ; r.metric(n) ; o["Instruction_Fetch_BW"] = n n = Metric_Cache_Memory_Bandwidth() ; r.metric(n) ; o["Cache_Memory_Bandwidth"] = n n = Metric_Cache_Memory_Latency() ; r.metric(n) ; o["Cache_Memory_Latency"] = n n = Metric_Memory_Data_TLBs() ; r.metric(n) ; o["Memory_Data_TLBs"] = n n = Metric_Memory_Synchronization() ; r.metric(n) ; o["Memory_Synchronization"] = n n = Metric_Compute_Bound_Est() ; r.metric(n) ; o["Compute_Bound_Est"] = n n = Metric_Irregular_Overhead() ; r.metric(n) ; o["Irregular_Overhead"] = n n = Metric_Other_Bottlenecks() ; r.metric(n) ; o["Other_Bottlenecks"] = n n = Metric_Branching_Overhead() ; r.metric(n) ; o["Branching_Overhead"] = n n = Metric_Useful_Work() ; r.metric(n) ; o["Useful_Work"] = n n = Metric_Core_Bound_Likely() ; r.metric(n) ; o["Core_Bound_Likely"] = n n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Slots_Utilization() ; r.metric(n) ; o["Slots_Utilization"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_FP_Arith_Utilization() ; r.metric(n) ; o["FP_Arith_Utilization"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_EPC() ; r.metric(n) ; o["EPC"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpFLOP() ; r.metric(n) ; o["IpFLOP"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_IpArith_Scalar_SP() ; r.metric(n) ; o["IpArith_Scalar_SP"] = n n = Metric_IpArith_Scalar_DP() ; r.metric(n) ; o["IpArith_Scalar_DP"] = n n = Metric_IpArith_AVX128() ; r.metric(n) ; o["IpArith_AVX128"] = n n = Metric_IpArith_AVX256() ; r.metric(n) ; o["IpArith_AVX256"] = n n = Metric_IpArith_AVX512() ; r.metric(n) ; o["IpArith_AVX512"] = n n = Metric_IpPause() ; r.metric(n) ; o["IpPause"] = n n = Metric_IpSWPF() ; r.metric(n) ; o["IpSWPF"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_IpAssist() ; r.metric(n) ; o["IpAssist"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_Fetch_LSD() ; r.metric(n) ; o["Fetch_LSD"] = n n = Metric_Fetch_DSB() ; r.metric(n) ; o["Fetch_DSB"] = n n = Metric_Fetch_MITE() ; r.metric(n) ; o["Fetch_MITE"] = n n = Metric_Fetch_UpC() ; r.metric(n) ; o["Fetch_UpC"] = n n = Metric_LSD_Coverage() ; r.metric(n) ; o["LSD_Coverage"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_DSB_Switch_Cost() ; r.metric(n) ; o["DSB_Switch_Cost"] = n n = Metric_DSB_Misses() ; r.metric(n) ; o["DSB_Misses"] = n n = Metric_DSB_Bandwidth() ; r.metric(n) ; o["DSB_Bandwidth"] = n n = Metric_ICache_Miss_Latency() ; r.metric(n) ; o["ICache_Miss_Latency"] = n n = Metric_IC_Misses() ; r.metric(n) ; o["IC_Misses"] = n n = Metric_IpDSB_Miss_Ret() ; r.metric(n) ; o["IpDSB_Miss_Ret"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_L2MPKI_Code() ; r.metric(n) ; o["L2MPKI_Code"] = n n = Metric_L2MPKI_Code_All() ; r.metric(n) ; o["L2MPKI_Code_All"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Cond_Ntaken() ; r.metric(n) ; o["IpMisp_Cond_Ntaken"] = n n = Metric_IpMisp_Cond_Taken() ; r.metric(n) ; o["IpMisp_Cond_Taken"] = n n = Metric_IpMisp_Ret() ; r.metric(n) ; o["IpMisp_Ret"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Branch_Misprediction_Cost() ; r.metric(n) ; o["Branch_Misprediction_Cost"] = n n = Metric_Spec_Clears_Ratio() ; r.metric(n) ; o["Spec_Clears_Ratio"] = n n = Metric_Cond_NT() ; r.metric(n) ; o["Cond_NT"] = n n = Metric_Cond_TK() ; r.metric(n) ; o["Cond_TK"] = n n = Metric_CallRet() ; r.metric(n) ; o["CallRet"] = n n = Metric_Jump() ; r.metric(n) ; o["Jump"] = n n = Metric_Other_Branches() ; r.metric(n) ; o["Other_Branches"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L1MPKI_Load() ; r.metric(n) ; o["L1MPKI_Load"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_All() ; r.metric(n) ; o["L2MPKI_All"] = n n = Metric_L2MPKI_Load() ; r.metric(n) ; o["L2MPKI_Load"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L2HPKI_Load() ; r.metric(n) ; o["L2HPKI_Load"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_FB_HPKI() ; r.metric(n) ; o["FB_HPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_L3_Cache_Access_BW() ; r.metric(n) ; o["L3_Cache_Access_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_Code_STLB_MPKI() ; r.metric(n) ; o["Code_STLB_MPKI"] = n n = Metric_Load_STLB_MPKI() ; r.metric(n) ; o["Load_STLB_MPKI"] = n n = Metric_Store_STLB_MPKI() ; r.metric(n) ; o["Store_STLB_MPKI"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Access_BW_2T() ; r.metric(n) ; o["L3_Cache_Access_BW_2T"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L3_Miss_Latency() ; r.metric(n) ; o["Load_L3_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_UC_Load_PKI() ; r.metric(n) ; o["UC_Load_PKI"] = n n = Metric_Bus_Lock_PKI() ; r.metric(n) ; o["Bus_Lock_PKI"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_Power_License0_Utilization() ; r.metric(n) ; o["Power_License0_Utilization"] = n n = Metric_Power_License1_Utilization() ; r.metric(n) ; o["Power_License1_Utilization"] = n n = Metric_Power_License2_Utilization() ; r.metric(n) ; o["Power_License2_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_Power() ; r.metric(n) ; o["Power"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Branch_Resteers"].Unknown_Branches = o["Unknown_Branches"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Bad_Speculation"].Retiring = o["Retiring"] o["Bad_Speculation"].Frontend_Bound = o["Frontend_Bound"] o["Bad_Speculation"].Backend_Bound = o["Backend_Bound"] o["Branch_Mispredicts"].Retiring = o["Retiring"] o["Branch_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Branch_Mispredicts"].Frontend_Bound = o["Frontend_Bound"] o["Branch_Mispredicts"].Backend_Bound = o["Backend_Bound"] o["Other_Mispredicts"].Retiring = o["Retiring"] o["Other_Mispredicts"].Backend_Bound = o["Backend_Bound"] o["Other_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Other_Mispredicts"].Frontend_Bound = o["Frontend_Bound"] o["Other_Mispredicts"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Retiring = o["Retiring"] o["Machine_Clears"].Frontend_Bound = o["Frontend_Bound"] o["Machine_Clears"].Backend_Bound = o["Backend_Bound"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Other_Nukes"].Machine_Clears = o["Machine_Clears"] o["Other_Nukes"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Nukes"].Retiring = o["Retiring"] o["Other_Nukes"].Backend_Bound = o["Backend_Bound"] o["Other_Nukes"].Bad_Speculation = o["Bad_Speculation"] o["Other_Nukes"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Retiring = o["Retiring"] o["Memory_Bound"].Backend_Bound = o["Backend_Bound"] o["Load_STLB_Hit"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Load_STLB_Hit"].DTLB_Load = o["DTLB_Load"] o["DRAM_Bound"].L2_Bound = o["L2_Bound"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Store_STLB_Hit"].DTLB_Store = o["DTLB_Store"] o["Store_STLB_Hit"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Retiring = o["Retiring"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Ports_Utilization"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Ports_Utilization"].Retiring = o["Retiring"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["FP_Arith"].Retiring = o["Retiring"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["X87_Use"].Retiring = o["Retiring"] o["FP_Scalar"].Retiring = o["Retiring"] o["FP_Vector"].Retiring = o["Retiring"] o["FP_Vector_128b"].Retiring = o["Retiring"] o["FP_Vector_256b"].Retiring = o["Retiring"] o["FP_Vector_512b"].Retiring = o["Retiring"] o["Memory_Operations"].Retiring = o["Retiring"] o["Memory_Operations"].Light_Operations = o["Light_Operations"] o["Memory_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Memory_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Branch_Instructions"].Retiring = o["Retiring"] o["Branch_Instructions"].Light_Operations = o["Light_Operations"] o["Branch_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Branch_Instructions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Other_Light_Ops"].Light_Operations = o["Light_Operations"] o["Other_Light_Ops"].Retiring = o["Retiring"] o["Other_Light_Ops"].FP_Arith = o["FP_Arith"] o["Other_Light_Ops"].Heavy_Operations = o["Heavy_Operations"] o["Other_Light_Ops"].FP_Vector = o["FP_Vector"] o["Other_Light_Ops"].FP_Scalar = o["FP_Scalar"] o["Other_Light_Ops"].Branch_Instructions = o["Branch_Instructions"] o["Other_Light_Ops"].X87_Use = o["X87_Use"] o["Other_Light_Ops"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Other_Light_Ops"].Memory_Operations = o["Memory_Operations"] o["Nop_Instructions"].Retiring = o["Retiring"] o["Nop_Instructions"].Light_Operations = o["Light_Operations"] o["Nop_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Nop_Instructions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Heavy_Operations"].Retiring = o["Retiring"] o["Heavy_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Few_Uops_Instructions"].Retiring = o["Retiring"] o["Few_Uops_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Few_Uops_Instructions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] o["Mispredictions"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Mispredictions"].LCP = o["LCP"] o["Mispredictions"].Retiring = o["Retiring"] o["Mispredictions"].Other_Mispredicts = o["Other_Mispredicts"] o["Mispredictions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Mispredictions"].Frontend_Bound = o["Frontend_Bound"] o["Mispredictions"].DSB_Switches = o["DSB_Switches"] o["Mispredictions"].Backend_Bound = o["Backend_Bound"] o["Mispredictions"].Branch_Resteers = o["Branch_Resteers"] o["Mispredictions"].ICache_Misses = o["ICache_Misses"] o["Mispredictions"].MS_Switches = o["MS_Switches"] o["Mispredictions"].Bad_Speculation = o["Bad_Speculation"] o["Mispredictions"].ITLB_Misses = o["ITLB_Misses"] o["Mispredictions"].Unknown_Branches = o["Unknown_Branches"] o["Mispredictions"].Fetch_Latency = o["Fetch_Latency"] o["Mispredictions"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Big_Code"].LCP = o["LCP"] o["Big_Code"].ICache_Misses = o["ICache_Misses"] o["Big_Code"].DSB_Switches = o["DSB_Switches"] o["Big_Code"].Branch_Resteers = o["Branch_Resteers"] o["Big_Code"].MS_Switches = o["MS_Switches"] o["Big_Code"].ITLB_Misses = o["ITLB_Misses"] o["Big_Code"].Unknown_Branches = o["Unknown_Branches"] o["Big_Code"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].Retiring = o["Retiring"] o["Instruction_Fetch_BW"].Other_Mispredicts = o["Other_Mispredicts"] o["Instruction_Fetch_BW"].DSB_Switches = o["DSB_Switches"] o["Instruction_Fetch_BW"].Assists = o["Assists"] o["Instruction_Fetch_BW"].Backend_Bound = o["Backend_Bound"] o["Instruction_Fetch_BW"].Branch_Resteers = o["Branch_Resteers"] o["Instruction_Fetch_BW"].Heavy_Operations = o["Heavy_Operations"] o["Instruction_Fetch_BW"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].ICache_Misses = o["ICache_Misses"] o["Instruction_Fetch_BW"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Instruction_Fetch_BW"].Frontend_Bound = o["Frontend_Bound"] o["Instruction_Fetch_BW"].Bad_Speculation = o["Bad_Speculation"] o["Instruction_Fetch_BW"].ITLB_Misses = o["ITLB_Misses"] o["Instruction_Fetch_BW"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Instruction_Fetch_BW"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Instruction_Fetch_BW"].LCP = o["LCP"] o["Instruction_Fetch_BW"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Instruction_Fetch_BW"].Clears_Resteers = o["Clears_Resteers"] o["Instruction_Fetch_BW"].MS_Switches = o["MS_Switches"] o["Instruction_Fetch_BW"].Unknown_Branches = o["Unknown_Branches"] o["Cache_Memory_Bandwidth"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Bandwidth"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Bandwidth"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Bandwidth"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Bandwidth"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Bandwidth"].Retiring = o["Retiring"] o["Cache_Memory_Bandwidth"].G4K_Aliasing = o["G4K_Aliasing"] o["Cache_Memory_Bandwidth"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Bandwidth"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Bandwidth"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Bandwidth"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Bandwidth"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Bandwidth"].Backend_Bound = o["Backend_Bound"] o["Cache_Memory_Bandwidth"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Bandwidth"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Bandwidth"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Bandwidth"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Bandwidth"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Bandwidth"].FB_Full = o["FB_Full"] o["Cache_Memory_Bandwidth"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Bandwidth"].DRAM_Bound = o["DRAM_Bound"] o["Cache_Memory_Latency"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Latency"].G4K_Aliasing = o["G4K_Aliasing"] o["Cache_Memory_Latency"].Retiring = o["Retiring"] o["Cache_Memory_Latency"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Latency"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Latency"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Latency"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Latency"].Store_Latency = o["Store_Latency"] o["Cache_Memory_Latency"].Backend_Bound = o["Backend_Bound"] o["Cache_Memory_Latency"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Latency"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Latency"].False_Sharing = o["False_Sharing"] o["Cache_Memory_Latency"].Streaming_Stores = o["Streaming_Stores"] o["Cache_Memory_Latency"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Latency"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Latency"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Latency"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Latency"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Latency"].FB_Full = o["FB_Full"] o["Cache_Memory_Latency"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Latency"].DTLB_Store = o["DTLB_Store"] o["Cache_Memory_Latency"].Split_Stores = o["Split_Stores"] o["Cache_Memory_Latency"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Latency"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Latency"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Data_TLBs"].L1_Bound = o["L1_Bound"] o["Memory_Data_TLBs"].DTLB_Load = o["DTLB_Load"] o["Memory_Data_TLBs"].False_Sharing = o["False_Sharing"] o["Memory_Data_TLBs"].G4K_Aliasing = o["G4K_Aliasing"] o["Memory_Data_TLBs"].Retiring = o["Retiring"] o["Memory_Data_TLBs"].DTLB_Store = o["DTLB_Store"] o["Memory_Data_TLBs"].L2_Bound = o["L2_Bound"] o["Memory_Data_TLBs"].Memory_Bound = o["Memory_Bound"] o["Memory_Data_TLBs"].Store_Bound = o["Store_Bound"] o["Memory_Data_TLBs"].Split_Loads = o["Split_Loads"] o["Memory_Data_TLBs"].L3_Bound = o["L3_Bound"] o["Memory_Data_TLBs"].FB_Full = o["FB_Full"] o["Memory_Data_TLBs"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Data_TLBs"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Memory_Data_TLBs"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Memory_Data_TLBs"].Store_Latency = o["Store_Latency"] o["Memory_Data_TLBs"].Split_Stores = o["Split_Stores"] o["Memory_Data_TLBs"].Lock_Latency = o["Lock_Latency"] o["Memory_Data_TLBs"].Backend_Bound = o["Backend_Bound"] o["Memory_Data_TLBs"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Synchronization"].L1_Bound = o["L1_Bound"] o["Memory_Synchronization"].False_Sharing = o["False_Sharing"] o["Memory_Synchronization"].Retiring = o["Retiring"] o["Memory_Synchronization"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Synchronization"].Machine_Clears = o["Machine_Clears"] o["Memory_Synchronization"].Data_Sharing = o["Data_Sharing"] o["Memory_Synchronization"].Memory_Bound = o["Memory_Bound"] o["Memory_Synchronization"].SQ_Full = o["SQ_Full"] o["Memory_Synchronization"].Store_Bound = o["Store_Bound"] o["Memory_Synchronization"].L3_Bound = o["L3_Bound"] o["Memory_Synchronization"].L2_Bound = o["L2_Bound"] o["Memory_Synchronization"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Synchronization"].Contested_Accesses = o["Contested_Accesses"] o["Memory_Synchronization"].DTLB_Store = o["DTLB_Store"] o["Memory_Synchronization"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Memory_Synchronization"].Store_Latency = o["Store_Latency"] o["Memory_Synchronization"].Split_Stores = o["Split_Stores"] o["Memory_Synchronization"].Backend_Bound = o["Backend_Bound"] o["Memory_Synchronization"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Synchronization"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Memory_Synchronization"].Other_Nukes = o["Other_Nukes"] o["Memory_Synchronization"].DRAM_Bound = o["DRAM_Bound"] o["Compute_Bound_Est"].Serializing_Operation = o["Serializing_Operation"] o["Compute_Bound_Est"].Ports_Utilization = o["Ports_Utilization"] o["Compute_Bound_Est"].Retiring = o["Retiring"] o["Compute_Bound_Est"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Compute_Bound_Est"].Memory_Bound = o["Memory_Bound"] o["Compute_Bound_Est"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Compute_Bound_Est"].Core_Bound = o["Core_Bound"] o["Compute_Bound_Est"].Backend_Bound = o["Backend_Bound"] o["Compute_Bound_Est"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Compute_Bound_Est"].Divider = o["Divider"] o["Compute_Bound_Est"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Ports_Utilization = o["Ports_Utilization"] o["Irregular_Overhead"].Retiring = o["Retiring"] o["Irregular_Overhead"].ICache_Misses = o["ICache_Misses"] o["Irregular_Overhead"].Heavy_Operations = o["Heavy_Operations"] o["Irregular_Overhead"].Frontend_Bound = o["Frontend_Bound"] o["Irregular_Overhead"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Irregular_Overhead"].Core_Bound = o["Core_Bound"] o["Irregular_Overhead"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Bad_Speculation = o["Bad_Speculation"] o["Irregular_Overhead"].ITLB_Misses = o["ITLB_Misses"] o["Irregular_Overhead"].Divider = o["Divider"] o["Irregular_Overhead"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Irregular_Overhead"].Serializing_Operation = o["Serializing_Operation"] o["Irregular_Overhead"].Machine_Clears = o["Machine_Clears"] o["Irregular_Overhead"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Irregular_Overhead"].LCP = o["LCP"] o["Irregular_Overhead"].Other_Mispredicts = o["Other_Mispredicts"] o["Irregular_Overhead"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Irregular_Overhead"].DSB_Switches = o["DSB_Switches"] o["Irregular_Overhead"].Memory_Bound = o["Memory_Bound"] o["Irregular_Overhead"].Assists = o["Assists"] o["Irregular_Overhead"].Backend_Bound = o["Backend_Bound"] o["Irregular_Overhead"].Branch_Resteers = o["Branch_Resteers"] o["Irregular_Overhead"].Clears_Resteers = o["Clears_Resteers"] o["Irregular_Overhead"].MS_Switches = o["MS_Switches"] o["Irregular_Overhead"].Other_Nukes = o["Other_Nukes"] o["Irregular_Overhead"].Unknown_Branches = o["Unknown_Branches"] o["Irregular_Overhead"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].Retiring = o["Retiring"] o["Other_Bottlenecks"].Data_Sharing = o["Data_Sharing"] o["Other_Bottlenecks"].L2_Bound = o["L2_Bound"] o["Other_Bottlenecks"].Contested_Accesses = o["Contested_Accesses"] o["Other_Bottlenecks"].L3_Bound = o["L3_Bound"] o["Other_Bottlenecks"].Machine_Clears = o["Machine_Clears"] o["Other_Bottlenecks"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Other_Bottlenecks"].Store_Latency = o["Store_Latency"] o["Other_Bottlenecks"].Other_Mispredicts = o["Other_Mispredicts"] o["Other_Bottlenecks"].DSB_Switches = o["DSB_Switches"] o["Other_Bottlenecks"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Other_Bottlenecks"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Other_Bottlenecks"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Other_Bottlenecks"].DTLB_Load = o["DTLB_Load"] o["Other_Bottlenecks"].ICache_Misses = o["ICache_Misses"] o["Other_Bottlenecks"].Streaming_Stores = o["Streaming_Stores"] o["Other_Bottlenecks"].Memory_Bound = o["Memory_Bound"] o["Other_Bottlenecks"].SQ_Full = o["SQ_Full"] o["Other_Bottlenecks"].Store_Bound = o["Store_Bound"] o["Other_Bottlenecks"].Bad_Speculation = o["Bad_Speculation"] o["Other_Bottlenecks"].FB_Full = o["FB_Full"] o["Other_Bottlenecks"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Other_Bottlenecks"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Other_Bottlenecks"].Split_Stores = o["Split_Stores"] o["Other_Bottlenecks"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Other_Bottlenecks"].Other_Nukes = o["Other_Nukes"] o["Other_Bottlenecks"].Unknown_Branches = o["Unknown_Branches"] o["Other_Bottlenecks"].DRAM_Bound = o["DRAM_Bound"] o["Other_Bottlenecks"].L1_Bound = o["L1_Bound"] o["Other_Bottlenecks"].G4K_Aliasing = o["G4K_Aliasing"] o["Other_Bottlenecks"].Core_Bound = o["Core_Bound"] o["Other_Bottlenecks"].Divider = o["Divider"] o["Other_Bottlenecks"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Other_Bottlenecks"].Assists = o["Assists"] o["Other_Bottlenecks"].Backend_Bound = o["Backend_Bound"] o["Other_Bottlenecks"].Branch_Resteers = o["Branch_Resteers"] o["Other_Bottlenecks"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Other_Bottlenecks"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Other_Bottlenecks"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].Ports_Utilization = o["Ports_Utilization"] o["Other_Bottlenecks"].False_Sharing = o["False_Sharing"] o["Other_Bottlenecks"].Heavy_Operations = o["Heavy_Operations"] o["Other_Bottlenecks"].Frontend_Bound = o["Frontend_Bound"] o["Other_Bottlenecks"].Serializing_Operation = o["Serializing_Operation"] o["Other_Bottlenecks"].MEM_Latency = o["MEM_Latency"] o["Other_Bottlenecks"].Split_Loads = o["Split_Loads"] o["Other_Bottlenecks"].ITLB_Misses = o["ITLB_Misses"] o["Other_Bottlenecks"].DTLB_Store = o["DTLB_Store"] o["Other_Bottlenecks"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Bottlenecks"].LCP = o["LCP"] o["Other_Bottlenecks"].Lock_Latency = o["Lock_Latency"] o["Other_Bottlenecks"].Clears_Resteers = o["Clears_Resteers"] o["Other_Bottlenecks"].MS_Switches = o["MS_Switches"] o["Other_Bottlenecks"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Useful_Work"].Assists = o["Assists"] o["Useful_Work"].Retiring = o["Retiring"] o["Useful_Work"].Heavy_Operations = o["Heavy_Operations"] o["Useful_Work"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Useful_Work"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Core_Bound_Likely"].Memory_Bound = o["Memory_Bound"] o["Core_Bound_Likely"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Core_Bound_Likely"].Core_Bound = o["Core_Bound"] o["Core_Bound_Likely"].Backend_Bound = o["Backend_Bound"] o["Core_Bound_Likely"].Retiring = o["Retiring"] o["Core_Bound_Likely"].Ports_Utilization = o["Ports_Utilization"] o["UopPI"].Retiring = o["Retiring"] o["UpTB"].Retiring = o["Retiring"] o["Retire"].Retiring = o["Retiring"] o["DSB_Misses"].LSD = o["LSD"] o["DSB_Misses"].MITE = o["MITE"] o["DSB_Misses"].LCP = o["LCP"] o["DSB_Misses"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Misses"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Misses"].DSB_Switches = o["DSB_Switches"] o["DSB_Misses"].Branch_Resteers = o["Branch_Resteers"] o["DSB_Misses"].ICache_Misses = o["ICache_Misses"] o["DSB_Misses"].MS_Switches = o["MS_Switches"] o["DSB_Misses"].ITLB_Misses = o["ITLB_Misses"] o["DSB_Misses"].DSB = o["DSB"] o["DSB_Misses"].Unknown_Branches = o["Unknown_Branches"] o["DSB_Misses"].Fetch_Latency = o["Fetch_Latency"] o["DSB_Bandwidth"].LSD = o["LSD"] o["DSB_Bandwidth"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Bandwidth"].MITE = o["MITE"] o["DSB_Bandwidth"].DSB = o["DSB"] o["DSB_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].LCP = o["LCP"] o["IC_Misses"].MS_Switches = o["MS_Switches"] o["IC_Misses"].ICache_Misses = o["ICache_Misses"] o["IC_Misses"].ITLB_Misses = o["ITLB_Misses"] o["IC_Misses"].Unknown_Branches = o["Unknown_Branches"] o["IC_Misses"].DSB_Switches = o["DSB_Switches"] o["IC_Misses"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].Retiring = o["Retiring"] o["Branch_Misprediction_Cost"].ICache_Misses = o["ICache_Misses"] o["Branch_Misprediction_Cost"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Branch_Misprediction_Cost"].Frontend_Bound = o["Frontend_Bound"] o["Branch_Misprediction_Cost"].Bad_Speculation = o["Bad_Speculation"] o["Branch_Misprediction_Cost"].ITLB_Misses = o["ITLB_Misses"] o["Branch_Misprediction_Cost"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Branch_Misprediction_Cost"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Branch_Misprediction_Cost"].LCP = o["LCP"] o["Branch_Misprediction_Cost"].Other_Mispredicts = o["Other_Mispredicts"] o["Branch_Misprediction_Cost"].DSB_Switches = o["DSB_Switches"] o["Branch_Misprediction_Cost"].Backend_Bound = o["Backend_Bound"] o["Branch_Misprediction_Cost"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].MS_Switches = o["MS_Switches"] o["Branch_Misprediction_Cost"].Unknown_Branches = o["Unknown_Branches"] o["Branch_Misprediction_Cost"].Fetch_Latency = o["Fetch_Latency"] # siblings cross-tree o["Mispredicts_Resteers"].sibling = (o["Branch_Mispredicts"],) o["Clears_Resteers"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["MS_Switches"].sibling = (o["Clears_Resteers"], o["Machine_Clears"], o["L1_Bound"], o["Serializing_Operation"], o["Mixing_Vectors"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Decoder0_Alone"].sibling = (o["Few_Uops_Instructions"],) o["Branch_Mispredicts"].sibling = (o["Mispredicts_Resteers"],) o["Machine_Clears"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"], o["Streaming_Stores"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"],) o["Streaming_Stores"].sibling = (o["FB_Full"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Serializing_Operation"].sibling = (o["MS_Switches"],) o["Mixing_Vectors"].sibling = (o["MS_Switches"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["Port_5"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["Port_6"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"], o["FP_Vector_512b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_512b"],) o["FP_Vector_512b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Few_Uops_Instructions"].sibling = (o["Decoder0_Alone"],) o["Microcode_Sequencer"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["Mispredictions"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["Cache_Memory_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],) o["Cache_Memory_Latency"].sibling = (o["L3_Hit_Latency"], o["MEM_Latency"],) o["Memory_Data_TLBs"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Memory_Synchronization"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Irregular_Overhead"].sibling = (o["MS_Switches"], o["Microcode_Sequencer"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Misses"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["Branch_Misprediction_Cost"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
238,593
Python
.py
5,504
37.395712
1,797
0.656737
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,882
bdx_server_ratios.py
andikleen_pmu-tools/bdx_server_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel Xeon E5 v4 (code named Broadwell EP) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 8 Mem_L2_Store_Cost = 9 Mem_L3_Weight = 7 Mem_STLB_Hit_Cost = 8 BAClear_Cost = 12 MS_Switches_Cost = 2 Avg_Assist_Cost = 66 Pipeline_Width = 4 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 61 Errata_Whitelist = "BDE69;BDE70" EBS_Mode = 0 DS = 1 # Aux. formulas def Backend_Bound_Cycles(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_TOTAL", level) + EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - Few_Uops_Executed_Threshold(self, EV, level) - Frontend_RS_Empty_Cycles(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) def Cycles_0_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:i1:c1", level)) / 2 if smt_enabled else(EV("CYCLE_ACTIVITY.STALLS_TOTAL", level) - Frontend_RS_Empty_Cycles(self, EV, level)) def Cycles_1_Port_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) - EV("UOPS_EXECUTED.CORE:c2", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level)) def Cycles_2_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c2", level) - EV("UOPS_EXECUTED.CORE:c3", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level)) def Cycles_3m_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c3", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) def Fetched_Uops(self, EV, level): return (EV("IDQ.DSB_UOPS", level) + EV("LSD.UOPS", level) + EV("IDQ.MITE_UOPS", level) + EV("IDQ.MS_UOPS", level)) def Few_Uops_Executed_Threshold(self, EV, level): EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) return EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) if (IPC(self, EV, level)> 1.8) else EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) + 2 * EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 * EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_ARITH_INST_RETIRED.VECTOR", level) def Frontend_RS_Empty_Cycles(self, EV, level): EV("RS_EVENTS.EMPTY_CYCLES", level) return EV("RS_EVENTS.EMPTY_CYCLES", level) if (self.Fetch_Latency.compute(EV)> 0.1) else 0 def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def ITLB_Miss_Cycles(self, EV, level): return (14 * EV("ITLB_MISSES.STLB_HIT", level) + EV("ITLB_MISSES.WALK_DURATION:c1", level) + 7 * EV("ITLB_MISSES.WALK_COMPLETED", level)) def LOAD_L1_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L2_HIT", level) + EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS", level) def LOAD_L1_MISS_NET(self, EV, level): return LOAD_L1_MISS(self, EV, level) + EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM", level) + EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM", level) + EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM", level) + EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD", level) def LOAD_L3_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_LCL_MEM(self, EV, level): return EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_FWD(self, EV, level): return EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_HITM(self, EV, level): return EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_RMT_MEM(self, EV, level): return EV("MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HITM(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def Mem_L3_Hit_Fraction(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) / (EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) + Mem_L3_Weight * EV("MEM_LOAD_UOPS_RETIRED.L3_MISS", level)) def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_UOPS_RETIRED.LOCK_LOADS", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) def Memory_Bound_Fraction(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", level) + EV("RESOURCE_STALLS.SB", level)) / Backend_Bound_Cycles(self, EV, level) def Mispred_Clears_Fraction(self, EV, level): return EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c4", level)) , level ) def SQ_Full_Cycles(self, EV, level): return (EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) / 2) if smt_enabled else EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) def Store_L2_Hit_Cycles(self, EV, level): return EV("L2_RQSTS.RFO_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 60 def Mem_XSNP_Hit_Cost(self, EV, level): return 43 def Mem_XSNP_None_Cost(self, EV, level): return 41 def Mem_Local_DRAM_Cost(self, EV, level): return 200 def Mem_Remote_DRAM_Cost(self, EV, level): return 310 def Mem_Remote_HitM_Cost(self, EV, level): return 200 def Mem_Remote_Fwd_Cost(self, EV, level): return 180 def Recovery_Cycles(self, EV, level): return (EV("INT_MISC.RECOVERY_CYCLES_ANY", level) / 2) if smt_enabled else EV("INT_MISC.RECOVERY_CYCLES", level) def Retire_Fraction(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return EV("UOPS_RETIRED.RETIRE_SLOTS", level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return 8 if smt_enabled else 4 # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return Pipeline_Width * CORE_CLKS(self, EV, level) # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common. def FP_Arith_Utilization(self, EV, level): return (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) / (2 * CORE_CLKS(self, EV, level)) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return ((EV("CPU_CLK_UNHALTED.THREAD", level) / 2) * (1 + EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_XCLK", level))) if ebs_mode else(EV("CPU_CLK_UNHALTED.THREAD_ANY", level) / 2) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408 def IpFLOP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / FLOP_Count(self, EV, level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_SP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_SINGLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_DP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX128(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX256(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.RETIRE_SLOTS:c1", level) def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = Instructions(self, EV, level) / (Retire_Fraction(self, EV, level) * EV("BR_MISP_EXEC.INDIRECT", level)) self.thresh = (val < 1000) return val # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / (EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level)) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache misses per kilo instruction for all request types (including speculative) def L2MPKI_All(self, EV, level): return 1000 * EV("L2_RQSTS.MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache misses per kilo instruction for all demand loads (including speculative) def L2MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("OFFCORE_REQUESTS.DEMAND_RFO", level) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all request types (including speculative) def L2HPKI_All(self, EV, level): return 1000 *(EV("L2_RQSTS.REFERENCES", level) - EV("L2_RQSTS.MISS", level)) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all demand loads (including speculative) def L2HPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_HIT", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L3_MISS", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_DURATION", level) + EV("DTLB_LOAD_MISSES.WALK_DURATION", level) + EV("DTLB_STORE_MISSES.WALK_DURATION", level) + 7 *(EV("DTLB_STORE_MISSES.WALK_COMPLETED", level) + EV("DTLB_LOAD_MISSES.WALK_COMPLETED", level) + EV("ITLB_MISSES.WALK_COMPLETED", level))) / (2 * CORE_CLKS(self, EV, level)) self.thresh = (val > 0.5) return val # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Measured Average Uncore Frequency for the SoC [GHz] def Uncore_Frequency(self, EV, level): return Socket_CLKS(self, EV, level) / 1e9 / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / (EV("CPU_CLK_UNHALTED.REF_XCLK_ANY", level) / 2) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return (64 *(EV("UNC_M_CAS_COUNT.RD", level) + EV("UNC_M_CAS_COUNT.WR", level)) / OneBillion) / Time(self, EV, level) # Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches. def MEM_Read_Latency(self, EV, level): return OneBillion *(EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:opc=0x182", level) / EV("UNC_C_TOR_INSERTS.MISS_OPCODE:opc=0x182", level)) / (Socket_CLKS(self, EV, level) / Time(self, EV, level)) # Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches def MEM_Parallel_Reads(self, EV, level): return EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:opc=0x182", level) / EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:opc=0x182:c1", level) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_C_CLOCKTICKS:one_unit", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['RS_EVENTS.EMPTY_END'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Pipeline_Width * EV("IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE.IFDATA_STALL", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['ITLB_MISSES.WALK_COMPLETED'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = ITLB_Miss_Cycles(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = BAClear_Cost *(EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) + EV("MACHINE_CLEARS.COUNT", 3) + EV("BACLEARS.ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class Mispredicts_Resteers: name = "Mispredicts_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP']) maxval = None def compute(self, EV): try: self.val = EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) * self.Branch_Resteers.compute(EV) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) + EV("MACHINE_CLEARS.COUNT", 4) + EV("BACLEARS.ANY", 4)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Mispredicts_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage.""" class Clears_Resteers: name = "Clears_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'MachineClears']) maxval = None def compute(self, EV): try: self.val = EV("MACHINE_CLEARS.COUNT", 4) * self.Branch_Resteers.compute(EV) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) + EV("MACHINE_CLEARS.COUNT", 4) + EV("BACLEARS.ANY", 4)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Clears_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears.""" class Unknown_Branches: name = "Unknown_Branches" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['BACLEARS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = self.Branch_Resteers.compute(EV) - self.Mispredicts_Resteers.compute(EV) - self.Clears_Resteers.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Unknown_Branches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (e.g. first time the branch is fetched or hitting BTB capacity limit) hence called Unknown Branches""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['IDQ.MS_SWITCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("IDQ.MS_SWITCHES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("ILD_STALL.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_MITE_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_MITE_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_DSB_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_DSB_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_ISSUED.ANY", 1) - Retired_Slots(self, EV, 1) + Pipeline_Width * Recovery_Cycles(self, EV, 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 2) * self.Bad_Speculation.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = 1 -(self.Frontend_Bound.compute(EV) + self.Bad_Speculation.compute(EV) + self.Retiring.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Memory_Bound_Fraction(self, EV, 2) * self.Backend_Bound.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L1_HIT:pp', 'MEM_LOAD_UOPS_RETIRED.HIT_LFB:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3) - EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT", 4) + EV("DTLB_LOAD_MISSES.WALK_DURATION:c1", 4) + 7 * EV("DTLB_LOAD_MISSES.WALK_COMPLETED", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.LOCK_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Lock_St_Fraction(self, EV, 4) * ORO_Demand_RFO_C1(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("LD_BLOCKS.NO_SR", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class G4K_Aliasing: name = "4K_Aliasing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("LD_BLOCKS_PARTIAL.ADDRESS_ALIAS", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "G4K_Aliasing zero division") return self.val desc = """ This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).. Consider reducing independent loads/stores accesses with 4K offsets. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("L1D_PEND_MISS.FB_FULL:c1", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", 3) - EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = Mem_L3_Hit_Fraction(self, EV, 3) * EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_HitM_Cost(self, EV, 4) * LOAD_XSNP_HITM(self, EV, 4) + Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_MISS(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_None_Cost(self, EV, 4) * LOAD_L3_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = SQ_Full_Cycles(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = (1 - Mem_L3_Hit_Fraction(self, EV, 3)) * EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Local_MEM: name = "Local_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_L3_MISS_RETIRED.LOCAL_DRAM:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Server']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Local_DRAM_Cost(self, EV, 5) * LOAD_LCL_MEM(self, EV, 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Local_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance.""" class Remote_MEM: name = "Remote_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_DRAM:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Remote_DRAM_Cost(self, EV, 5) * LOAD_RMT_MEM(self, EV, 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations.""" class Remote_Cache: name = "Remote_Cache" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM:pp', 'MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_FWD:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore', 'Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Remote_HitM_Cost(self, EV, 5) * LOAD_RMT_HITM(self, EV, 5) + Mem_Remote_Fwd_Cost(self, EV, 5) * LOAD_RMT_FWD(self, EV, 5)) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_Cache zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations.""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_UOPS_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SB", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_UOPS_L3_MISS_RETIRED.REMOTE_HITM:pp', 'OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE', 'OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Remote_HitM_Cost(self, EV, 4) * EV("OFFCORE_RESPONSE.DEMAND_RFO.LLC_MISS.REMOTE_HITM", 4) + Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OFFCORE_RESPONSE.DEMAND_RFO.LLC_HIT.HITM_OTHER_CORE", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 2 * EV("MEM_UOPS_RETIRED.SPLIT_STORES", 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_STORE_MISSES.STLB_HIT", 4) + EV("DTLB_STORE_MISSES.WALK_DURATION:c1", 4) + 7 * EV("DTLB_STORE_MISSES.WALK_COMPLETED", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.FPU_DIV_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.FPU_DIV_ACTIVE", 3) / CORE_CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = (Backend_Bound_Cycles(self, EV, 3) - EV("RESOURCE_STALLS.SB", 3) - EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_0_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_1_Port_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_2_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_3m_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_0", 5) + EV("UOPS_DISPATCHED_PORT.PORT_1", 5) + EV("UOPS_DISPATCHED_PORT.PORT_5", 5) + EV("UOPS_DISPATCHED_PORT.PORT_6", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU and 2nd branch""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_5: name = "Port_5" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_5'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_5", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_5 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ALU. See section 'Handling Port 5 Pressure' in Optimization Manual:. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Port_6: name = "Port_6" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_6'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_6", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_6 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 Primary Branch and simple ALU""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_2", 5) + EV("UOPS_DISPATCHED_PORT.PORT_3", 5) + EV("UOPS_DISPATCHED_PORT.PORT_7", 5) - EV("UOPS_DISPATCHED_PORT.PORT_4", 5)) / (2 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Port_2: name = "Port_2" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_2'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_2", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_2 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 Loads and Store-address""" class Port_3: name = "Port_3" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_3'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_3", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_3 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 Loads and Store-address""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Port_4: name = "Port_4" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_4'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_4 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data)""" class Port_7: name = "Port_7" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_7'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_7", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_7 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 simple Store-address""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.RETIRE_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. . Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("INST_RETIRED.X87", 4) * UopPI(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Microcode_Sequencer.compute(EV) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['IDQ.MS_UOPS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = Retire_Fraction(self, EV, 3) * EV("IDQ.MS_UOPS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['OTHER_ASSISTS.ANY_WB_ASSIST'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("OTHER_ASSISTS.ANY_WB_ASSIST", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_FP_Arith_Utilization: name = "FP_Arith_Utilization" domain = "Core_Metric" maxval = 2.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = FP_Arith_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FP_Arith_Utilization zero division") desc = """ Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common.""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpFLOP: name = "IpFLOP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpFLOP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpFLOP zero division") desc = """ Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_IpArith_Scalar_SP: name = "IpArith_Scalar_SP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_SP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_SP zero division") desc = """ Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_DP: name = "IpArith_Scalar_DP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_DP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_DP zero division") desc = """ Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX128: name = "IpArith_AVX128" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX128(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX128 zero division") desc = """ Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX256: name = "IpArith_AVX256" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX256(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX256 zero division") desc = """ Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_All: name = "L2MPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_All zero division") desc = """ L2 cache misses per kilo instruction for all request types (including speculative)""" class Metric_L2MPKI_Load: name = "L2MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Load zero division") desc = """ L2 cache misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L2HPKI_All: name = "L2HPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_All zero division") desc = """ L2 cache hits per kilo instruction for all request types (including speculative)""" class Metric_L2HPKI_Load: name = "L2HPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_Load zero division") desc = """ L2 cache hits per kilo instruction for all demand loads (including speculative)""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_Uncore_Frequency: name = "Uncore_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Uncore_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Uncore_Frequency zero division") desc = """ Measured Average Uncore Frequency for the SoC [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_MEM_Read_Latency: name = "MEM_Read_Latency" domain = "NanoSeconds" maxval = 1000 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryLat', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Read_Latency zero division") desc = """ Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches.""" class Metric_MEM_Parallel_Reads: name = "MEM_Parallel_Reads" domain = "SystemMetric" maxval = 100 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Parallel_Reads(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Parallel_Reads zero division") desc = """ Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = Mispredicts_Resteers() ; r.run(n) ; o["Mispredicts_Resteers"] = n n = Clears_Resteers() ; r.run(n) ; o["Clears_Resteers"] = n n = Unknown_Branches() ; r.run(n) ; o["Unknown_Branches"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = G4K_Aliasing() ; r.run(n) ; o["G4K_Aliasing"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Local_MEM() ; r.run(n) ; o["Local_MEM"] = n n = Remote_MEM() ; r.run(n) ; o["Remote_MEM"] = n n = Remote_Cache() ; r.run(n) ; o["Remote_Cache"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_5() ; r.run(n) ; o["Port_5"] = n n = Port_6() ; r.run(n) ; o["Port_6"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Port_2() ; r.run(n) ; o["Port_2"] = n n = Port_3() ; r.run(n) ; o["Port_3"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Port_4() ; r.run(n) ; o["Port_4"] = n n = Port_7() ; r.run(n) ; o["Port_7"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["Mispredicts_Resteers"].parent = o["Branch_Resteers"] o["Clears_Resteers"].parent = o["Branch_Resteers"] o["Unknown_Branches"].parent = o["Branch_Resteers"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["DSB"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["G4K_Aliasing"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Local_MEM"].parent = o["MEM_Latency"] o["Remote_MEM"].parent = o["MEM_Latency"] o["Remote_Cache"].parent = o["MEM_Latency"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_5"].parent = o["ALU_Op_Utilization"] o["Port_6"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_2"].parent = o["Load_Op_Utilization"] o["Port_3"].parent = o["Load_Op_Utilization"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_4"].parent = o["Store_Op_Utilization"] o["Port_7"].parent = o["Store_Op_Utilization"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["Heavy_Operations"].parent = o["Retiring"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_FP_Arith_Utilization() ; r.metric(n) ; o["FP_Arith_Utilization"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpFLOP() ; r.metric(n) ; o["IpFLOP"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_IpArith_Scalar_SP() ; r.metric(n) ; o["IpArith_Scalar_SP"] = n n = Metric_IpArith_Scalar_DP() ; r.metric(n) ; o["IpArith_Scalar_DP"] = n n = Metric_IpArith_AVX128() ; r.metric(n) ; o["IpArith_AVX128"] = n n = Metric_IpArith_AVX256() ; r.metric(n) ; o["IpArith_AVX256"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_All() ; r.metric(n) ; o["L2MPKI_All"] = n n = Metric_L2MPKI_Load() ; r.metric(n) ; o["L2MPKI_Load"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L2HPKI_All() ; r.metric(n) ; o["L2HPKI_All"] = n n = Metric_L2HPKI_Load() ; r.metric(n) ; o["L2HPKI_Load"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_Uncore_Frequency() ; r.metric(n) ; o["Uncore_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_MEM_Read_Latency() ; r.metric(n) ; o["MEM_Read_Latency"] = n n = Metric_MEM_Parallel_Reads() ; r.metric(n) ; o["MEM_Parallel_Reads"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Mispredicts_Resteers"].Branch_Resteers = o["Branch_Resteers"] o["Clears_Resteers"].Branch_Resteers = o["Branch_Resteers"] o["Unknown_Branches"].Clears_Resteers = o["Clears_Resteers"] o["Unknown_Branches"].Branch_Resteers = o["Branch_Resteers"] o["Unknown_Branches"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Branch_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Backend_Bound"].Retiring = o["Retiring"] o["Backend_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Backend_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Retiring = o["Retiring"] o["Memory_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Backend_Bound = o["Backend_Bound"] o["Memory_Bound"].Fetch_Latency = o["Fetch_Latency"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Core_Bound"].Retiring = o["Retiring"] o["Core_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Core_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Core_Bound"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilization"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilized_0"].Fetch_Latency = o["Fetch_Latency"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["Heavy_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] # siblings cross-tree o["Mispredicts_Resteers"].sibling = (o["Branch_Mispredicts"],) o["Clears_Resteers"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["MS_Switches"].sibling = (o["Clears_Resteers"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Branch_Mispredicts"].sibling = (o["Mispredicts_Resteers"],) o["Machine_Clears"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Remote_Cache"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Remote_Cache"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"],) o["Split_Stores"].sibling = (o["Port_4"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_5"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_6"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_4"].sibling = (o["Split_Stores"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"],) o["Microcode_Sequencer"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
148,580
Python
.py
3,656
34.950492
320
0.656843
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,883
tl-xlsx.py
andikleen_pmu-tools/tl-xlsx.py
#!/usr/bin/env python3 # Copyright (c) 2020, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. # # convert toplev output to xlsx files using xlsxwriter # toplev.py --all --valcsv xv.log --perf-output xp.log -A -a --split-output --per-socket --global --summary \ # --per-core --per-thread -x, -o x.log -I 1000 sleep 10 # tl_xlsx.py --valcsv xv.log --perf xp.log --socket x-socket.log --global x-global.log --core x-core.log --thread x-thread.log x.xlsx from __future__ import print_function import sys import argparse import re import collections import csv try: import xlsxwriter except ImportError: sys.exit("Please install xlswriter with 'pip%s install xlsxwriter'" % sys.version[0]) import gen_level ap = argparse.ArgumentParser(description="Convert toplev CSV files to xlsx") ap.add_argument('xlsxfile', help="xlsx output name") ap.add_argument('--socket', type=argparse.FileType('r'), help="toplev socket csv file", metavar="csvfile") ap.add_argument('--global', type=argparse.FileType('r'), help="toplev global csv file", dest='global_', metavar="csvfile") ap.add_argument('--core', type=argparse.FileType('r'), help="toplev core csv file", metavar="csvfile") ap.add_argument('--program', type=argparse.FileType('r'), help="toplev program csv file", metavar="csvfile") ap.add_argument('--thread', type=argparse.FileType('r'), help="toplev thread csv file", metavar="csvfile") ap.add_argument('--add', nargs=2, help="toplev thread generic csv file. Specify csvfile and sheet name as two arguments", metavar="name", action="append") ap.add_argument('--valcsv', type=argparse.FileType('r'), help="toplev valcsv input file", metavar="csvfile") ap.add_argument('--perf', type=argparse.FileType('r'), help="toplev perf values csv file") ap.add_argument('--cpuinfo', type=argparse.FileType('r'), help="cpuinfo file") ap.add_argument('--chart', help="add sheet with plots of normalized sheet. argument is normalied sheet name", action="append") ap.add_argument('--no-summary', help='Do not generate summary charts', action='store_true') args = ap.parse_args() workbook = xlsxwriter.Workbook(args.xlsxfile, {'constant_memory': True}) bold = workbook.add_format({'bold': True}) valueformat = workbook.add_format({'num_format': '###,###,###,###,##0.0'}) valueformat_bold = workbook.add_format({'num_format': '###,###,###,###,##0.0', 'bold': True}) #valueformat.set_num_format(1) def set_columns(worksheet, c, lengths): for col, j in enumerate(c): if j == "Value": j = " " * 18 if j == "Description": j = "Descr" lengths[col] = max(len(j) + 5, lengths[col]) worksheet.set_column(col, col, lengths[col]) worksheets = {} rows = {} headers = {} def get_worksheet(name): if name in worksheets: return worksheets[name] worksheet = workbook.add_worksheet(name) worksheets[name] = worksheet rows[name] = 1 return worksheet def to_float(n): if re.match(r'-?[,.0-9]+$', n): n = float(n) return n def create_sheet(name, infh, delimiter=',', version=None): lengths = collections.defaultdict(lambda: 0) worksheet = get_worksheet(name) cf = csv.reader(infh, delimiter=delimiter) row = 0 title = {} titlerow = [] summary = False for c in cf: if len(c) > 0 and len(c[0]) > 0 and c[0][0] == "#": version = c continue if row == 0: title = collections.OrderedDict() for i, k in enumerate(c): title[k] = i titlerow = c headers[name] = title if row < 10: set_columns(worksheet, c, lengths) if not summary and len(c) > 0 and c[0] == "SUMMARY": if args.no_summary: continue rows[name] = row sname = name + " summary" worksheet = get_worksheet(sname) set_columns(worksheet, c, lengths) worksheet.write_row(0, 0, titlerow) headers[sname] = titlerow row = rows[sname] summary = True elif summary and len(c) > 0 and c[0] != "SUMMARY" and c[0][0] != "#": worksheet = get_worksheet(name) summary = False rows[sname] = row row = rows[name] c = list(map(to_float, c)) worksheet.write_row(row, 0, c) isbn = False if "Bottleneck" in title: bn = title["Bottleneck"] if len(c) > bn and c[bn] == "<==": worksheet.write(row, title["Area"], c[title["Area"]], bold) worksheet.write(row, title["Value"], c[title["Value"]], bold) if "CPUs" in title: worksheet.write(row, title["CPUs"], c[title["CPUs"]], bold) isbn = True worksheet.write(row, bn, c[bn], bold) if "Value" in title and len(c) > title["Value"] and isinstance(c[title["Value"]], float): worksheet.write_number(row, title["Value"], c[title["Value"]], valueformat_bold if isbn else valueformat) elif "0" in title: num = 0 while num in title: col = title["%d" % num] if len(c) > col and re.match(r'[0-9]+', c[col]): worksheet.write_number(row, col, float(c[col]), valueformat) num += 1 row += 1 return version ROW_SCALE = 18 MIN_ROWS = 15 GRAPH_ROWS = 15 def gen_chart(source): if source not in headers: print("source %s for chart not found" % source, file=sys.stderr) return worksheet = get_worksheet(source + " chart") charts = collections.OrderedDict() for n, ind in headers[source].items(): if n == "Timestamp": continue ns = n.split() if len(ns) > 1: prefix = ns[0] + " " nn = " ".join(ns[1:]) else: prefix = '' nn = n if gen_level.is_metric(nn): chart = workbook.add_chart({'type': 'line'}) charts[n] = chart chart.set_title({'name': n}) else: key = n[:n.rindex('.')] if '.' in n else prefix if key not in charts: charts[key] = workbook.add_chart( {'type': 'column', 'subtype': 'percent_stacked' }) chart = charts[key] chart.set_title({ 'name': '%s Level %d %s' % (prefix, n.count('.') + 1, key[key.index(' '):] if ' ' in key else key) }) chart.set_x_axis({'name': 'Timestamp'}) chart.add_series({ 'name': [source, 0, ind], 'categories': [source, 1, 0, rows[source], 0], 'values': [source, 1, ind, rows[source], ind] }) chart.set_size({'width': (rows[source] + MIN_ROWS ) * ROW_SCALE}) chart.set_legend({ 'overlay': True, 'layout': { 'x': 0.01, 'y': 0.01, 'width': 0.12, 'height': 0.12 }, 'fill': { 'none': True, 'transparency': True } }) row = 1 for j in charts.keys(): worksheet.insert_chart('A%d' % row, charts[j]) row += GRAPH_ROWS version = None if args.global_: version = create_sheet("global", args.global_, version=version) if args.socket: version = create_sheet("socket", args.socket, version=version) if args.core: version = create_sheet("core", args.core, version=version) if args.thread: version = create_sheet("thread", args.thread, version=version) if args.program: version = create_sheet("prog", args.program, version=version) if args.add: for fn, name in args.add: version = create_sheet(name, open(fn), version=version) if args.chart: for cname in args.chart: gen_chart(cname) if args.valcsv: create_sheet("event values", args.valcsv) if args.perf: create_sheet("raw perf output", args.perf, delimiter=';') if args.cpuinfo: create_sheet("cpuinfo", args.cpuinfo, delimiter=':') if version: worksheet = workbook.add_worksheet("version") worksheet.write_row(0, 0, version) workbook.close()
8,633
Python
.py
207
34.120773
133
0.597741
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,884
cputop.py
andikleen_pmu-tools/cputop.py
#!/usr/bin/env python3 # query cpu topology and print all matching cpu numbers # cputop "query" ["format"] # query is a python expression, using variables: # socket, core, thread, type, cpu # or "offline" to query all offline cpus # or "atom" or "core" to select core types # type can be "atom" or "core" # cpu is the cpu number # format is a printf format with %d # %d will be replaced with the cpu number # format can be offline to offline the cpu or online to online # Author: Andi Kleen from __future__ import print_function import sys import os import re import argparse import glob def numfile(fn): f = open(fn, "r") v = int(f.read()) f.close() return v outstr = "" def output(p, fmt): if fmt: if fmt == "taskset": global outstr if outstr: outstr += "," else: outstr += "taskset -c " outstr += "%d" % p else: print(fmt % (p,)) else: print(p) ap = argparse.ArgumentParser(description=''' query cpu topology and print all matching cpu numbers cputop "query" ["format"] query is a python expression, using variables: socket, core, thread, type, cpu type is "core" or "atom" on a hybrid system cpu is the cpu number or "offline" to query all offline cpus format is a printf format with %d %d will be replaced with the cpu number, or online/offline to generate online/offline commands, or taskset to generate taskset command line''', epilog=''' Examples: print all cores on socket 0 cputop "socket == 0" print all first threads in each core on socket 0 cputop "thread == 0 and socket == 0" disable all second threads (disable hyper threading) cputop "thread == 1" offline reenable all offlined cpus cputop offline online print all online cpus cputop True ''', formatter_class=argparse.RawTextHelpFormatter) ap.add_argument('expr', help='python expression with socket/core/thread') ap.add_argument('fmt', help='Output format string with %%d, or online/offline', nargs='?') args = ap.parse_args() if args.expr == "atom": args.expr = 'type == "atom"' if args.expr == "core": args.expr = 'type == "core"' special = { "offline": "echo 0 > /sys/devices/system/cpu/cpu%d/online", "online": "echo 1 > /sys/devices/system/cpu/cpu%d/online", } if args.fmt in special: args.fmt = special[args.fmt] types = dict() for fn in glob.glob("/sys/devices/cpu_*/cpus"): typ = os.path.basename(fn.replace("/cpus", "")).replace("cpu_","") cpus = open(fn).read() for j in cpus.split(","): m = re.match(r'(\d+)(-\d+)?', j) if m is None: continue if m.group(2): for k in range(int(m.group(1)), int(m.group(2)[1:])+1): types[k] = typ else: types[int(m.group(1))] = typ base = "/sys/devices/system/cpu/" p = {} l = os.listdir(base) for d in l: m = re.match(r"cpu([0-9]+)", d) if not m: continue proc = int(m.group(1)) top = base + d + "/topology" if not os.path.exists(top): if args.expr == "offline": output(proc, args.fmt) continue socket = numfile(top + "/physical_package_id") core = numfile(top + "/core_id") n = 0 while (socket, core, n) in p: n += 1 p[(socket, core, n)] = proc if args.expr == "offline": sys.exit(0) for j in sorted(p.keys()): socket, core, thread = j cpu = p[j] type = "any" if cpu in types: type = types[cpu] if eval(args.expr): output(p[j], args.fmt) if outstr: print(outstr)
3,577
Python
.py
118
25.923729
90
0.634331
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,885
spr_server_ratios.py
andikleen_pmu-tools/spr_server_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel Xeon Scalable Processors 4th gen (code name Sapphire Rapids) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 1 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 topdown_use_fixed = False def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 12 Mem_L2_Store_Cost = 10 Mem_STLB_Hit_Cost = 7 MS_Switches_Cost = 3 Avg_Assist_Cost = ( 99 *3 + 63 + 30 ) / 5 Pipeline_Width = 6 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 61 Errata_Whitelist = "SPR121;SPR103" Memory = 0 PMM_App_Direct = 1 if Memory == 1 else 0 HBM = 1 if Memory > 1 else 0 PERF_METRICS_MSR = 1 FP16 = 1 DS = 1 # Aux. formulas def Br_DoI_Jumps(self, EV, level): return EV("BR_INST_RETIRED.NEAR_TAKEN", level) - EV("BR_INST_RETIRED.COND_TAKEN", level) - 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) def Branching_Retired(self, EV, level): return (EV("BR_INST_RETIRED.ALL_BRANCHES", level) + 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("INST_RETIRED.NOP", level)) / SLOTS(self, EV, level) def Serialize_Core(self, EV, level): return self.Core_Bound.compute(EV) * (self.Serializing_Operation.compute(EV) + EV("RS.EMPTY:u1", level) / CLKS(self, EV, level) * self.Ports_Utilized_0.compute(EV)) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.AMX_Busy.compute(EV) + self.Divider.compute(EV)) def Umisp(self, EV, level): return 10 * self.Microcode_Sequencer.compute(EV) * self.Other_Mispredicts.compute(EV) / self.Branch_Mispredicts.compute(EV) def Assist(self, EV, level): return (self.Microcode_Sequencer.compute(EV) / (self.Microcode_Sequencer.compute(EV) + self.Few_Uops_Instructions.compute(EV))) * (self.Assists.compute(EV) / self.Microcode_Sequencer.compute(EV)) def Assist_Frontend(self, EV, level): return (1 - EV("INST_RETIRED.REP_ITERATION", level) / EV("UOPS_RETIRED.MS:c1", level)) * (self.Fetch_Latency.compute(EV) * (self.MS_Switches.compute(EV) + self.Branch_Resteers.compute(EV) * (self.Clears_Resteers.compute(EV) + self.Mispredicts_Resteers.compute(EV) * self.Other_Mispredicts.compute(EV) / self.Branch_Mispredicts.compute(EV)) / (self.Clears_Resteers.compute(EV) + self.Unknown_Branches.compute(EV) + self.Mispredicts_Resteers.compute(EV))) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) def Assist_Retired(self, EV, level): return Assist(self, EV, level) * self.Heavy_Operations.compute(EV) def Core_Bound_Cycles(self, EV, level): return self.Ports_Utilized_0.compute(EV) * CLKS(self, EV, level) + Few_Uops_Executed_Threshold(self, EV, level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE_CYCLES_GE_1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.THREAD:c1", level) # factor used for metrics associating fixed costs for FB Hits - according to probability theory if all FB Hits come at a random rate in original L1_Miss cost interval then the average cost for each one is 0.5 of the fixed cost def FB_Factor(self, EV, level): return 1 + FBHit_per_L1Miss(self, EV, level) / 2 def FBHit_per_L1Miss(self, EV, level): return EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("MEM_LOAD_RETIRED.L1_MISS", level) def Fetched_Uops(self, EV, level): return EV("UOPS_ISSUED.ANY", level) def Few_Uops_Executed_Threshold(self, EV, level): return EV("EXE_ACTIVITY.1_PORTS_UTIL", level) + self.Retiring.compute(EV) * EV("EXE_ACTIVITY.2_PORTS_UTIL:u0xc", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", level) EV("FP_ARITH_INST_RETIRED.SCALAR", level) EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", level) EV("FP_ARITH_INST_RETIRED2.SCALAR_HALF", level) EV("FP_ARITH_INST_RETIRED.8_FLOPS", level) EV("FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF", level) EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level) EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", level) EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) return (1 *(EV("FP_ARITH_INST_RETIRED.SCALAR", level) + EV("FP_ARITH_INST_RETIRED2.SCALAR_HALF", level)) + 2 *(EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED2.COMPLEX_SCALAR_HALF", level)) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 *(EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", level) + EV("FP_ARITH_INST_RETIRED.8_FLOPS", level)) + 16 *(EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", level) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level)) + 32 * EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", level)) if FP16 else(1 * EV("FP_ARITH_INST_RETIRED.SCALAR", level) + 2 * EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 * EV("FP_ARITH_INST_RETIRED.8_FLOPS", level) + 16 * EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level)) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): EV("FP_ARITH_INST_RETIRED2.SCALAR", level) EV("FP_ARITH_INST_RETIRED.SCALAR", level) return EV("FP_ARITH_INST_RETIRED.SCALAR", level) + EV("FP_ARITH_INST_RETIRED2.SCALAR", level) if FP16 else EV("FP_ARITH_INST_RETIRED.SCALAR", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): EV("FP_ARITH_INST_RETIRED.VECTOR", level) EV("FP_ARITH_INST_RETIRED2.VECTOR", level) return EV("FP_ARITH_INST_RETIRED.VECTOR", level) + EV("FP_ARITH_INST_RETIRED2.VECTOR", level) if FP16 else EV("FP_ARITH_INST_RETIRED.VECTOR", level) def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def Light_Ops_Sum(self, EV, level): return self.FP_Arith.compute(EV) + self.Int_Operations.compute(EV) + self.Memory_Operations.compute(EV) + self.Fused_Instructions.compute(EV) + self.Non_Fused_Branches.compute(EV) def LOAD_L3_HIT(self, EV, level): return EV("MEM_LOAD_RETIRED.L3_HIT", level) * FB_Factor(self, EV, level) def LOAD_LCL_MEM(self, EV, level): return EV("MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM", level) * (1 + FBHit_per_L1Miss(self, EV, level)) def LOAD_LCL_PMM(self, EV, level): EV("MEM_LOAD_RETIRED.LOCAL_PMM", level) return EV("MEM_LOAD_RETIRED.LOCAL_PMM", level) * (1 + FBHit_per_L1Miss(self, EV, level)) if DS else 0 def LOAD_RMT_FWD(self, EV, level): EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD", level) return EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD", level) * (1 + FBHit_per_L1Miss(self, EV, level)) if DS else 0 def LOAD_RMT_HITM(self, EV, level): EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM", level) return EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM", level) * (1 + FBHit_per_L1Miss(self, EV, level)) if DS else 0 def LOAD_RMT_MEM(self, EV, level): EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM", level) return EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM", level) * (1 + FBHit_per_L1Miss(self, EV, level)) if DS else 0 def LOAD_RMT_PMM(self, EV, level): EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM", level) return EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM", level) * (1 + FBHit_per_L1Miss(self, EV, level)) if DS else 0 def LOAD_XSNP_HIT(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", level) + EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", level) * (1 - True_XSNP_HitM_Fraction(self, EV, level)) def LOAD_XSNP_HITM(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", level) * True_XSNP_HitM_Fraction(self, EV, level) def LOAD_XSNP_MISS(self, EV, level): return EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", level) def MEM_Bound_Ratio(self, EV, level): return EV("MEMORY_ACTIVITY.STALLS_L3_MISS", level) / CLKS(self, EV, level) def Mem_DDR_Hit_Fraction(self, EV, level): return (19 * LOAD_RMT_MEM(self, EV, level) + 10 *(LOAD_LCL_MEM(self, EV, level) + LOAD_RMT_FWD(self, EV, level) + LOAD_RMT_HITM(self, EV, level))) / ((19 * LOAD_RMT_MEM(self, EV, level) + 10 *(LOAD_LCL_MEM(self, EV, level) + LOAD_RMT_FWD(self, EV, level) + LOAD_RMT_HITM(self, EV, level))) + (25 * LOAD_LCL_PMM(self, EV, level) + 33 * LOAD_RMT_PMM(self, EV, level))) if DS else 1 def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_INST_RETIRED.LOCK_LOADS", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) def Mispred_Clears_Fraction(self, EV, level): return self.Branch_Mispredicts.compute(EV) / self.Bad_Speculation.compute(EV) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c4", level)) , level ) def Store_L2_Hit_Cycles(self, EV, level): return EV("MEM_STORE_RETIRED.L2_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def True_XSNP_HitM_Fraction(self, EV, level): return EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM", level) / (EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM", level) + EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD", level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 81 * Core_Frequency(self, EV, level) def Mem_XSNP_Hit_Cost(self, EV, level): return 79 * Core_Frequency(self, EV, level) def Mem_XSNP_None_Cost(self, EV, level): return 37 * Core_Frequency(self, EV, level) def Mem_Local_DRAM_Cost(self, EV, level): return 109 * Core_Frequency(self, EV, level) def Mem_Remote_DRAM_Cost(self, EV, level): return 190 * Core_Frequency(self, EV, level) def Mem_Remote_HitM_Cost(self, EV, level): return 170 * Core_Frequency(self, EV, level) def Mem_Remote_Fwd_Cost(self, EV, level): return 170 * Core_Frequency(self, EV, level) def Mem_L2_Hit_Cost(self, EV, level): return 4.4 * Core_Frequency(self, EV, level) def PERF_METRICS_SUM(self, EV, level): return (EV("PERF_METRICS.FRONTEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BAD_SPECULATION", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.RETIRING", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BACKEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) def Retire_Fraction(self, EV, level): return EV("UOPS_RETIRED.SLOTS", level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return self.Retiring.compute(EV) * SLOTS(self, EV, level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return num_cores * num_sockets * num_threads if num_cores else 224 /(2 - smt_enabled ) # A system parameter for dependent-loads (pointer chasing like access pattern) of the workload. An integer fraction in range from 0 (no dependent loads) to 100 (all loads are dependent loads) def Dependent_Loads_Weight(self, EV, level): return 20 # Total pipeline cost of Branch Misprediction related bottlenecks def Mispredictions(self, EV, level): val = 100 *(1 - Umisp(self, EV, level)) * (self.Branch_Mispredicts.compute(EV) + self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses) def Big_Code(self, EV, level): val = 100 * self.Fetch_Latency.compute(EV) * (self.ITLB_Misses.compute(EV) + self.ICache_Misses.compute(EV) + self.Unknown_Branches.compute(EV)) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end) def Instruction_Fetch_BW(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) - (1 - Umisp(self, EV, level)) * self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) - Assist_Frontend(self, EV, level)) - Big_Code(self, EV, level) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks def Cache_Memory_Bandwidth(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.MEM_Bandwidth.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.SQ_Full.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.FB_Full.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Latency related bottlenecks def Cache_Memory_Latency(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.MEM_Latency.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.L3_Hit_Latency.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * self.L2_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.Store_Latency.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.L1_Hit_Latency.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs) def Memory_Data_TLBs(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / max(self.Memory_Bound.compute(EV) , (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV)))) * (self.DTLB_Load.compute(EV) / max(self.L1_Bound.compute(EV) , (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.DTLB_Store.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors) def Memory_Synchronization(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * ((self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * (self.Contested_Accesses.compute(EV) + self.Data_Sharing.compute(EV)) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)) + (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.PMM_Bound.compute(EV) + self.L2_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV))) * self.False_Sharing.compute(EV) / ((self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)) - self.Store_Latency.compute(EV))) + self.Machine_Clears.compute(EV) * (1 - self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV)))) self.thresh = (val > 10) return val # Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy. def Compute_Bound_Est(self, EV, level): val = 100 *((self.Core_Bound.compute(EV) * self.Divider.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.AMX_Busy.compute(EV) + self.Divider.compute(EV))) + (self.Core_Bound.compute(EV) * self.AMX_Busy.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.AMX_Busy.compute(EV) + self.Divider.compute(EV))) + (self.Core_Bound.compute(EV) * (self.Ports_Utilization.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.AMX_Busy.compute(EV) + self.Divider.compute(EV))) * (self.Ports_Utilized_3m.compute(EV) / (self.Ports_Utilized_0.compute(EV) + self.Ports_Utilized_1.compute(EV) + self.Ports_Utilized_2.compute(EV) + self.Ports_Utilized_3m.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments) def Irregular_Overhead(self, EV, level): val = 100 *(Assist_Frontend(self, EV, level) + Umisp(self, EV, level) * self.Branch_Mispredicts.compute(EV) + (self.Machine_Clears.compute(EV) * self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV))) + Serialize_Core(self, EV, level) + Assist_Retired(self, EV, level)) self.thresh = (val > 10) return val # Total pipeline cost of remaining bottlenecks in the back-end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls. def Other_Bottlenecks(self, EV, level): val = 100 -(Big_Code(self, EV, level) + Instruction_Fetch_BW(self, EV, level) + Mispredictions(self, EV, level) + Cache_Memory_Bandwidth(self, EV, level) + Cache_Memory_Latency(self, EV, level) + Memory_Data_TLBs(self, EV, level) + Memory_Synchronization(self, EV, level) + Compute_Bound_Est(self, EV, level) + Irregular_Overhead(self, EV, level) + Branching_Overhead(self, EV, level) + Useful_Work(self, EV, level)) self.thresh = (val > 20) return val # Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations def Branching_Overhead(self, EV, level): val = 100 * Branching_Retired(self, EV, level) self.thresh = (val > 5) return val # Total pipeline cost of "useful operations" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead. def Useful_Work(self, EV, level): val = 100 *(self.Retiring.compute(EV) - Branching_Retired(self, EV, level) - Assist_Retired(self, EV, level)) self.thresh = (val > 20) return val # Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled def Core_Bound_Likely(self, EV, level): val = 100 *(1 - self.Core_Bound.compute(EV) / self.Ports_Utilization.compute(EV) if self.Core_Bound.compute(EV)< self.Ports_Utilization.compute(EV) else 1) if SMT_2T_Utilization(self, EV, level)> 0.5 else 0 self.thresh = (val > 0.5) return val # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return EV("TOPDOWN.SLOTS", level) if topdown_use_fixed else EV("TOPDOWN.SLOTS", level) # Fraction of Physical Core issue-slots utilized by this Logical Processor def Slots_Utilization(self, EV, level): return SLOTS(self, EV, level) / (EV("TOPDOWN.SLOTS:percore", level) / 2) if smt_enabled else 1 # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common. def FP_Arith_Utilization(self, EV, level): return (EV("FP_ARITH_DISPATCHED.PORT_0", level) + EV("FP_ARITH_DISPATCHED.PORT_1", level) + EV("FP_ARITH_DISPATCHED.PORT_5", level)) / (2 * CORE_CLKS(self, EV, level)) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # uops Executed per Cycle def EPC(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / CLKS(self, EV, level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.DISTRIBUTED", level) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408 def IpFLOP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / FLOP_Count(self, EV, level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Half-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_HP(self, EV, level): EV("FP_ARITH_INST_RETIRED2.SCALAR", level) EV("INST_RETIRED.ANY", level) val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED2.SCALAR", level) if FP16 else 0 self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_SP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_SINGLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_DP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX128(self, EV, level): EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level) EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) EV("INST_RETIRED.ANY", level) EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", level) val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level) + EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", level)) if FP16 else EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX256(self, EV, level): EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", level) EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) EV("INST_RETIRED.ANY", level) EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level) val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level) + EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", level)) if FP16 else EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX512(self, EV, level): EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level) EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", level) EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", level) EV("INST_RETIRED.ANY", level) val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level) + EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", level)) if FP16 else EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per PAUSE (lower number means higher occurrence rate) def IpPause(self, EV, level): return Instructions(self, EV, level) / EV("CPU_CLK_UNHALTED.PAUSE_INST", level) # Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate) def IpSWPF(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("SW_PREFETCH_ACCESS.T0:u0xF", level) self.thresh = (val < 100) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.SLOTS:c1", level) # Estimated fraction of retirement-cycles dealing with repeat instructions def Strings_Cycles(self, EV, level): val = EV("INST_RETIRED.REP_ITERATION", level) / EV("UOPS_RETIRED.SLOTS:c1", level) self.thresh = (val > 0.1) return val # Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate) def IpAssist(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("ASSISTS.ANY", level) self.thresh = (val < 100000) return val def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Average number of uops fetched from LSD per cycle def Fetch_LSD(self, EV, level): return EV("LSD.UOPS", level) / EV("LSD.CYCLES_ACTIVE", level) # Average number of uops fetched from DSB per cycle def Fetch_DSB(self, EV, level): return EV("IDQ.DSB_UOPS", level) / EV("IDQ.DSB_CYCLES_ANY", level) # Average number of uops fetched from MITE per cycle def Fetch_MITE(self, EV, level): return EV("IDQ.MITE_UOPS", level) / EV("IDQ.MITE_CYCLES_ANY", level) # Average number of Uops issued by front-end when it issued something def Fetch_UpC(self, EV, level): return EV("UOPS_ISSUED.ANY", level) / EV("UOPS_ISSUED.ANY:c1", level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Average number of cycles the front-end was delayed due to an Unknown Branch detection. See Unknown_Branches node. def Unknown_Branch_Cost(self, EV, level): return EV("INT_MISC.UNKNOWN_BRANCH_CYCLES", level) / EV("INT_MISC.UNKNOWN_BRANCH_CYCLES:c1:e1", level) # Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details. def DSB_Switch_Cost(self, EV, level): return EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", level) / EV("DSB2MITE_SWITCHES.PENALTY_CYCLES:c1:e1", level) # Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.DSB_Switches.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) + self.Fetch_Bandwidth.compute(EV) * self.MITE.compute(EV) / (self.MITE.compute(EV) + self.DSB.compute(EV))) self.thresh = (val > 10) return val # Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Bandwidth(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) * (self.Fetch_Bandwidth.compute(EV) / (self.Fetch_Bandwidth.compute(EV) + self.Fetch_Latency.compute(EV))) * (self.DSB.compute(EV) / (self.MITE.compute(EV) + self.DSB.compute(EV)))) self.thresh = (val > 10) return val # Average Latency for L1 instruction cache misses def ICache_Miss_Latency(self, EV, level): return EV("ICACHE_DATA.STALLS", level) / EV("ICACHE_DATA.STALLS:c1:e1", level) # Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck. def IC_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.ICache_Misses.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 5) return val # Instructions per non-speculative DSB miss (lower number means higher occurrence rate) def IpDSB_Miss_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FRONTEND_RETIRED.ANY_DSB_MISS", level) self.thresh = (val < 50) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # L2 cache true code cacheline misses per kilo instruction def L2MPKI_Code(self, EV, level): return 1000 * EV("FRONTEND_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache speculative code cacheline misses per kilo instruction def L2MPKI_Code_All(self, EV, level): return 1000 * EV("L2_RQSTS.CODE_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional non-taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Ntaken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_NTAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Taken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_TAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate). def IpMisp_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.RET", level) self.thresh = (val < 500) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.INDIRECT", level) self.thresh = (val < 1000) return val # Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear) def Branch_Misprediction_Cost(self, EV, level): return Mispredictions(self, EV, level) * SLOTS(self, EV, level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / 100 # Speculative to Retired ratio of all clears (covering Mispredicts and nukes) def Spec_Clears_Ratio(self, EV, level): return EV("INT_MISC.CLEARS_COUNT", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) # Fraction of branches that are non-taken conditionals def Cond_NT(self, EV, level): return EV("BR_INST_RETIRED.COND_NTAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are taken conditionals def Cond_TK(self, EV, level): return EV("BR_INST_RETIRED.COND_TAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are CALL or RET def CallRet(self, EV, level): return (EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("BR_INST_RETIRED.NEAR_RETURN", level)) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are unconditional (direct or indirect) jumps def Jump(self, EV, level): return Br_DoI_Jumps(self, EV, level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches of other types (not individually covered by other metrics in Info.Branches group) def Other_Branches(self, EV, level): return 1 -(Cond_NT(self, EV, level) + Cond_TK(self, EV, level) + CallRet(self, EV, level) + Jump(self, EV, level)) # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("MEM_LOAD_COMPLETED.L1_MISS_ANY", level) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L1 cache true misses per kilo instruction for all demand loads (including speculative) def L1MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.ALL_DEMAND_DATA_RD", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for all request types (including speculative) def L2MPKI_All(self, EV, level): return 1000 * EV("L2_RQSTS.MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for all demand loads (including speculative) def L2MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("L2_RQSTS.RFO_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all request types (including speculative) def L2HPKI_All(self, EV, level): return 1000 *(EV("L2_RQSTS.REFERENCES", level) - EV("L2_RQSTS.MISS", level)) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all demand loads (including speculative) def L2HPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_HIT", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L3_MISS", level) / EV("INST_RETIRED.ANY", level) # Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries) def FB_HPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) def L3_Cache_Access_BW(self, EV, level): return 64 * EV("OFFCORE_REQUESTS.ALL_REQUESTS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_PENDING", level) + EV("DTLB_LOAD_MISSES.WALK_PENDING", level) + EV("DTLB_STORE_MISSES.WALK_PENDING", level)) / (4 * CORE_CLKS(self, EV, level)) self.thresh = (val > 0.5) return val # STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Code_STLB_MPKI(self, EV, level): return 1000 * EV("ITLB_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Load_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_LOAD_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Store_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_STORE_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average per-core data access bandwidth to the L3 cache [GB / sec] def L3_Cache_Access_BW_2T(self, EV, level): return L3_Cache_Access_BW(self, EV, level) # Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory) def L2_Evictions_Silent_PKI(self, EV, level): return 1000 * EV("L2_LINES_OUT.SILENT", level) / Instructions(self, EV, level) # Rate of non silent evictions from the L2 cache per Kilo instruction def L2_Evictions_NonSilent_PKI(self, EV, level): return 1000 * EV("L2_LINES_OUT.NON_SILENT", level) / Instructions(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Latency for L3 cache miss demand Loads def Load_L3_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.L3_MISS_DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.L3_MISS_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD:c1", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Off-core accesses per kilo instruction for reads-to-core requests (speculative; including in-core HW prefetches) def Offcore_Read_Any_PKI(self, EV, level): return 1000 * EV("OCR.READS_TO_CORE.ANY_RESPONSE", level) / Instructions(self, EV, level) # L3 cache misses per kilo instruction for reads-to-core requests (speculative; including in-core HW prefetches) def Offcore_Read_L3M_PKI(self, EV, level): return 1000 * EV("OCR.READS_TO_CORE.L3_MISS", level) / Instructions(self, EV, level) # Off-core accesses per kilo instruction for modified write requests def Offcore_MWrite_Any_PKI(self, EV, level): return 1000 * EV("OCR.MODIFIED_WRITE.ANY_RESPONSE", level) / Instructions(self, EV, level) # Un-cacheable retired load per kilo instruction def UC_Load_PKI(self, EV, level): return 1000 * EV("MEM_LOAD_MISC_RETIRED.UC", level) / EV("INST_RETIRED.ANY", level) # "Bus lock" per kilo instruction def Bus_Lock_PKI(self, EV, level): return 1000 * EV("SQ_MISC.BUS_LOCK", level) / EV("INST_RETIRED.ANY", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Measured Average Uncore Frequency for the SoC [GHz] def Uncore_Frequency(self, EV, level): return Socket_CLKS(self, EV, level) / 1e9 / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_DISTRIBUTED", level) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Fraction of cycles the processor is waiting yet unhalted; covering legacy PAUSE instruction, as well as C0.1 / C0.2 power-performance optimized states. Sample code of TPAUSE: https://github.com/torvalds/linux/blob/master/arch/x86/lib/delay.c#L105. If running on Linux, please check the power control interface: https://github.com/torvalds/linux/blob/master/arch/x86/kernel/cpu/umwait.c and https://github.com/torvalds/linux/blob/master/Documentation/ABI/testing/sysfs-devices-system-cpu#L587 def C0_Wait(self, EV, level): val = EV("CPU_CLK_UNHALTED.C0_WAIT", level) / CLKS(self, EV, level) self.thresh = (val > 0.05) return val # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return (64 *(EV("UNC_M_CAS_COUNT.RD", level) + EV("UNC_M_CAS_COUNT.WR", level)) / OneBillion) / Time(self, EV, level) # Average Off-core access BW for Reads-to-Core (R2C). R2C account for demand or prefetch load/RFO/code access that fill data into the Core caches. def R2C_Offcore_BW(self, EV, level): return 64 * EV("OCR.READS_TO_CORE.ANY_RESPONSE", level) / 1e9 / Time(self, EV, level) # Average L3-cache miss BW for Reads-to-Core (R2C). This covering going to DRAM or other memory off-chip memory tears. See R2C_Offcore_BW. def R2C_L3M_BW(self, EV, level): return 64 * EV("OCR.READS_TO_CORE.L3_MISS", level) / 1e9 / Time(self, EV, level) # Average DRAM BW for Reads-to-Core (R2C) covering for memory attached to local- and remote-socket. See R2C_Offcore_BW. def R2C_DRAM_BW(self, EV, level): return 64 * EV("OCR.READS_TO_CORE.DRAM", level) / 1e9 / Time(self, EV, level) # Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches. memory-controller only def MEM_Read_Latency(self, EV, level): return OneBillion *(EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD", level) / EV("UNC_CHA_TOR_INSERTS.IA_MISS_DRD", level)) / (Socket_CLKS(self, EV, level) / Time(self, EV, level)) # Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches def MEM_Parallel_Reads(self, EV, level): return EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD", level) / EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD:c1", level) # Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches def MEM_PMM_Read_Latency(self, EV, level): EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM", level) EV("UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM", level) return (OneBillion *(EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_PMM", level) / EV("UNC_CHA_TOR_INSERTS.IA_MISS_DRD_PMM", level)) / EV("UNC_CHA_CLOCKTICKS:one_unit", level)) if PMM_App_Direct else 0 # Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches def MEM_DRAM_Read_Latency(self, EV, level): return OneBillion *(EV("UNC_CHA_TOR_OCCUPANCY.IA_MISS_DRD_DDR", level) / EV("UNC_CHA_TOR_INSERTS.IA_MISS_DRD_DDR", level)) / EV("UNC_CHA_CLOCKTICKS:one_unit", level) # Average 3DXP Memory Bandwidth Use for reads [GB / sec] def PMM_Read_BW(self, EV, level): return ((64 * EV("UNC_M_PMM_RPQ_INSERTS", level) / OneBillion) / Time(self, EV, level)) if PMM_App_Direct else 0 # Average 3DXP Memory Bandwidth Use for Writes [GB / sec] def PMM_Write_BW(self, EV, level): return ((64 * EV("UNC_M_PMM_WPQ_INSERTS", level) / OneBillion) / Time(self, EV, level)) if PMM_App_Direct else 0 # Average IO (network or disk) Bandwidth Use for Reads [GB / sec]. Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU def IO_Read_BW(self, EV, level): return EV("UNC_CHA_TOR_INSERTS.IO_PCIRDCUR", level) * 64 / OneBillion / Time(self, EV, level) # Average IO (network or disk) Bandwidth Use for Writes [GB / sec]. Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU def IO_Write_BW(self, EV, level): return (EV("UNC_CHA_TOR_INSERTS.IO_ITOM", level) + EV("UNC_CHA_TOR_INSERTS.IO_ITOMCACHENEAR", level)) * 64 / OneBillion / Time(self, EV, level) # Cross-socket Ultra Path Interconnect (UPI) data transmit bandwidth for data only [MB / sec] def UPI_Data_Transmit_BW(self, EV, level): return EV("UNC_UPI_TxL_FLITS.ALL_DATA", level) * 64 / 9 / 1000000 # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_CHA_CLOCKTICKS:one_unit", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_4:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.FRONTEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) - EV("INT_MISC.UOP_DROPPING", 1) / SLOTS(self, EV, 1) if topdown_use_fixed else(EV("IDQ_BUBBLES.CORE", 1) - EV("INT_MISC.UOP_DROPPING", 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_16:pp', 'FRONTEND_RETIRED.LATENCY_GE_8:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = ((EV("PERF_METRICS.FETCH_LATENCY", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) - EV("INT_MISC.UOP_DROPPING", 2) / SLOTS(self, EV, 2)) if topdown_use_fixed else(EV("IDQ_BUBBLES.CYCLES_0_UOPS_DELIV.CORE", 2) * Pipeline_Width - EV("INT_MISC.UOP_DROPPING", 2)) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.L2_MISS:pp', 'FRONTEND_RETIRED.L1I_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_DATA.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.STLB_MISS:pp', 'FRONTEND_RETIRED.ITLB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_TAG.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.CLEAR_RESTEER_CYCLES", 3) / CLKS(self, EV, 3) + self.Unknown_Branches.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class Mispredicts_Resteers: name = "Mispredicts_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 4) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Mispredicts_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage.""" class Clears_Resteers: name = "Clears_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'MachineClears']) maxval = None def compute(self, EV): try: self.val = (1 - Mispred_Clears_Fraction(self, EV, 4)) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Clears_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears.""" class Unknown_Branches: name = "Unknown_Branches" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['FRONTEND_RETIRED.UNKNOWN_BRANCH'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.UNKNOWN_BRANCH_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Unknown_Branches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (e.g. first time the branch is fetched or hitting BTB capacity limit) hence called Unknown Branches""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.MS_FLOWS'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("UOPS_RETIRED.MS:c1:e1", 3) / Retire_Fraction(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DECODE.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.DSB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_2:pp'] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.ANY_DSB_MISS'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.MITE_CYCLES_ANY", 3) - EV("IDQ.MITE_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Decoder0_Alone: name = "Decoder0_Alone" domain = "Slots_Estimated" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("INST_DECODED.DECODERS:c1", 4) - EV("INST_DECODED.DECODERS:c2", 4)) / CORE_CLKS(self, EV, 4) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Decoder0_Alone zero division") return self.val desc = """ This metric represents fraction of cycles where decoder-0 was the only active decoder""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.DSB_CYCLES_ANY", 3) - EV("IDQ.DSB_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = max(1 -(self.Frontend_Bound.compute(EV) + self.Backend_Bound.compute(EV) + self.Retiring.compute(EV)) , 0 ) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['TOPDOWN.BR_MISPREDICT_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.BRANCH_MISPREDICTS", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("TOPDOWN.BR_MISPREDICT_SLOTS", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Other_Mispredicts: name = "Other_Mispredicts" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'BrMispredicts']) maxval = None def compute(self, EV): try: self.val = max(self.Branch_Mispredicts.compute(EV) * (1 - EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) / (EV("INT_MISC.CLEARS_COUNT", 3) - EV("MACHINE_CLEARS.COUNT", 3))) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Mispredicts zero division") return self.val desc = """ This metric estimates fraction of slots the CPU was stalled due to other cases of misprediction (non-retired x86 branches or other types).""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Other_Nukes: name = "Other_Nukes" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'Machine_Clears']) maxval = None def compute(self, EV): try: self.val = max(self.Machine_Clears.compute(EV) * (1 - EV("MACHINE_CLEARS.MEMORY_ORDERING", 3) / EV("MACHINE_CLEARS.COUNT", 3)) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Nukes zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Nukes (Machine Clears) not related to memory ordering.""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = ['TOPDOWN.BACKEND_BOUND_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.BACKEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) if topdown_use_fixed else EV("TOPDOWN.BACKEND_BOUND_SLOTS", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.MEMORY_BOUND", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("TOPDOWN.MEMORY_BOUND_SLOTS", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT:pp', 'MEM_LOAD_RETIRED.FB_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3) - EV("MEMORY_ACTIVITY.STALLS_L1D_MISS", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = min(Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT:c1", 4) + EV("DTLB_LOAD_MISSES.WALK_ACTIVE", 4) , max(EV("CYCLE_ACTIVITY.CYCLES_MEM_ANY", 4) - EV("MEMORY_ACTIVITY.CYCLES_L1D_MISS", 4) , 0)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Load_STLB_Hit: name = "Load_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = self.DTLB_Load.compute(EV) - self.Load_STLB_Miss.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)""" class Load_STLB_Miss: name = "Load_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_LOAD_MISSES.WALK_ACTIVE", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class L1_Hit_Latency: name = "L1_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = min(2 *(EV("MEM_INST_RETIRED.ALL_LOADS", 4) - EV("MEM_LOAD_RETIRED.FB_HIT", 4) - EV("MEM_LOAD_RETIRED.L1_MISS", 4)) * Dependent_Loads_Weight(self, EV, 4) / 100 , max(EV("CYCLE_ACTIVITY.CYCLES_MEM_ANY", 4) - EV("MEMORY_ACTIVITY.CYCLES_L1D_MISS", 4) , 0)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Hit_Latency zero division") return self.val desc = """ This metric roughly estimates fraction of cycles with demand load accesses that hit the L1 cache. The short latency of the L1 data cache may be exposed in pointer-chasing memory access patterns as an example.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.LOCK_LOADS'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (16 * max(0 , EV("MEM_INST_RETIRED.LOCK_LOADS", 4) - EV("L2_RQSTS.ALL_RFO", 4)) + Mem_Lock_St_Fraction(self, EV, 4) * (Mem_L2_Store_Cost * EV("L2_RQSTS.RFO_HIT", 4) + ORO_Demand_RFO_C1(self, EV, 4))) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("LD_BLOCKS.NO_SR", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = EV("L1D_PEND_MISS.FB_FULL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("MEMORY_ACTIVITY.STALLS_L1D_MISS", 3) - EV("MEMORY_ACTIVITY.STALLS_L2_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("MEMORY_ACTIVITY.STALLS_L2_MISS", 3) - EV("MEMORY_ACTIVITY.STALLS_L3_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD', 'MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = ((Mem_XSNP_HitM_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_HITM(self, EV, 4) + (Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_MISS(self, EV, 4)) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_XSNP_HIT(self, EV, 4) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_None_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * LOAD_L3_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = (EV("XQ.FULL_CYCLES", 4) + EV("L1D_PEND_MISS.L2_STALLS", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = MEM_Bound_Ratio(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MBA_Stalls: name = "MBA_Stalls" domain = "Clocks" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBW', 'Offcore', 'Server']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.MBA_STALLS", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MBA_Stalls zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to memory bandwidth Allocation feature (RDT's memory bandwidth throttling).""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Local_MEM: name = "Local_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM'] errcount = 0 sibling = None metricgroup = frozenset(['Server']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Local_DRAM_Cost(self, EV, 5) - Mem_XSNP_None_Cost(self, EV, 5)) * EV("MEM_LOAD_L3_MISS_RETIRED.LOCAL_DRAM", 5) * FB_Factor(self, EV, 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Local_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from local memory. Caching will improve the latency and increase performance.""" class Remote_MEM: name = "Remote_MEM" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_Remote_DRAM_Cost(self, EV, 5) - Mem_XSNP_None_Cost(self, EV, 5)) * EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM", 5) * FB_Factor(self, EV, 5) / CLKS(self, EV, 5) if DS else 0 EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_DRAM", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_MEM zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote memory. This is caused often due to non-optimal NUMA allocations.""" class Remote_Cache: name = "Remote_Cache" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = ['MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM:pp', 'MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore', 'Server', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = ((Mem_Remote_HitM_Cost(self, EV, 5) - Mem_XSNP_None_Cost(self, EV, 5)) * EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM", 5) + (Mem_Remote_Fwd_Cost(self, EV, 5) - Mem_XSNP_None_Cost(self, EV, 5)) * EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD", 5)) * FB_Factor(self, EV, 5) / CLKS(self, EV, 5) if DS else 0 EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_HITM", 5) EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_FWD", 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Remote_Cache zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling loads from remote cache in other sockets including synchronizations issues. This is caused often due to non-optimal NUMA allocations.""" class PMM_Bound: name = "PMM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'Server', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = (((1 - Mem_DDR_Hit_Fraction(self, EV, 3)) * MEM_Bound_Ratio(self, EV, 3)) if (OneMillion *(EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM", 3) + EV("MEM_LOAD_RETIRED.LOCAL_PMM", 3))> EV("MEM_LOAD_RETIRED.L1_MISS", 3)) else 0) if PMM_App_Direct else 0 EV("MEM_LOAD_RETIRED.LOCAL_PMM", 3) EV("MEM_LOAD_RETIRED.L1_MISS", 3) EV("MEM_LOAD_L3_MISS_RETIRED.REMOTE_PMM", 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "PMM_Bound zero division") return self.val desc = """ This metric roughly estimates (based on idle latencies) how often the CPU was stalled on accesses to external 3D-Xpoint (Crystal Ridge, a.k.a. IXP) memory by loads, PMM stands for Persistent Memory Module. . Consider moving data-structure from AEP to DDR memory for better latency/bandwidth.""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_INST_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.BOUND_ON_STORES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.SPLIT_STORES", 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class Streaming_Stores: name = "Streaming_Stores" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.STREAMING_WR.ANY_RESPONSE'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBW', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = 9 * EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) / CLKS(self, EV, 4) if DS else 0 EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Streaming_Stores zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_STORE_MISSES.STLB_HIT:c1", 4) + EV("DTLB_STORE_MISSES.WALK_ACTIVE", 4)) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Store_STLB_Hit: name = "Store_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = self.DTLB_Store.compute(EV) - self.Store_STLB_Miss.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second- level TLB (STLB)""" class Store_STLB_Miss: name = "Store_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_STORE_MISSES.WALK_ACTIVE", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.DIVIDER_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.DIV_ACTIVE", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Serializing_Operation: name = "Serializing_Operation" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['RESOURCE_STALLS.SCOREBOARD'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SCOREBOARD", 3) / CLKS(self, EV, 3) + self.C02_WAIT.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Serializing_Operation zero division") return self.val desc = """ This metric represents fraction of cycles the CPU issue- pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out- of-order execution which may limit performance.""" class Slow_Pause: name = "Slow_Pause" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['CPU_CLK_UNHALTED.PAUSE_INST'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.PAUSE", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Slow_Pause zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions.""" class C01_WAIT: name = "C01_WAIT" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['C0Wait']) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.C01", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "C01_WAIT zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due staying in C0.1 power-performance optimized state (Faster wakeup time; Smaller power savings).""" class C02_WAIT: name = "C02_WAIT" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['C0Wait']) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.C02", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "C02_WAIT zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due staying in C0.2 power-performance optimized state (Slower wakeup time; Larger power savings).""" class Memory_Fence: name = "Memory_Fence" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("MISC2_RETIRED.LFENCE", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Fence zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to LFENCE Instructions.""" class AMX_Busy: name = "AMX_Busy" domain = "Core_Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'Compute', 'HPC', 'Server']) maxval = None def compute(self, EV): try: self.val = EV("EXE.AMX_BUSY", 3) / CORE_CLKS(self, EV, 3) self.thresh = (self.val > 0.5) and self.parent.thresh except ZeroDivisionError: handle_error(self, "AMX_Busy zero division") return self.val desc = """ This metric estimates fraction of cycles where the Advanced Matrix eXtensions (AMX) execution engine was busy with tile (arithmetic) operations""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Core_Bound_Cycles(self, EV, 3) / CLKS(self, EV, 3) if (EV("ARITH.DIV_ACTIVE", 3)<(EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) - EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3))) else Few_Uops_Executed_Threshold(self, EV, 3) / CLKS(self, EV, 3) EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3) EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) EV("ARITH.DIV_ACTIVE", 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = (EV("EXE_ACTIVITY.EXE_BOUND_0_PORTS", 4) + max(EV("RS.EMPTY:u1", 4) - EV("RESOURCE_STALLS.SCOREBOARD", 4) , 0)) / CLKS(self, EV, 4) * (EV("CYCLE_ACTIVITY.STALLS_TOTAL", 4) - EV("EXE_ACTIVITY.BOUND_ON_LOADS", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Mixing_Vectors: name = "Mixing_Vectors" domain = "Clocks" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 160 * EV("ASSISTS.SSE_AVX_MIX", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error(self, "Mixing_Vectors zero division") return self.val desc = """ This metric estimates penalty in terms of percentage of cycles. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.1_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.1_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.2_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.2_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['UOPS_EXECUTED.CYCLES_GE_3'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_EXECUTED.CYCLES_GE_3", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_0", 5) + EV("UOPS_DISPATCHED.PORT_1", 5) + EV("UOPS_DISPATCHED.PORT_5_11", 5) + EV("UOPS_DISPATCHED.PORT_6", 5)) / (5 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU and 2nd branch""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_6: name = "Port_6" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_6'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_6", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_6 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 Primary Branch and simple ALU""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_2_3_10'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_2_3_10", 5) / (3 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_7_8'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_4_9", 5) + EV("UOPS_DISPATCHED.PORT_7_8", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.RETIRING", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) if topdown_use_fixed else EV("UOPS_RETIRED.SLOTS", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. .. Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) * EV("UOPS_EXECUTED.X87", 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5) + EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", 5)) / Retired_Slots(self, EV, 5) if FP16 else(EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5) EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) EV("FP_ARITH_INST_RETIRED2.128B_PACKED_HALF", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5) + EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", 5)) / Retired_Slots(self, EV, 5) if FP16 else(EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5) EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) EV("FP_ARITH_INST_RETIRED2.256B_PACKED_HALF", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_512b: name = "FP_Vector_512b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", 5) + EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", 5)) / Retired_Slots(self, EV, 5) if FP16 else(EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) EV("FP_ARITH_INST_RETIRED.512B_PACKED_SINGLE", 5) EV("FP_ARITH_INST_RETIRED2.512B_PACKED_HALF", 5) EV("FP_ARITH_INST_RETIRED.512B_PACKED_DOUBLE", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_512b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 512-bit wide vectors. May overcount due to FMA double counting.""" class Int_Operations: name = "Int_Operations" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Int_Vector_128b.compute(EV) + self.Int_Vector_256b.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Operations zero division") return self.val desc = """ This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired). Vector/Matrix Int operations and shuffles are counted. Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain.""" class Int_Vector_128b: name = "Int_Vector_128b" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'IntVector', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = (EV("INT_VEC_RETIRED.ADD_128", 4) + EV("INT_VEC_RETIRED.VNNI_128", 4)) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Vector_128b zero division") return self.val desc = """ This metric represents 128-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.""" class Int_Vector_256b: name = "Int_Vector_256b" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'IntVector', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = (EV("INT_VEC_RETIRED.ADD_256", 4) + EV("INT_VEC_RETIRED.MUL_256", 4) + EV("INT_VEC_RETIRED.VNNI_256", 4)) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Vector_256b zero division") return self.val desc = """ This metric represents 256-bit vector Integer ADD/SUB/SAD/MUL or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.""" class Memory_Operations: name = "Memory_Operations" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("MEM_UOP_RETIRED.ANY", 3) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.""" class Fused_Instructions: name = "Fused_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Branches', 'BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INST_RETIRED.MACRO_FUSED", 3) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fused_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. CMP+JCC or DEC+JCC are common examples of legacy fusions. {}. See section 'Optimizing for Macro-fusion' in Optimization Manual:""" class Non_Fused_Branches: name = "Non_Fused_Branches" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Branches', 'BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * (EV("BR_INST_RETIRED.ALL_BRANCHES", 3) - EV("INST_RETIRED.MACRO_FUSED", 3)) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Non_Fused_Branches zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non- conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.""" class Other_Light_Ops: name = "Other_Light_Ops" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Light_Operations.compute(EV) - Light_Ops_Sum(self, EV, 3)) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Light_Ops zero division") return self.val desc = """ This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting""" class Nop_Instructions: name = "Nop_Instructions" domain = "Slots" area = "RET" level = 4 htoff = False sample = ['INST_RETIRED.NOP'] errcount = 0 sibling = None metricgroup = frozenset(['BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INST_RETIRED.NOP", 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Nop_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.. Improve Codegen by correctly placing NOPs outside hot sections (e.g. outside loop body).""" class Shuffles_256b: name = "Shuffles_256b" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INT_VEC_RETIRED.SHUFFLES", 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Shuffles_256b zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring Shuffle operations of 256-bit vector size (FP or Integer). Shuffles may incur slow cross \"vector lane\" data transfers.""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['UOPS_RETIRED.HEAVY'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.HEAVY_OPERATIONS", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("UOPS_RETIRED.HEAVY", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences. .""" class Few_Uops_Instructions: name = "Few_Uops_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Heavy_Operations.compute(EV) - self.Microcode_Sequencer.compute(EV)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Few_Uops_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to five uops. This highly-correlates with the number of uops in such instructions.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['UOPS_RETIRED.MS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_RETIRED.MS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['ASSISTS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("ASSISTS.ANY", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class Page_Faults: name = "Page_Faults" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 99 * EV("ASSISTS.PAGE_FAULT", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error(self, "Page_Faults zero division") return self.val desc = """ This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults. A Page Fault may apply on first application access to a memory page. Note operating system handling of page faults accounts for the majority of its cost.""" class FP_Assists: name = "FP_Assists" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = 30 * EV("ASSISTS.FP", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "FP_Assists zero division") return self.val desc = """ This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists. FP Assist may apply when working with very small floating point values (so-called Denormals).. Consider DAZ (Denormals Are Zero) and/or FTZ (Flush To Zero) options in your compiler; \"-ffast-math\" with -O2 in GCC for example. This option may improve performance if the denormal values are not critical in your application. Also note that the DAZ and FTZ modes are not compatible with the IEEE Standard 754.. https://www.intel.com/content/www/us/en/develop/docume ntation/vtune-help/top/reference/cpu-metrics-reference/bad- speculation-back-end-bound-pipeline-slots/fp-assists.html""" class AVX_Assists: name = "AVX_Assists" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = 63 * EV("ASSISTS.SSE_AVX_MIX", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "AVX_Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops as a result of handing SSE to AVX* or AVX* to SSE transition Assists.""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = ['FRONTEND_RETIRED.MS_FLOWS'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_Mispredictions: name = "Mispredictions" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts', 'BvMP']) sibling = None def compute(self, EV): try: self.val = Mispredictions(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Mispredictions zero division") desc = """ Total pipeline cost of Branch Misprediction related bottlenecks""" class Metric_Big_Code: name = "Big_Code" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBC', 'BigFootprint', 'Fed', 'Frontend', 'IcMiss', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Big_Code(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Big_Code zero division") desc = """ Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)""" class Metric_Instruction_Fetch_BW: name = "Instruction_Fetch_BW" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvFB', 'Fed', 'FetchBW', 'Frontend']) sibling = None def compute(self, EV): try: self.val = Instruction_Fetch_BW(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Instruction_Fetch_BW zero division") desc = """ Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end)""" class Metric_Cache_Memory_Bandwidth: name = "Cache_Memory_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMB', 'Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Bandwidth(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Bandwidth zero division") desc = """ Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks""" class Metric_Cache_Memory_Latency: name = "Cache_Memory_Latency" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvML', 'Mem', 'MemoryLat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Latency(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Latency zero division") desc = """ Total pipeline cost of external Memory- or Cache-Latency related bottlenecks""" class Metric_Memory_Data_TLBs: name = "Memory_Data_TLBs" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMT', 'Mem', 'MemoryTLB', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Data_TLBs(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Memory_Data_TLBs zero division") desc = """ Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)""" class Metric_Memory_Synchronization: name = "Memory_Synchronization" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMS', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Synchronization(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Memory_Synchronization zero division") desc = """ Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors)""" class Metric_Compute_Bound_Est: name = "Compute_Bound_Est" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvCB', 'Cor']) sibling = None def compute(self, EV): try: self.val = Compute_Bound_Est(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Compute_Bound_Est zero division") desc = """ Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy.""" class Metric_Irregular_Overhead: name = "Irregular_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BvIO', 'Cor', 'Ret']) sibling = None def compute(self, EV): try: self.val = Irregular_Overhead(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Irregular_Overhead zero division") desc = """ Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments)""" class Metric_Other_Bottlenecks: name = "Other_Bottlenecks" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvOB', 'Cor', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Other_Bottlenecks(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Other_Bottlenecks zero division") desc = """ Total pipeline cost of remaining bottlenecks in the back- end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls.""" class Metric_Branching_Overhead: name = "Branching_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBO', 'Ret']) sibling = None def compute(self, EV): try: self.val = Branching_Overhead(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "Branching_Overhead zero division") desc = """ Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations""" class Metric_Useful_Work: name = "Useful_Work" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvUW', 'Ret']) sibling = None def compute(self, EV): try: self.val = Useful_Work(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Useful_Work zero division") desc = """ Total pipeline cost of \"useful operations\" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead.""" class Metric_Core_Bound_Likely: name = "Core_Bound_Likely" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Botlnk.L0" metricgroup = frozenset(['Cor', 'SMT']) sibling = None def compute(self, EV): try: self.val = Core_Bound_Likely(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Core_Bound_Likely zero division") desc = """ Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Slots_Utilization: name = "Slots_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Slots_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Slots_Utilization zero division") desc = """ Fraction of Physical Core issue-slots utilized by this Logical Processor""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_FP_Arith_Utilization: name = "FP_Arith_Utilization" domain = "Core_Metric" maxval = 2.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = FP_Arith_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FP_Arith_Utilization zero division") desc = """ Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common.""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_EPC: name = "EPC" domain = "Metric" maxval = 20.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = EPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "EPC zero division") desc = """ uops Executed per Cycle""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpFLOP: name = "IpFLOP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpFLOP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpFLOP zero division") desc = """ Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_IpArith_Scalar_HP: name = "IpArith_Scalar_HP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType', 'Server']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_HP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_HP zero division") desc = """ Instructions per FP Arithmetic Scalar Half-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_SP: name = "IpArith_Scalar_SP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_SP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_SP zero division") desc = """ Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_DP: name = "IpArith_Scalar_DP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_DP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_DP zero division") desc = """ Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX128: name = "IpArith_AVX128" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX128(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX128 zero division") desc = """ Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX256: name = "IpArith_AVX256" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX256(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX256 zero division") desc = """ Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX512: name = "IpArith_AVX512" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX512(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX512 zero division") desc = """ Instructions per FP Arithmetic AVX 512-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpPause: name = "IpPause" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpPause(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpPause zero division") desc = """ Instructions per PAUSE (lower number means higher occurrence rate)""" class Metric_IpSWPF: name = "IpSWPF" domain = "Inst_Metric" maxval = 1000 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Prefetches']) sibling = None def compute(self, EV): try: self.val = IpSWPF(self, EV, 0) self.thresh = (self.val < 100) except ZeroDivisionError: handle_error_metric(self, "IpSWPF zero division") desc = """ Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate)""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_Strings_Cycles: name = "Strings_Cycles" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['MicroSeq', 'Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Strings_Cycles(self, EV, 0) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error_metric(self, "Strings_Cycles zero division") desc = """ Estimated fraction of retirement-cycles dealing with repeat instructions""" class Metric_IpAssist: name = "IpAssist" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['MicroSeq', 'Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = IpAssist(self, EV, 0) self.thresh = (self.val < 100000) except ZeroDivisionError: handle_error_metric(self, "IpAssist zero division") desc = """ Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate)""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_Fetch_LSD: name = "Fetch_LSD" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_LSD(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_LSD zero division") desc = """ Average number of uops fetched from LSD per cycle""" class Metric_Fetch_DSB: name = "Fetch_DSB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_DSB(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_DSB zero division") desc = """ Average number of uops fetched from DSB per cycle""" class Metric_Fetch_MITE: name = "Fetch_MITE" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_MITE(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_MITE zero division") desc = """ Average number of uops fetched from MITE per cycle""" class Metric_Fetch_UpC: name = "Fetch_UpC" domain = "Metric" maxval = 6.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_UpC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_UpC zero division") desc = """ Average number of Uops issued by front-end when it issued something""" class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_Unknown_Branch_Cost: name = "Unknown_Branch_Cost" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = Unknown_Branch_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Unknown_Branch_Cost zero division") desc = """ Average number of cycles the front-end was delayed due to an Unknown Branch detection. See Unknown_Branches node.""" class Metric_DSB_Switch_Cost: name = "DSB_Switch_Cost" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss']) sibling = None def compute(self, EV): try: self.val = DSB_Switch_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DSB_Switch_Cost zero division") desc = """ Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details.""" class Metric_DSB_Misses: name = "DSB_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = DSB_Misses(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Misses zero division") desc = """ Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_DSB_Bandwidth: name = "DSB_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSB', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Bandwidth(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Bandwidth zero division") desc = """ Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_ICache_Miss_Latency: name = "ICache_Miss_Latency" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = ICache_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ICache_Miss_Latency zero division") desc = """ Average Latency for L1 instruction cache misses""" class Metric_IC_Misses: name = "IC_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = IC_Misses(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "IC_Misses zero division") desc = """ Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck.""" class Metric_IpDSB_Miss_Ret: name = "IpDSB_Miss_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = IpDSB_Miss_Ret(self, EV, 0) self.thresh = (self.val < 50) except ZeroDivisionError: handle_error_metric(self, "IpDSB_Miss_Ret zero division") desc = """ Instructions per non-speculative DSB miss (lower number means higher occurrence rate)""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_L2MPKI_Code: name = "L2MPKI_Code" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code zero division") desc = """ L2 cache true code cacheline misses per kilo instruction""" class Metric_L2MPKI_Code_All: name = "L2MPKI_Code_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code_All zero division") desc = """ L2 cache speculative code cacheline misses per kilo instruction""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Cond_Ntaken: name = "IpMisp_Cond_Ntaken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Ntaken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Ntaken zero division") desc = """ Instructions per retired Mispredicts for conditional non- taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Cond_Taken: name = "IpMisp_Cond_Taken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Taken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Taken zero division") desc = """ Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Ret: name = "IpMisp_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Ret(self, EV, 0) self.thresh = (self.val < 500) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Ret zero division") desc = """ Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Branch_Misprediction_Cost: name = "Branch_Misprediction_Cost" domain = "Core_Metric" maxval = 300 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Branch_Misprediction_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Branch_Misprediction_Cost zero division") desc = """ Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)""" class Metric_Spec_Clears_Ratio: name = "Spec_Clears_Ratio" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Spec_Clears_Ratio(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Spec_Clears_Ratio zero division") desc = """ Speculative to Retired ratio of all clears (covering Mispredicts and nukes)""" class Metric_Cond_NT: name = "Cond_NT" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_NT(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_NT zero division") desc = """ Fraction of branches that are non-taken conditionals""" class Metric_Cond_TK: name = "Cond_TK" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_TK(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_TK zero division") desc = """ Fraction of branches that are taken conditionals""" class Metric_CallRet: name = "CallRet" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = CallRet(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CallRet zero division") desc = """ Fraction of branches that are CALL or RET""" class Metric_Jump: name = "Jump" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Jump(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Jump zero division") desc = """ Fraction of branches that are unconditional (direct or indirect) jumps""" class Metric_Other_Branches: name = "Other_Branches" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Other_Branches(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Other_Branches zero division") desc = """ Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L1MPKI_Load: name = "L1MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI_Load zero division") desc = """ L1 cache true misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_All: name = "L2MPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_All zero division") desc = """ L2 cache true misses per kilo instruction for all request types (including speculative)""" class Metric_L2MPKI_Load: name = "L2MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Load zero division") desc = """ L2 cache true misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L2HPKI_All: name = "L2HPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_All zero division") desc = """ L2 cache hits per kilo instruction for all request types (including speculative)""" class Metric_L2HPKI_Load: name = "L2HPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_Load zero division") desc = """ L2 cache hits per kilo instruction for all demand loads (including speculative)""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_FB_HPKI: name = "FB_HPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = FB_HPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FB_HPKI zero division") desc = """ Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss- handling entries)""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Access_BW: name = "L3_Cache_Access_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_Code_STLB_MPKI: name = "Code_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Fed', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Code_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Code_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Load_STLB_MPKI: name = "Load_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Load_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Store_STLB_MPKI: name = "Store_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Store_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Store_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_L3_Cache_Access_BW_2T: name = "L3_Cache_Access_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW_2T zero division") desc = """ Average per-core data access bandwidth to the L3 cache [GB / sec]""" class Metric_L2_Evictions_Silent_PKI: name = "L2_Evictions_Silent_PKI" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['L2Evicts', 'Mem', 'Server']) sibling = None def compute(self, EV): try: self.val = L2_Evictions_Silent_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Evictions_Silent_PKI zero division") desc = """ Rate of silent evictions from the L2 cache per Kilo instruction where the evicted lines are dropped (no writeback to L3 or memory)""" class Metric_L2_Evictions_NonSilent_PKI: name = "L2_Evictions_NonSilent_PKI" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['L2Evicts', 'Mem', 'Server']) sibling = None def compute(self, EV): try: self.val = L2_Evictions_NonSilent_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Evictions_NonSilent_PKI zero division") desc = """ Rate of non silent evictions from the L2 cache per Kilo instruction""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L3_Miss_Latency: name = "Load_L3_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L3_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L3_Miss_Latency zero division") desc = """ Average Latency for L3 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_Offcore_Read_Any_PKI: name = "Offcore_Read_Any_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['CacheHits', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Offcore_Read_Any_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Offcore_Read_Any_PKI zero division") desc = """ Off-core accesses per kilo instruction for reads-to-core requests (speculative; including in-core HW prefetches)""" class Metric_Offcore_Read_L3M_PKI: name = "Offcore_Read_L3M_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Offcore']) sibling = None def compute(self, EV): try: self.val = Offcore_Read_L3M_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Offcore_Read_L3M_PKI zero division") desc = """ L3 cache misses per kilo instruction for reads-to-core requests (speculative; including in-core HW prefetches)""" class Metric_Offcore_MWrite_Any_PKI: name = "Offcore_MWrite_Any_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Offcore']) sibling = None def compute(self, EV): try: self.val = Offcore_MWrite_Any_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Offcore_MWrite_Any_PKI zero division") desc = """ Off-core accesses per kilo instruction for modified write requests""" class Metric_UC_Load_PKI: name = "UC_Load_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = UC_Load_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "UC_Load_PKI zero division") desc = """ Un-cacheable retired load per kilo instruction""" class Metric_Bus_Lock_PKI: name = "Bus_Lock_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = Bus_Lock_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Bus_Lock_PKI zero division") desc = """ \"Bus lock\" per kilo instruction""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_Uncore_Frequency: name = "Uncore_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Uncore_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Uncore_Frequency zero division") desc = """ Measured Average Uncore Frequency for the SoC [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_C0_Wait: name = "C0_Wait" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['C0Wait']) sibling = None def compute(self, EV): try: self.val = C0_Wait(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "C0_Wait zero division") desc = """ Fraction of cycles the processor is waiting yet unhalted; covering legacy PAUSE instruction, as well as C0.1 / C0.2 power-performance optimized states. Sample code of TPAUSE: h ttps://github.com/torvalds/linux/blob/master/arch/x86/lib/de lay.c""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_R2C_Offcore_BW: name = "R2C_Offcore_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.Memory.SoC" metricgroup = frozenset(['HPC', 'Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = R2C_Offcore_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "R2C_Offcore_BW zero division") desc = """ Average Off-core access BW for Reads-to-Core (R2C). R2C account for demand or prefetch load/RFO/code access that fill data into the Core caches.""" class Metric_R2C_L3M_BW: name = "R2C_L3M_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.Memory.SoC" metricgroup = frozenset(['HPC', 'Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = R2C_L3M_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "R2C_L3M_BW zero division") desc = """ Average L3-cache miss BW for Reads-to-Core (R2C). This covering going to DRAM or other memory off-chip memory tears. See R2C_Offcore_BW.""" class Metric_R2C_DRAM_BW: name = "R2C_DRAM_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.Memory.SoC" metricgroup = frozenset(['HPC', 'Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = R2C_DRAM_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "R2C_DRAM_BW zero division") desc = """ Average DRAM BW for Reads-to-Core (R2C) covering for memory attached to local- and remote-socket. See R2C_Offcore_BW.""" class Metric_MEM_Read_Latency: name = "MEM_Read_Latency" domain = "NanoSeconds" maxval = 1000 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryLat', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Read_Latency zero division") desc = """ Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches. memory-controller only""" class Metric_MEM_Parallel_Reads: name = "MEM_Parallel_Reads" domain = "SystemMetric" maxval = 100 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Parallel_Reads(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Parallel_Reads zero division") desc = """ Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches""" class Metric_MEM_PMM_Read_Latency: name = "MEM_PMM_Read_Latency" domain = "NanoSeconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['MemOffcore', 'MemoryLat', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = MEM_PMM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_PMM_Read_Latency zero division") desc = """ Average latency of data read request to external 3D X-Point memory [in nanoseconds]. Accounts for demand loads and L1/L2 data-read prefetches""" class Metric_MEM_DRAM_Read_Latency: name = "MEM_DRAM_Read_Latency" domain = "NanoSeconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['MemOffcore', 'MemoryLat', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = MEM_DRAM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_DRAM_Read_Latency zero division") desc = """ Average latency of data read request to external DRAM memory [in nanoseconds]. Accounts for demand loads and L1/L2 data- read prefetches""" class Metric_PMM_Read_BW: name = "PMM_Read_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['MemOffcore', 'MemoryBW', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = PMM_Read_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "PMM_Read_BW zero division") desc = """ Average 3DXP Memory Bandwidth Use for reads [GB / sec]""" class Metric_PMM_Write_BW: name = "PMM_Write_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['MemOffcore', 'MemoryBW', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = PMM_Write_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "PMM_Write_BW zero division") desc = """ Average 3DXP Memory Bandwidth Use for Writes [GB / sec]""" class Metric_IO_Read_BW: name = "IO_Read_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['IoBW', 'MemOffcore', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = IO_Read_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IO_Read_BW zero division") desc = """ Average IO (network or disk) Bandwidth Use for Reads [GB / sec]. Bandwidth of IO reads that are initiated by end device controllers that are requesting memory from the CPU""" class Metric_IO_Write_BW: name = "IO_Write_BW" domain = "GB/sec" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['IoBW', 'MemOffcore', 'SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = IO_Write_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IO_Write_BW zero division") desc = """ Average IO (network or disk) Bandwidth Use for Writes [GB / sec]. Bandwidth of IO writes that are initiated by end device controllers that are writing memory to the CPU""" class Metric_UPI_Data_Transmit_BW: name = "UPI_Data_Transmit_BW" domain = "MB/sec" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC', 'Server']) sibling = None def compute(self, EV): try: self.val = UPI_Data_Transmit_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "UPI_Data_Transmit_BW zero division") desc = """ Cross-socket Ultra Path Interconnect (UPI) data transmit bandwidth for data only [MB / sec]""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = Mispredicts_Resteers() ; r.run(n) ; o["Mispredicts_Resteers"] = n n = Clears_Resteers() ; r.run(n) ; o["Clears_Resteers"] = n n = Unknown_Branches() ; r.run(n) ; o["Unknown_Branches"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = Decoder0_Alone() ; r.run(n) ; o["Decoder0_Alone"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Other_Mispredicts() ; r.run(n) ; o["Other_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Other_Nukes() ; r.run(n) ; o["Other_Nukes"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Load_STLB_Hit() ; r.run(n) ; o["Load_STLB_Hit"] = n n = Load_STLB_Miss() ; r.run(n) ; o["Load_STLB_Miss"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = L1_Hit_Latency() ; r.run(n) ; o["L1_Hit_Latency"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MBA_Stalls() ; r.run(n) ; o["MBA_Stalls"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Local_MEM() ; r.run(n) ; o["Local_MEM"] = n n = Remote_MEM() ; r.run(n) ; o["Remote_MEM"] = n n = Remote_Cache() ; r.run(n) ; o["Remote_Cache"] = n n = PMM_Bound() ; r.run(n) ; o["PMM_Bound"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = Streaming_Stores() ; r.run(n) ; o["Streaming_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Store_STLB_Hit() ; r.run(n) ; o["Store_STLB_Hit"] = n n = Store_STLB_Miss() ; r.run(n) ; o["Store_STLB_Miss"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Serializing_Operation() ; r.run(n) ; o["Serializing_Operation"] = n n = Slow_Pause() ; r.run(n) ; o["Slow_Pause"] = n n = C01_WAIT() ; r.run(n) ; o["C01_WAIT"] = n n = C02_WAIT() ; r.run(n) ; o["C02_WAIT"] = n n = Memory_Fence() ; r.run(n) ; o["Memory_Fence"] = n n = AMX_Busy() ; r.run(n) ; o["AMX_Busy"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Mixing_Vectors() ; r.run(n) ; o["Mixing_Vectors"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_6() ; r.run(n) ; o["Port_6"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = FP_Vector_512b() ; r.run(n) ; o["FP_Vector_512b"] = n n = Int_Operations() ; r.run(n) ; o["Int_Operations"] = n n = Int_Vector_128b() ; r.run(n) ; o["Int_Vector_128b"] = n n = Int_Vector_256b() ; r.run(n) ; o["Int_Vector_256b"] = n n = Memory_Operations() ; r.run(n) ; o["Memory_Operations"] = n n = Fused_Instructions() ; r.run(n) ; o["Fused_Instructions"] = n n = Non_Fused_Branches() ; r.run(n) ; o["Non_Fused_Branches"] = n n = Other_Light_Ops() ; r.run(n) ; o["Other_Light_Ops"] = n n = Nop_Instructions() ; r.run(n) ; o["Nop_Instructions"] = n n = Shuffles_256b() ; r.run(n) ; o["Shuffles_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Few_Uops_Instructions() ; r.run(n) ; o["Few_Uops_Instructions"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = Page_Faults() ; r.run(n) ; o["Page_Faults"] = n n = FP_Assists() ; r.run(n) ; o["FP_Assists"] = n n = AVX_Assists() ; r.run(n) ; o["AVX_Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["Mispredicts_Resteers"].parent = o["Branch_Resteers"] o["Clears_Resteers"].parent = o["Branch_Resteers"] o["Unknown_Branches"].parent = o["Branch_Resteers"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["Decoder0_Alone"].parent = o["MITE"] o["DSB"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Other_Mispredicts"].parent = o["Branch_Mispredicts"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Other_Nukes"].parent = o["Machine_Clears"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Load_STLB_Hit"].parent = o["DTLB_Load"] o["Load_STLB_Miss"].parent = o["DTLB_Load"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["L1_Hit_Latency"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MBA_Stalls"].parent = o["MEM_Bandwidth"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Local_MEM"].parent = o["MEM_Latency"] o["Remote_MEM"].parent = o["MEM_Latency"] o["Remote_Cache"].parent = o["MEM_Latency"] o["PMM_Bound"].parent = o["Memory_Bound"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["Streaming_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Store_STLB_Hit"].parent = o["DTLB_Store"] o["Store_STLB_Miss"].parent = o["DTLB_Store"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Serializing_Operation"].parent = o["Core_Bound"] o["Slow_Pause"].parent = o["Serializing_Operation"] o["C01_WAIT"].parent = o["Serializing_Operation"] o["C02_WAIT"].parent = o["Serializing_Operation"] o["Memory_Fence"].parent = o["Serializing_Operation"] o["AMX_Busy"].parent = o["Core_Bound"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Mixing_Vectors"].parent = o["Ports_Utilized_0"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_6"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["FP_Vector_512b"].parent = o["FP_Vector"] o["Int_Operations"].parent = o["Light_Operations"] o["Int_Vector_128b"].parent = o["Int_Operations"] o["Int_Vector_256b"].parent = o["Int_Operations"] o["Memory_Operations"].parent = o["Light_Operations"] o["Fused_Instructions"].parent = o["Light_Operations"] o["Non_Fused_Branches"].parent = o["Light_Operations"] o["Other_Light_Ops"].parent = o["Light_Operations"] o["Nop_Instructions"].parent = o["Other_Light_Ops"] o["Shuffles_256b"].parent = o["Other_Light_Ops"] o["Heavy_Operations"].parent = o["Retiring"] o["Few_Uops_Instructions"].parent = o["Heavy_Operations"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["Page_Faults"].parent = o["Assists"] o["FP_Assists"].parent = o["Assists"] o["AVX_Assists"].parent = o["Assists"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_Mispredictions() ; r.metric(n) ; o["Mispredictions"] = n n = Metric_Big_Code() ; r.metric(n) ; o["Big_Code"] = n n = Metric_Instruction_Fetch_BW() ; r.metric(n) ; o["Instruction_Fetch_BW"] = n n = Metric_Cache_Memory_Bandwidth() ; r.metric(n) ; o["Cache_Memory_Bandwidth"] = n n = Metric_Cache_Memory_Latency() ; r.metric(n) ; o["Cache_Memory_Latency"] = n n = Metric_Memory_Data_TLBs() ; r.metric(n) ; o["Memory_Data_TLBs"] = n n = Metric_Memory_Synchronization() ; r.metric(n) ; o["Memory_Synchronization"] = n n = Metric_Compute_Bound_Est() ; r.metric(n) ; o["Compute_Bound_Est"] = n n = Metric_Irregular_Overhead() ; r.metric(n) ; o["Irregular_Overhead"] = n n = Metric_Other_Bottlenecks() ; r.metric(n) ; o["Other_Bottlenecks"] = n n = Metric_Branching_Overhead() ; r.metric(n) ; o["Branching_Overhead"] = n n = Metric_Useful_Work() ; r.metric(n) ; o["Useful_Work"] = n n = Metric_Core_Bound_Likely() ; r.metric(n) ; o["Core_Bound_Likely"] = n n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Slots_Utilization() ; r.metric(n) ; o["Slots_Utilization"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_FP_Arith_Utilization() ; r.metric(n) ; o["FP_Arith_Utilization"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_EPC() ; r.metric(n) ; o["EPC"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpFLOP() ; r.metric(n) ; o["IpFLOP"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_IpArith_Scalar_HP() ; r.metric(n) ; o["IpArith_Scalar_HP"] = n n = Metric_IpArith_Scalar_SP() ; r.metric(n) ; o["IpArith_Scalar_SP"] = n n = Metric_IpArith_Scalar_DP() ; r.metric(n) ; o["IpArith_Scalar_DP"] = n n = Metric_IpArith_AVX128() ; r.metric(n) ; o["IpArith_AVX128"] = n n = Metric_IpArith_AVX256() ; r.metric(n) ; o["IpArith_AVX256"] = n n = Metric_IpArith_AVX512() ; r.metric(n) ; o["IpArith_AVX512"] = n n = Metric_IpPause() ; r.metric(n) ; o["IpPause"] = n n = Metric_IpSWPF() ; r.metric(n) ; o["IpSWPF"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_Strings_Cycles() ; r.metric(n) ; o["Strings_Cycles"] = n n = Metric_IpAssist() ; r.metric(n) ; o["IpAssist"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_Fetch_LSD() ; r.metric(n) ; o["Fetch_LSD"] = n n = Metric_Fetch_DSB() ; r.metric(n) ; o["Fetch_DSB"] = n n = Metric_Fetch_MITE() ; r.metric(n) ; o["Fetch_MITE"] = n n = Metric_Fetch_UpC() ; r.metric(n) ; o["Fetch_UpC"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_Unknown_Branch_Cost() ; r.metric(n) ; o["Unknown_Branch_Cost"] = n n = Metric_DSB_Switch_Cost() ; r.metric(n) ; o["DSB_Switch_Cost"] = n n = Metric_DSB_Misses() ; r.metric(n) ; o["DSB_Misses"] = n n = Metric_DSB_Bandwidth() ; r.metric(n) ; o["DSB_Bandwidth"] = n n = Metric_ICache_Miss_Latency() ; r.metric(n) ; o["ICache_Miss_Latency"] = n n = Metric_IC_Misses() ; r.metric(n) ; o["IC_Misses"] = n n = Metric_IpDSB_Miss_Ret() ; r.metric(n) ; o["IpDSB_Miss_Ret"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_L2MPKI_Code() ; r.metric(n) ; o["L2MPKI_Code"] = n n = Metric_L2MPKI_Code_All() ; r.metric(n) ; o["L2MPKI_Code_All"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Cond_Ntaken() ; r.metric(n) ; o["IpMisp_Cond_Ntaken"] = n n = Metric_IpMisp_Cond_Taken() ; r.metric(n) ; o["IpMisp_Cond_Taken"] = n n = Metric_IpMisp_Ret() ; r.metric(n) ; o["IpMisp_Ret"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Branch_Misprediction_Cost() ; r.metric(n) ; o["Branch_Misprediction_Cost"] = n n = Metric_Spec_Clears_Ratio() ; r.metric(n) ; o["Spec_Clears_Ratio"] = n n = Metric_Cond_NT() ; r.metric(n) ; o["Cond_NT"] = n n = Metric_Cond_TK() ; r.metric(n) ; o["Cond_TK"] = n n = Metric_CallRet() ; r.metric(n) ; o["CallRet"] = n n = Metric_Jump() ; r.metric(n) ; o["Jump"] = n n = Metric_Other_Branches() ; r.metric(n) ; o["Other_Branches"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L1MPKI_Load() ; r.metric(n) ; o["L1MPKI_Load"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_All() ; r.metric(n) ; o["L2MPKI_All"] = n n = Metric_L2MPKI_Load() ; r.metric(n) ; o["L2MPKI_Load"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L2HPKI_All() ; r.metric(n) ; o["L2HPKI_All"] = n n = Metric_L2HPKI_Load() ; r.metric(n) ; o["L2HPKI_Load"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_FB_HPKI() ; r.metric(n) ; o["FB_HPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_L3_Cache_Access_BW() ; r.metric(n) ; o["L3_Cache_Access_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_Code_STLB_MPKI() ; r.metric(n) ; o["Code_STLB_MPKI"] = n n = Metric_Load_STLB_MPKI() ; r.metric(n) ; o["Load_STLB_MPKI"] = n n = Metric_Store_STLB_MPKI() ; r.metric(n) ; o["Store_STLB_MPKI"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Access_BW_2T() ; r.metric(n) ; o["L3_Cache_Access_BW_2T"] = n n = Metric_L2_Evictions_Silent_PKI() ; r.metric(n) ; o["L2_Evictions_Silent_PKI"] = n n = Metric_L2_Evictions_NonSilent_PKI() ; r.metric(n) ; o["L2_Evictions_NonSilent_PKI"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L3_Miss_Latency() ; r.metric(n) ; o["Load_L3_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_Offcore_Read_Any_PKI() ; r.metric(n) ; o["Offcore_Read_Any_PKI"] = n n = Metric_Offcore_Read_L3M_PKI() ; r.metric(n) ; o["Offcore_Read_L3M_PKI"] = n n = Metric_Offcore_MWrite_Any_PKI() ; r.metric(n) ; o["Offcore_MWrite_Any_PKI"] = n n = Metric_UC_Load_PKI() ; r.metric(n) ; o["UC_Load_PKI"] = n n = Metric_Bus_Lock_PKI() ; r.metric(n) ; o["Bus_Lock_PKI"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_Uncore_Frequency() ; r.metric(n) ; o["Uncore_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_C0_Wait() ; r.metric(n) ; o["C0_Wait"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_R2C_Offcore_BW() ; r.metric(n) ; o["R2C_Offcore_BW"] = n n = Metric_R2C_L3M_BW() ; r.metric(n) ; o["R2C_L3M_BW"] = n n = Metric_R2C_DRAM_BW() ; r.metric(n) ; o["R2C_DRAM_BW"] = n n = Metric_MEM_Read_Latency() ; r.metric(n) ; o["MEM_Read_Latency"] = n n = Metric_MEM_Parallel_Reads() ; r.metric(n) ; o["MEM_Parallel_Reads"] = n n = Metric_MEM_PMM_Read_Latency() ; r.metric(n) ; o["MEM_PMM_Read_Latency"] = n n = Metric_MEM_DRAM_Read_Latency() ; r.metric(n) ; o["MEM_DRAM_Read_Latency"] = n n = Metric_PMM_Read_BW() ; r.metric(n) ; o["PMM_Read_BW"] = n n = Metric_PMM_Write_BW() ; r.metric(n) ; o["PMM_Write_BW"] = n n = Metric_IO_Read_BW() ; r.metric(n) ; o["IO_Read_BW"] = n n = Metric_IO_Write_BW() ; r.metric(n) ; o["IO_Write_BW"] = n n = Metric_UPI_Data_Transmit_BW() ; r.metric(n) ; o["UPI_Data_Transmit_BW"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Branch_Resteers"].Unknown_Branches = o["Unknown_Branches"] o["Mispredicts_Resteers"].Retiring = o["Retiring"] o["Mispredicts_Resteers"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Mispredicts_Resteers"].Bad_Speculation = o["Bad_Speculation"] o["Mispredicts_Resteers"].Frontend_Bound = o["Frontend_Bound"] o["Mispredicts_Resteers"].Backend_Bound = o["Backend_Bound"] o["Clears_Resteers"].Retiring = o["Retiring"] o["Clears_Resteers"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Clears_Resteers"].Bad_Speculation = o["Bad_Speculation"] o["Clears_Resteers"].Frontend_Bound = o["Frontend_Bound"] o["Clears_Resteers"].Backend_Bound = o["Backend_Bound"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Bad_Speculation"].Retiring = o["Retiring"] o["Bad_Speculation"].Frontend_Bound = o["Frontend_Bound"] o["Bad_Speculation"].Backend_Bound = o["Backend_Bound"] o["Other_Mispredicts"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Retiring = o["Retiring"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Frontend_Bound = o["Frontend_Bound"] o["Machine_Clears"].Backend_Bound = o["Backend_Bound"] o["Other_Nukes"].Machine_Clears = o["Machine_Clears"] o["Other_Nukes"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Nukes"].Retiring = o["Retiring"] o["Other_Nukes"].Backend_Bound = o["Backend_Bound"] o["Other_Nukes"].Bad_Speculation = o["Bad_Speculation"] o["Other_Nukes"].Frontend_Bound = o["Frontend_Bound"] o["Load_STLB_Hit"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Load_STLB_Hit"].DTLB_Load = o["DTLB_Load"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Store_STLB_Hit"].DTLB_Store = o["DTLB_Store"] o["Store_STLB_Hit"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Serializing_Operation"].C02_WAIT = o["C02_WAIT"] o["Ports_Utilization"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Ports_Utilization"].Retiring = o["Retiring"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["FP_Arith"].Retiring = o["Retiring"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["X87_Use"].Retiring = o["Retiring"] o["FP_Scalar"].Retiring = o["Retiring"] o["FP_Vector"].Retiring = o["Retiring"] o["FP_Vector_128b"].Retiring = o["Retiring"] o["FP_Vector_256b"].Retiring = o["Retiring"] o["FP_Vector_512b"].Retiring = o["Retiring"] o["Int_Operations"].Retiring = o["Retiring"] o["Int_Operations"].Int_Vector_256b = o["Int_Vector_256b"] o["Int_Operations"].Int_Vector_128b = o["Int_Vector_128b"] o["Int_Vector_128b"].Retiring = o["Retiring"] o["Int_Vector_256b"].Retiring = o["Retiring"] o["Memory_Operations"].Retiring = o["Retiring"] o["Memory_Operations"].Light_Operations = o["Light_Operations"] o["Memory_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Fused_Instructions"].Retiring = o["Retiring"] o["Fused_Instructions"].Light_Operations = o["Light_Operations"] o["Fused_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Non_Fused_Branches"].Retiring = o["Retiring"] o["Non_Fused_Branches"].Light_Operations = o["Light_Operations"] o["Non_Fused_Branches"].Heavy_Operations = o["Heavy_Operations"] o["Other_Light_Ops"].Light_Operations = o["Light_Operations"] o["Other_Light_Ops"].Retiring = o["Retiring"] o["Other_Light_Ops"].Heavy_Operations = o["Heavy_Operations"] o["Other_Light_Ops"].Int_Operations = o["Int_Operations"] o["Other_Light_Ops"].Non_Fused_Branches = o["Non_Fused_Branches"] o["Other_Light_Ops"].FP_Arith = o["FP_Arith"] o["Other_Light_Ops"].Fused_Instructions = o["Fused_Instructions"] o["Other_Light_Ops"].Int_Vector_128b = o["Int_Vector_128b"] o["Other_Light_Ops"].FP_Vector = o["FP_Vector"] o["Other_Light_Ops"].FP_Scalar = o["FP_Scalar"] o["Other_Light_Ops"].X87_Use = o["X87_Use"] o["Other_Light_Ops"].Int_Vector_256b = o["Int_Vector_256b"] o["Other_Light_Ops"].Memory_Operations = o["Memory_Operations"] o["Nop_Instructions"].Retiring = o["Retiring"] o["Nop_Instructions"].Light_Operations = o["Light_Operations"] o["Nop_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Shuffles_256b"].Retiring = o["Retiring"] o["Shuffles_256b"].Light_Operations = o["Light_Operations"] o["Shuffles_256b"].Heavy_Operations = o["Heavy_Operations"] o["Few_Uops_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Few_Uops_Instructions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] o["Mispredictions"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Mispredictions"].LCP = o["LCP"] o["Mispredictions"].Retiring = o["Retiring"] o["Mispredictions"].Other_Mispredicts = o["Other_Mispredicts"] o["Mispredictions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Mispredictions"].Frontend_Bound = o["Frontend_Bound"] o["Mispredictions"].DSB_Switches = o["DSB_Switches"] o["Mispredictions"].Backend_Bound = o["Backend_Bound"] o["Mispredictions"].Branch_Resteers = o["Branch_Resteers"] o["Mispredictions"].ICache_Misses = o["ICache_Misses"] o["Mispredictions"].MS_Switches = o["MS_Switches"] o["Mispredictions"].Bad_Speculation = o["Bad_Speculation"] o["Mispredictions"].ITLB_Misses = o["ITLB_Misses"] o["Mispredictions"].Unknown_Branches = o["Unknown_Branches"] o["Mispredictions"].Fetch_Latency = o["Fetch_Latency"] o["Mispredictions"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Big_Code"].LCP = o["LCP"] o["Big_Code"].ICache_Misses = o["ICache_Misses"] o["Big_Code"].DSB_Switches = o["DSB_Switches"] o["Big_Code"].Branch_Resteers = o["Branch_Resteers"] o["Big_Code"].MS_Switches = o["MS_Switches"] o["Big_Code"].ITLB_Misses = o["ITLB_Misses"] o["Big_Code"].Unknown_Branches = o["Unknown_Branches"] o["Big_Code"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].Retiring = o["Retiring"] o["Instruction_Fetch_BW"].Other_Mispredicts = o["Other_Mispredicts"] o["Instruction_Fetch_BW"].DSB_Switches = o["DSB_Switches"] o["Instruction_Fetch_BW"].Backend_Bound = o["Backend_Bound"] o["Instruction_Fetch_BW"].Branch_Resteers = o["Branch_Resteers"] o["Instruction_Fetch_BW"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].ICache_Misses = o["ICache_Misses"] o["Instruction_Fetch_BW"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Instruction_Fetch_BW"].Frontend_Bound = o["Frontend_Bound"] o["Instruction_Fetch_BW"].Bad_Speculation = o["Bad_Speculation"] o["Instruction_Fetch_BW"].ITLB_Misses = o["ITLB_Misses"] o["Instruction_Fetch_BW"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Instruction_Fetch_BW"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Instruction_Fetch_BW"].LCP = o["LCP"] o["Instruction_Fetch_BW"].Clears_Resteers = o["Clears_Resteers"] o["Instruction_Fetch_BW"].MS_Switches = o["MS_Switches"] o["Instruction_Fetch_BW"].Unknown_Branches = o["Unknown_Branches"] o["Cache_Memory_Bandwidth"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Bandwidth"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Bandwidth"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Bandwidth"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Bandwidth"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Bandwidth"].PMM_Bound = o["PMM_Bound"] o["Cache_Memory_Bandwidth"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Bandwidth"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Bandwidth"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Bandwidth"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Bandwidth"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Bandwidth"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Bandwidth"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Bandwidth"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Bandwidth"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Bandwidth"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Bandwidth"].FB_Full = o["FB_Full"] o["Cache_Memory_Bandwidth"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Bandwidth"].DRAM_Bound = o["DRAM_Bound"] o["Cache_Memory_Latency"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Latency"].PMM_Bound = o["PMM_Bound"] o["Cache_Memory_Latency"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Latency"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Latency"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Latency"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Latency"].Store_Latency = o["Store_Latency"] o["Cache_Memory_Latency"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Latency"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Latency"].False_Sharing = o["False_Sharing"] o["Cache_Memory_Latency"].Streaming_Stores = o["Streaming_Stores"] o["Cache_Memory_Latency"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Latency"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Latency"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Latency"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Latency"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Latency"].FB_Full = o["FB_Full"] o["Cache_Memory_Latency"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Latency"].DTLB_Store = o["DTLB_Store"] o["Cache_Memory_Latency"].Split_Stores = o["Split_Stores"] o["Cache_Memory_Latency"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Latency"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Latency"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Data_TLBs"].L1_Bound = o["L1_Bound"] o["Memory_Data_TLBs"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Memory_Data_TLBs"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Memory_Data_TLBs"].DTLB_Load = o["DTLB_Load"] o["Memory_Data_TLBs"].Store_Latency = o["Store_Latency"] o["Memory_Data_TLBs"].Split_Stores = o["Split_Stores"] o["Memory_Data_TLBs"].PMM_Bound = o["PMM_Bound"] o["Memory_Data_TLBs"].DTLB_Store = o["DTLB_Store"] o["Memory_Data_TLBs"].L2_Bound = o["L2_Bound"] o["Memory_Data_TLBs"].Memory_Bound = o["Memory_Bound"] o["Memory_Data_TLBs"].Lock_Latency = o["Lock_Latency"] o["Memory_Data_TLBs"].Store_Bound = o["Store_Bound"] o["Memory_Data_TLBs"].False_Sharing = o["False_Sharing"] o["Memory_Data_TLBs"].Split_Loads = o["Split_Loads"] o["Memory_Data_TLBs"].L3_Bound = o["L3_Bound"] o["Memory_Data_TLBs"].FB_Full = o["FB_Full"] o["Memory_Data_TLBs"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Data_TLBs"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Synchronization"].L1_Bound = o["L1_Bound"] o["Memory_Synchronization"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Synchronization"].False_Sharing = o["False_Sharing"] o["Memory_Synchronization"].Retiring = o["Retiring"] o["Memory_Synchronization"].PMM_Bound = o["PMM_Bound"] o["Memory_Synchronization"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Synchronization"].Machine_Clears = o["Machine_Clears"] o["Memory_Synchronization"].Data_Sharing = o["Data_Sharing"] o["Memory_Synchronization"].Memory_Bound = o["Memory_Bound"] o["Memory_Synchronization"].SQ_Full = o["SQ_Full"] o["Memory_Synchronization"].Store_Bound = o["Store_Bound"] o["Memory_Synchronization"].L3_Bound = o["L3_Bound"] o["Memory_Synchronization"].L2_Bound = o["L2_Bound"] o["Memory_Synchronization"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Synchronization"].Contested_Accesses = o["Contested_Accesses"] o["Memory_Synchronization"].DTLB_Store = o["DTLB_Store"] o["Memory_Synchronization"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Memory_Synchronization"].Store_Latency = o["Store_Latency"] o["Memory_Synchronization"].Split_Stores = o["Split_Stores"] o["Memory_Synchronization"].Backend_Bound = o["Backend_Bound"] o["Memory_Synchronization"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Memory_Synchronization"].Other_Nukes = o["Other_Nukes"] o["Memory_Synchronization"].DRAM_Bound = o["DRAM_Bound"] o["Compute_Bound_Est"].Serializing_Operation = o["Serializing_Operation"] o["Compute_Bound_Est"].Ports_Utilization = o["Ports_Utilization"] o["Compute_Bound_Est"].C02_WAIT = o["C02_WAIT"] o["Compute_Bound_Est"].Retiring = o["Retiring"] o["Compute_Bound_Est"].AMX_Busy = o["AMX_Busy"] o["Compute_Bound_Est"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Compute_Bound_Est"].Memory_Bound = o["Memory_Bound"] o["Compute_Bound_Est"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Compute_Bound_Est"].Core_Bound = o["Core_Bound"] o["Compute_Bound_Est"].Backend_Bound = o["Backend_Bound"] o["Compute_Bound_Est"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Compute_Bound_Est"].Divider = o["Divider"] o["Compute_Bound_Est"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Heavy_Operations = o["Heavy_Operations"] o["Irregular_Overhead"].Ports_Utilization = o["Ports_Utilization"] o["Irregular_Overhead"].C02_WAIT = o["C02_WAIT"] o["Irregular_Overhead"].Retiring = o["Retiring"] o["Irregular_Overhead"].ICache_Misses = o["ICache_Misses"] o["Irregular_Overhead"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Irregular_Overhead"].Frontend_Bound = o["Frontend_Bound"] o["Irregular_Overhead"].Serializing_Operation = o["Serializing_Operation"] o["Irregular_Overhead"].Core_Bound = o["Core_Bound"] o["Irregular_Overhead"].Bad_Speculation = o["Bad_Speculation"] o["Irregular_Overhead"].ITLB_Misses = o["ITLB_Misses"] o["Irregular_Overhead"].Divider = o["Divider"] o["Irregular_Overhead"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Irregular_Overhead"].Memory_Bound = o["Memory_Bound"] o["Irregular_Overhead"].Machine_Clears = o["Machine_Clears"] o["Irregular_Overhead"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Irregular_Overhead"].LCP = o["LCP"] o["Irregular_Overhead"].Other_Mispredicts = o["Other_Mispredicts"] o["Irregular_Overhead"].AMX_Busy = o["AMX_Busy"] o["Irregular_Overhead"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Irregular_Overhead"].DSB_Switches = o["DSB_Switches"] o["Irregular_Overhead"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Assists = o["Assists"] o["Irregular_Overhead"].Backend_Bound = o["Backend_Bound"] o["Irregular_Overhead"].Branch_Resteers = o["Branch_Resteers"] o["Irregular_Overhead"].Clears_Resteers = o["Clears_Resteers"] o["Irregular_Overhead"].MS_Switches = o["MS_Switches"] o["Irregular_Overhead"].Other_Nukes = o["Other_Nukes"] o["Irregular_Overhead"].Unknown_Branches = o["Unknown_Branches"] o["Irregular_Overhead"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].L1_Bound = o["L1_Bound"] o["Other_Bottlenecks"].C02_WAIT = o["C02_WAIT"] o["Other_Bottlenecks"].Retiring = o["Retiring"] o["Other_Bottlenecks"].PMM_Bound = o["PMM_Bound"] o["Other_Bottlenecks"].Data_Sharing = o["Data_Sharing"] o["Other_Bottlenecks"].L2_Bound = o["L2_Bound"] o["Other_Bottlenecks"].Core_Bound = o["Core_Bound"] o["Other_Bottlenecks"].Ports_Utilization = o["Ports_Utilization"] o["Other_Bottlenecks"].Contested_Accesses = o["Contested_Accesses"] o["Other_Bottlenecks"].Divider = o["Divider"] o["Other_Bottlenecks"].L3_Bound = o["L3_Bound"] o["Other_Bottlenecks"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Other_Bottlenecks"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Other_Bottlenecks"].FB_Full = o["FB_Full"] o["Other_Bottlenecks"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Other_Bottlenecks"].Store_Latency = o["Store_Latency"] o["Other_Bottlenecks"].Other_Mispredicts = o["Other_Mispredicts"] o["Other_Bottlenecks"].DSB_Switches = o["DSB_Switches"] o["Other_Bottlenecks"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Other_Bottlenecks"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Other_Bottlenecks"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Other_Bottlenecks"].Assists = o["Assists"] o["Other_Bottlenecks"].Backend_Bound = o["Backend_Bound"] o["Other_Bottlenecks"].Branch_Resteers = o["Branch_Resteers"] o["Other_Bottlenecks"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Other_Bottlenecks"].Heavy_Operations = o["Heavy_Operations"] o["Other_Bottlenecks"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].DTLB_Load = o["DTLB_Load"] o["Other_Bottlenecks"].False_Sharing = o["False_Sharing"] o["Other_Bottlenecks"].ICache_Misses = o["ICache_Misses"] o["Other_Bottlenecks"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Other_Bottlenecks"].Frontend_Bound = o["Frontend_Bound"] o["Other_Bottlenecks"].Machine_Clears = o["Machine_Clears"] o["Other_Bottlenecks"].Streaming_Stores = o["Streaming_Stores"] o["Other_Bottlenecks"].Memory_Bound = o["Memory_Bound"] o["Other_Bottlenecks"].SQ_Full = o["SQ_Full"] o["Other_Bottlenecks"].Store_Bound = o["Store_Bound"] o["Other_Bottlenecks"].Split_Loads = o["Split_Loads"] o["Other_Bottlenecks"].Bad_Speculation = o["Bad_Speculation"] o["Other_Bottlenecks"].ITLB_Misses = o["ITLB_Misses"] o["Other_Bottlenecks"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Other_Bottlenecks"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Other_Bottlenecks"].Serializing_Operation = o["Serializing_Operation"] o["Other_Bottlenecks"].DTLB_Store = o["DTLB_Store"] o["Other_Bottlenecks"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Bottlenecks"].LCP = o["LCP"] o["Other_Bottlenecks"].Split_Stores = o["Split_Stores"] o["Other_Bottlenecks"].AMX_Busy = o["AMX_Busy"] o["Other_Bottlenecks"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Other_Bottlenecks"].Lock_Latency = o["Lock_Latency"] o["Other_Bottlenecks"].MEM_Latency = o["MEM_Latency"] o["Other_Bottlenecks"].Clears_Resteers = o["Clears_Resteers"] o["Other_Bottlenecks"].MS_Switches = o["MS_Switches"] o["Other_Bottlenecks"].Other_Nukes = o["Other_Nukes"] o["Other_Bottlenecks"].Unknown_Branches = o["Unknown_Branches"] o["Other_Bottlenecks"].DRAM_Bound = o["DRAM_Bound"] o["Useful_Work"].Retiring = o["Retiring"] o["Useful_Work"].Heavy_Operations = o["Heavy_Operations"] o["Useful_Work"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Useful_Work"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Useful_Work"].Assists = o["Assists"] o["Core_Bound_Likely"].Memory_Bound = o["Memory_Bound"] o["Core_Bound_Likely"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Core_Bound_Likely"].Core_Bound = o["Core_Bound"] o["Core_Bound_Likely"].Ports_Utilization = o["Ports_Utilization"] o["Core_Bound_Likely"].Retiring = o["Retiring"] o["Core_Bound_Likely"].Backend_Bound = o["Backend_Bound"] o["UopPI"].Retiring = o["Retiring"] o["UpTB"].Retiring = o["Retiring"] o["Retire"].Retiring = o["Retiring"] o["DSB_Misses"].MITE = o["MITE"] o["DSB_Misses"].LCP = o["LCP"] o["DSB_Misses"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Misses"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Misses"].DSB_Switches = o["DSB_Switches"] o["DSB_Misses"].Branch_Resteers = o["Branch_Resteers"] o["DSB_Misses"].ICache_Misses = o["ICache_Misses"] o["DSB_Misses"].MS_Switches = o["MS_Switches"] o["DSB_Misses"].ITLB_Misses = o["ITLB_Misses"] o["DSB_Misses"].DSB = o["DSB"] o["DSB_Misses"].Unknown_Branches = o["Unknown_Branches"] o["DSB_Misses"].Fetch_Latency = o["Fetch_Latency"] o["DSB_Bandwidth"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Bandwidth"].DSB = o["DSB"] o["DSB_Bandwidth"].MITE = o["MITE"] o["DSB_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].LCP = o["LCP"] o["IC_Misses"].MS_Switches = o["MS_Switches"] o["IC_Misses"].ICache_Misses = o["ICache_Misses"] o["IC_Misses"].ITLB_Misses = o["ITLB_Misses"] o["IC_Misses"].Unknown_Branches = o["Unknown_Branches"] o["IC_Misses"].DSB_Switches = o["DSB_Switches"] o["IC_Misses"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].Retiring = o["Retiring"] o["Branch_Misprediction_Cost"].ICache_Misses = o["ICache_Misses"] o["Branch_Misprediction_Cost"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Branch_Misprediction_Cost"].Frontend_Bound = o["Frontend_Bound"] o["Branch_Misprediction_Cost"].Bad_Speculation = o["Bad_Speculation"] o["Branch_Misprediction_Cost"].ITLB_Misses = o["ITLB_Misses"] o["Branch_Misprediction_Cost"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Branch_Misprediction_Cost"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Branch_Misprediction_Cost"].LCP = o["LCP"] o["Branch_Misprediction_Cost"].Other_Mispredicts = o["Other_Mispredicts"] o["Branch_Misprediction_Cost"].DSB_Switches = o["DSB_Switches"] o["Branch_Misprediction_Cost"].Backend_Bound = o["Backend_Bound"] o["Branch_Misprediction_Cost"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].MS_Switches = o["MS_Switches"] o["Branch_Misprediction_Cost"].Unknown_Branches = o["Unknown_Branches"] o["Branch_Misprediction_Cost"].Fetch_Latency = o["Fetch_Latency"] # siblings cross-tree o["Mispredicts_Resteers"].sibling = (o["Branch_Mispredicts"],) o["Clears_Resteers"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["MS_Switches"].sibling = (o["Clears_Resteers"], o["Machine_Clears"], o["L1_Bound"], o["Serializing_Operation"], o["Mixing_Vectors"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Decoder0_Alone"].sibling = (o["Few_Uops_Instructions"],) o["Branch_Mispredicts"].sibling = (o["Mispredicts_Resteers"],) o["Machine_Clears"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"], o["Streaming_Stores"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["Remote_Cache"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Remote_Cache"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Remote_Cache"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"], o["Remote_Cache"],) o["Streaming_Stores"].sibling = (o["FB_Full"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Serializing_Operation"].sibling = (o["MS_Switches"],) o["Mixing_Vectors"].sibling = (o["MS_Switches"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_6"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_512b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector_512b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Int_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_256b"],) o["Int_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["FP_Vector_512b"], o["Int_Vector_128b"],) o["Few_Uops_Instructions"].sibling = (o["Decoder0_Alone"],) o["Microcode_Sequencer"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["Mispredictions"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["Cache_Memory_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],) o["Cache_Memory_Latency"].sibling = (o["L3_Hit_Latency"], o["MEM_Latency"],) o["Memory_Data_TLBs"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Memory_Synchronization"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Irregular_Overhead"].sibling = (o["MS_Switches"], o["Microcode_Sequencer"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Misses"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["Branch_Misprediction_Cost"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
274,697
Python
.py
6,251
37.977764
1,910
0.656158
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,886
tl_cpu.py
andikleen_pmu-tools/tl_cpu.py
# Copyright (c) 2012-2020, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. # # toplev CPU detection # # Environment variables to override (mainly for testing): # FORCECPU=cpu Force CPU type (e.g. skl). Also --force-cpu # FORCECOUNTERS=n Force number of generic counters # FORCEHT=0/1 Force SMT mode # HYPERVISOR=0/1 Force hypervisor mode (also --force-hypervisor) # CPUINFO=file Read cpu information from file instead of /proc/cpuinfo. Also --force-cpuinfo # REDUCED_COUNTERS=n Reduce counters by n # from collections import defaultdict, Counter import os import re import glob import sys if sys.version_info.major == 3: import typing # noqa from typing import Set, List, Dict, Tuple # noqa modelid_map = { (0x8e, ): "KBLR", (0x9e, ): "CFL", } cpus_8gpc = set(["icl", "tgl", "icx", "spr", "sprmax"]) # XXX handle hybrid ADL def num_offline_cpus(): cpus = glob.glob("/sys/devices/system/cpu/cpu[0-9]*/online") offline = 0 for fn in cpus: with open(fn, "r") as f: if int(f.read()) == 0: offline += 1 return offline def reduced_counters(): rc = os.getenv("REDUCED_COUNTERS") if rc: return int(rc) val = 1 fn = "/sys/devices/cpu/allow_tsx_force_abort" if os.path.exists(fn): with open(fn, "r") as f: val = int(f.read()) return 1 if val == 0 else 0 class Env(object): def __init__(self): self.forcecpu = os.getenv("FORCECPU") self.forcecounters = os.getenv("FORCECOUNTERS") self.forceht = os.getenv("FORCEHT") self.hypervisor = os.getenv("HYPERVISOR") self.cpuinfo = os.getenv("CPUINFO") class CPU(object): """Detect the CPU.""" # overrides for easy regression tests def force_cpu(self, known_cpus): force = self.env.forcecpu if not force: return False self.cpu = "" for i in known_cpus: if force == i[0]: self.cpu = i[0] break if self.cpu is None: print("Unknown FORCECPU ",force) return True def force_ht(self): ht = self.env.forceht if ht: self.ht = True if int(ht) else False return True return False def __init__(self, known_cpus=(), nocheck=False, env=Env()): self.vendor = "" self.env = env self.model = 0 self.cpu = "" self.realcpu = "simple" self.ht = False self.counters = {} # type: Dict[str,int] self.has_tsx = False self.hypervisor = False self.force_hypervisor = False if self.env.hypervisor: self.hypervisor = True self.force_hypervisor = True self.freq = 0.0 self.threads = 0 forced_cpu = self.force_cpu(known_cpus) forced_ht = self.force_ht() cores = Counter() # type: typing.Counter[Tuple[int,int]] sockets = Counter() # type: typing.Counter[int] self.coreids = defaultdict(list) self.cputocore = {} self.cputothread = {} self.sockettocpus = defaultdict(list) self.cputosocket = {} self.allcpus = [] self.step = 0 self.name = "" cpuinfo = self.env.cpuinfo if cpuinfo is None: cpuinfo = "/proc/cpuinfo" with open(cpuinfo, "r") as f: seen = set() for l in f: n = l.split() if len(n) < 3: continue if n[0] == 'processor': seen.add("processor") cpunum = int(n[2]) self.allcpus.append(cpunum) elif (n[0], n[2]) == ("vendor_id", "GenuineIntel"): self.vendor = n[2] seen.add("vendor_id") elif (len(n) > 3 and (n[0], n[1], n[3]) == ("cpu", "family", "6")): seen.add("cpu family") elif (n[0], n[1]) == ("model", ":"): seen.add("model") self.model = int(n[2]) elif (n[0], n[1]) == ("model", "name"): seen.add("model name") m = re.search(r"@ (\d+\.\d+)GHz", l) if m: self.freq = float(m.group(1)) self.name = " ".join(n[3:]) elif (n[0], n[1]) == ("physical", "id"): physid = int(n[3]) sockets[physid] += 1 self.sockettocpus[physid].append(cpunum) self.cputosocket[cpunum] = physid elif (n[0], n[1]) == ("core", "id"): coreid = int(n[3]) key = (physid, coreid,) cores[key] += 1 self.threads = max(self.threads, cores[key]) if self.threads > 1 and not forced_ht: self.ht = True self.coreids[key].append(cpunum) self.cputocore[cpunum] = key self.cputothread[cpunum] = self.coreids[key].index(cpunum) elif n[0] == "flags": seen.add("flags") self.has_tsx = "rtm" in n if "hypervisor" in n: self.hypervisor = True elif n[0] == "stepping": seen.add("stepping") self.step = int(n[2]) if len(seen) >= 7: for i in known_cpus: if self.model in i[1] or (self.model, self.step) in i[1]: self.realcpu = i[0] if not forced_cpu: self.cpu = i[0] break self.force_counters() self.limit4_counters = { "cpu": "none" } self.standard_counters = { "cpu": tuple(("0,1,2,3",)) } if self.cpu.startswith("adl") or self.cpu.startswith("mtl"): atom_counters = 6 if self.cpu.startswith("adl") else 8 newcounters = { "cpu_core": 8, "cpu": 4, "cpu_atom": atom_counters, } self.standard_counters = { "cpu_core": ("0,1,2,3,4,5,6,7", "0,1,2,3", ), "cpu": ("0,1,2,3,4,5,6,7", "0,1,2,3", ), "cpu_atom": ("0,1,2,3,4,5", ) if atom_counters == 6 else ("0,1,2,3,4,5,6,7,", ) } self.limit4_counters = { "cpu_core": "0,1,2,3", "cpu_atom": "none", "cpu": "0,1,2,3" } elif self.cpu == "slm": newcounters = { "cpu": 2 } self.standard_counters = { "cpu": ("0,1",) } # when running in a hypervisor always assume worst case HT in on # also when CPUs are offline assume SMT is on elif self.ht or self.hypervisor or (num_offline_cpus() > 0 and not nocheck) or self.cpu in cpus_8gpc: if self.cpu in cpus_8gpc or (self.cpu == "simple" and self.realcpu in cpus_8gpc): newcounters = {"cpu": 8 } self.standard_counters = { "cpu": ("0,1,2,3,4,5,6,7", "0,1,2,3", ) } self.limit4_counters = { "cpu": "0,1,2,3" } else: newcounters = { "cpu": 4 } elif self.cpu == "ehl": newcounters = { "cpu": 4 } else: newcounters = { "cpu": 8 } if not self.counters: self.counters = newcounters if not nocheck and not self.env.forcecounters: for j in ("cpu", "cpu_core", "cpu_atom"): if j in self.counters: self.counters[j] -= reduced_counters() if self.cpu in cpus_8gpc: self.standard_counters = { "cpu": ("0,1,2,3,4,5,6,7", "0,1,2,3", ) } self.limit4_counters = { "cpu": "0,1,2,3" } if self.cpu == "spr" and "Max" in self.name: self.cpu = "sprmax" try: self.pmu_name = open("/sys/devices/cpu/caps/pmu_name").read().strip() except IOError: self.pmu_name = "" self.sockets = len(sockets.keys()) self.modelid = None mid = (self.model,) self.true_name = self.cpu if mid in modelid_map: self.modelid = modelid_map[mid] self.true_name = self.modelid.lower() # XXX match steppings here too def force_counters(self): cnt = self.env.forcecounters if cnt: cntn = int(cnt) if self.realcpu == "adl" or self.realcpu == "mtl": self.counters = { "cpu_core": cntn, "cpu_atom": cntn, "cpu": cntn } else: self.counters = { "cpu": cntn }
9,244
Python
.py
234
28.047009
109
0.506336
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,887
gen_level.py
andikleen_pmu-tools/gen_level.py
# generate levels for events from the model # utility module for other tools l1 = set(("Frontend_Bound", "Backend_Bound", "Retiring", "Bad_Speculation")) def get_level(name): is_node = name in l1 or "." in name level = name.count(".") + 1 if is_node: return level return 0 def is_metric(name): return get_level(name) == 0 def level_name(name): if name.count(".") > 0: f = name.split(".")[:-1] n = ".".join(f) elif is_metric(name): return "CPU-METRIC" # XXX split else: n = "TopLevel" n = n.replace(" ", "_") return n
600
Python
.py
21
23.619048
76
0.586806
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,888
dummyarith.py
andikleen_pmu-tools/dummyarith.py
# dummy arithmetic type without any errors, for collecting # the events from the model. Otherwise divisions by zero cause # early exits class DummyArith(object): def __add__(self, o): return self __sub__ = __add__ __mul__ = __add__ __div__ = __add__ __truediv__ = __add__ __rsub__ = __add__ __radd__ = __add__ __rmul__ = __add__ __rdiv__ = __add__ __rtruediv__ = __add__ def __lt__(self, o): return True __eq__ = __lt__ __ne__ = __lt__ __gt__ = __lt__ __ge__ = __lt__ __or__ = __add__ __and__ = __add__ __min__ = __add__ __max__ = __add__
633
Python
.py
25
20.56
62
0.455446
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,889
toplev.py
andikleen_pmu-tools/toplev.py
#!/usr/bin/env python3 # Copyright (c) 2012-2022, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. # # Measure a workload using the topdown performance model: # estimate on which part of the CPU pipeline it bottlenecks. # # Must find ocperf in python module path. # Handles a variety of perf and kernel versions, but older ones have various # limitations. # Environment variables for overrides (see also tl_cpu/ocperf): # TOPOLOGY=file Read sysfs topology from file. Also --force-topology # PERF=exe Force perf binary to run # FORCE_NMI_WATCHDOG=1 Force NMI watchdog mode # KERNEL_VERSION=... Force kernel version (e.g. 5.0) # FORCEMETRICS=1 Force fixed metrics and slots # TLSEED=n Set seed for --subset sample: sampling # DURATION_TIME=0 Force not using duration_time from __future__ import print_function, division import sys import os import re import textwrap import platform import pty import subprocess import argparse import time import types import csv import bisect import random import json import io import glob from dummyarith import DummyArith from copy import copy from fnmatch import fnmatch from math import isnan from collections import defaultdict, Counter, OrderedDict from itertools import compress, groupby, chain from listutils import cat_unique, dedup, filternot, not_list, append_dict, \ zip_longest, flatten, findprefix, dummy_dict from objutils import has, safe_ref, map_fields, ref_or from tl_stat import ComputeStat, ValStat, combine_valstat import tl_cpu import tl_output import ocperf import event_download from tl_uval import UVal from tl_io import flex_open_r, flex_open_w, popentext, warn, warn_once, \ warn_once_no_assert, print_once, \ obj_debug_print, debug_print, warn_no_assert, \ set_args as io_set_args if sys.version_info.major == 3: import typing # noqa from typing import Set, List, Dict, Any, Tuple, DefaultDict # noqa known_cpus = ( ("snb", (42, )), ("jkt", (45, )), ("ivb", (58, )), ("ivt", (62, )), ("hsw", (60, 70, 69 )), ("hsx", (63, )), ("slm", (55, 77, 76, )), ("bdw", (61, 71, )), ("bdx", (79, 86, )), ("simple", ()), ("skl", (94, 78, 142, 158, 165, 166, )), ("knl", (87, )), ("skx", ((85, 4,), )), ("clx", ((85, 5,), (85, 6,), (85, 7,), (85, 8,), (85, 9,), (85, 10,), )), ("icl", (126, 125, 157, 167, )), # RKL as ICL for now ("tgl", (140, 141, )), ("icx", (106, 108, )), ("adl", (154, 151, 183, )), # RPL as ADL ("adl-glc", (154, )), ("adl-grt", (154, )), ("spr", (143, )), ("ehl", (150, )), ("sprmax", ()), ("mtl", (170, 186, )), ("mtl-cmt", (170, 186, )), ("mtl-rwc", (170, 186, )), ) eventlist_alias = { } # type: Dict[str,str] tsx_cpus = ("hsw", "hsx", "bdw", "skl", "skx", "clx", "icl", "icx", "spr", "sprmax") hybrid_cpus = ("adl", "mtl",) atom_hybrid_cpus = ("adl-grt", "mtl-cmt",) non_json_events = set(("dummy", "duration_time")) tma_mgroups = set() # type: Set[str] # tunables (tunable with --tune) DEDUP_AREA = "Info.Bot*" DEDUP_NODE = "" BOTTLENECK_LEVEL_INC = 1 IDLE_MARKER_THRESHOLD = 0.05 SIB_THRESH = 0.05 PERF_SKIP_WINDOW = 15 KEEP_UNREF = False INAME = False FUZZYINPUT = False # handle kernels that don't support all events unsup_pebs = ( ("BR_MISP_RETIRED.ALL_BRANCHES:pp", (("hsw",), (3, 18), None)), ("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM:pp", (("hsw",), (3, 18), None)), ("MEM_LOAD_UOPS_RETIRED.L3_MISS:pp", (("hsw",), (3, 18), None)), ) ivb_ht_39 = (("ivb", "ivt"), (4, 1), (3, 9)) # uncomment if you removed commit 741a698f420c3 from kernel #ivb_ht_39 = ((), None, None) # both kernel bugs and first time a core was supported # disable events if the kernel does not support them properly # this does not handle backports (override with --force-events) unsup_events = ( # commit 36bbb2f2988a29 ("OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.HITM_OTHER_CORE", (("hsw", "hsx"), (3, 18), None)), # commit 741a698f420c3 broke it, commit e979121b1b and later fixed it ("MEM_LOAD_UOPS_L*_HIT_RETIRED.*", ivb_ht_39), ("MEM_LOAD_UOPS_RETIRED.*", ivb_ht_39), ("MEM_LOAD_UOPS_L*_MISS_RETIRED.*", ivb_ht_39), ("MEM_UOPS_RETIRED.*", ivb_ht_39), # commit 5e176213a6b2bc # the event works, but it cannot put into the same group as # any other CYCLE_ACTIVITY.* event. For now black list, but # could also special case this in the group scheduler. ("CYCLE_ACTIVITY.STALLS_TOTAL", (("bdw", (4, 4), None))), # commit 91f1b70582c62576 ("CYCLE_ACTIVITY.*", (("bdw"), (4, 1), None)), ("L1D_PEND_MISS.PENDING", (("bdw"), (4, 1), None)), # commit 6113af14c8 ("CYCLE_ACTIVITY:CYCLES_LDM_PENDING", (("ivb", "ivt"), (3, 12), None)), # commit f8378f52596477 ("CYCLE_ACTIVITY.*", (("snb", "jkt"), (3, 9), None)), # commit 0499bd867bd17c (ULT) or commit 3a632cb229bfb18 (other) # technically most haswells are 3.10, but ULT is 3.11 ("L1D_PEND_MISS.PENDING", (("hsw",), (3, 11), None)), ("L1D_PEND_MISS.PENDING", (("hsx"), (3, 10), None)), # commit c420f19b9cdc ("CYCLE_ACTIVITY.*_L1D_PENDING", (("hsw", "hsx"), (4, 1), None)), ("CYCLE_ACTIVITY.CYCLES_NO_EXECUTE", (("hsw", "hsx"), (4, 1), None)), # commit 3a632cb229b ("CYCLE_ACTIVITY.*", (("hsw", "hsx"), (3, 11), None))) FIXED_BASE = 50 METRICS_BASE = 100 SPECIAL_END = 130 limited_counters_base = { "instructions": FIXED_BASE + 0, "cycles": FIXED_BASE + 1, "ref-cycles": FIXED_BASE + 2, "slots": FIXED_BASE + 3, "cpu/slots/": FIXED_BASE + 3, "cpu_core/slots/": FIXED_BASE + 3, "topdown.slots": FIXED_BASE + 3, "cpu_core/topdown-fe-bound/": METRICS_BASE + 0, "cpu/topdown-fe-bound/": METRICS_BASE + 0, "cpu_core/topdown-be-bound/": METRICS_BASE + 1, "cpu/topdown-be-bound/": METRICS_BASE + 1, "cpu_core/topdown-bad-spec/": METRICS_BASE + 2, "cpu/topdown-bad-spec/": METRICS_BASE + 2, "cpu_core/topdown-retiring/": METRICS_BASE + 3, "cpu/topdown-retiring/": METRICS_BASE + 3, "cpu_core/topdown-heavy-ops/": METRICS_BASE + 4, "cpu/topdown-heavy-ops/": METRICS_BASE + 4, "cpu/topdown-br-mispredict/": METRICS_BASE + 5, "cpu_core/topdown-br-mispredict/": METRICS_BASE + 5, "cpu_core/topdown-mem-bound/": METRICS_BASE + 6, "cpu/topdown-mem-bound/": METRICS_BASE + 6, "cpu/topdown-fetch-lat/": METRICS_BASE + 7, "cpu_core/topdown-fetch-lat/": METRICS_BASE + 7, "cpu/cycles-ct/": 2, "cpu_core/cycles-ct/": 2, } promotable_limited = set(( "instructions", "cycles", "slots", "cpu/slots/", "cpu_core/slots/") ) Undef = UVal("undef", 0) # should clean those up, but they're all read-only after initialization anyways global smt_mode global output_numcpus global metrics_own_group global run_l1_parallel class EventContextBase(object): def __init__(self): self.constraint_fixes = {} self.constraint_patterns = [] self.errata_whitelist = [] self.outgroup_events = set() self.sched_ignore_events = set() self.require_pebs_events = set() self.core_domains = set() self.limited_counters = {} self.limited_set = set() self.fixed_events = set() self.errata_events = {} self.errata_warn_events = {} self.limit4_events = set() self.notfound_cache = {} self.rmap_cache = {} self.slots_available = False self.standard_counters = tuple("") # type: Tuple[str, ...] self.counters = 0 self.limit4_counters = "" self.force_metrics = False self.metrics_override = False def init_emap(self, emap): self.emap = emap class EventContext(EventContextBase): """Event related context for a given target CPU.""" def __init__(self, pmu): EventContextBase.__init__(self) self.outgroup_events = set(["dummy", "duration_time", "msr/tsc/"]) self.core_domains = set(["Slots", "CoreClocks", "CoreMetric", "Core_Execution", "Core_Clocks", "Core_Metric"]) self.limited_counters = dict(limited_counters_base) self.limited_set = set(self.limited_counters.keys()) self.fixed_events = set([x for x in self.limited_counters if FIXED_BASE <= self.limited_counters[x] <= SPECIAL_END]) if (pmu is None or pmu not in cpu.counters or pmu not in cpu.standard_counters or pmu not in cpu.limit4_counters): pmu = "cpu" self.standard_counters = cpu.standard_counters[pmu] self.counters = cpu.counters[pmu] self.limit4_counters = cpu.limit4_counters[pmu] smt_mode = False def works(x): return os.system(x + " >/dev/null 2>/dev/null") == 0 exists_cache = {} # type: Dict[str,bool] def cached_exists(fn): if fn in exists_cache: return exists_cache[fn] found = os.path.exists(fn) exists_cache[fn] = found return found def safe_int(x): try: return int(x) except ValueError: return 0 def event_nocheck(): return args.import_ or args.no_check class PerfFeatures(object): """Adapt to the quirks of various perf versions.""" def __init__(self, args): pmu = "cpu" if os.path.exists("/sys/devices/cpu_core"): pmu = "cpu_core" p = os.getenv("PERF") self.perf = p if p else "perf" ret = os.system(self.perf + " stat --log-fd 3 3>/dev/null true") if ret: # work around the insane perf setup on Debian derivates # it fails if the perf isn't the same as the kernel # look for the underlying perf installs, if any # perf is compatible, so just pick the newest if ret == 512: l = sorted(glob.glob("/usr/lib/linux-tools*/perf"), key=lambda x: [int(t) if t.isdigit() else t for t in re.split(r'(\d+)', x)]) if len(l) > 0: self.perf = l[0] ret = os.system(self.perf + " stat --log-fd 3 3>/dev/null true") if ret: sys.exit("perf binary is too old/not installed or perf is disabled in /proc/sys/kernel/perf_event_paranoid") self.logfd_supported = ret == 0 self.supports_power = ( not args.no_uncore and not args.force_hypervisor and os.path.exists("/sys/devices/power/events/energy-cores")) with os.popen(self.perf + " --version") as f: v = f.readline().split() perf_version = tuple(map(safe_int, v[2].split(".")[:2])) if len(v) >= 3 else (0,0) self.supports_percore = (perf_version >= (5,7) or works(self.perf + " stat --percore-show-thread true")) dt = os.getenv("DURATION_TIME") if dt: self.supports_duration_time = int(dt) else: self.supports_duration_time = (perf_version >= (5,2) or works(self.perf + " stat -e duration_time true")) # guests don't support offcore response if event_nocheck(): self.has_max_precise = True self.max_precise = 3 else: self.has_max_precise = os.path.exists("/sys/devices/%s/caps/max_precise" % pmu) if self.has_max_precise: self.max_precise = int(open("/sys/devices/%s/caps/max_precise" % pmu).read()) if args.exclusive and not args.print and not (perf_version >= (5,10) or works(self.perf + " stat -e '{branches,branches,branches,branches}:e' true")): sys.exit("perf binary does not support :e exclusive modifier") def kv_to_key(v): return v[0] * 100 + v[1] def unsup_event(e, table, kernel_version, min_kernel=None): if ":" in e: e = e[:e.find(":")] for j in table: if fnmatch(e, j[0]) and cpu.realcpu in j[1][0]: break else: return False v = j[1] if v[1] and kernel_version < kv_to_key(v[1]): if min_kernel: min_kernel.append(v[1]) return True if v[2] and kernel_version >= kv_to_key(v[2]): return False return False def remove_qual(ev): def get_group(prefix, m): if m.group(1): return prefix + m.group(1) return prefix return re.sub(r':(p?)[ku]+', lambda m: get_group("", m), re.sub(r'/(p?)[ku]+', lambda m: get_group("/", m), ev)) def limited_overflow(evlist, num): class GenericCounters: def __init__(self): self.num = 0 def gen_overflow(c, gc, inc): if c in promotable_limited: gc.num += inc - 1 return False return True assigned = Counter([ectx.limited_counters[remove_qual(x)] for x in evlist if remove_qual(x) in ectx.limited_counters]) gc = GenericCounters() return any([x > 1 and gen_overflow(k, gc, x) for k, x in assigned.items()]), gc.num # we limit to 3 events because one could be taken up by the nmi watchdog # and also there's some kernel issue that sometimes only 3 fit on ICL LIMIT4_MAX_EVENTS = 3 # limited to first four counters on ICL+ def limit4_overflow(evlist): return sum([remove_qual(x) in ectx.limit4_events for x in evlist]) > LIMIT4_MAX_EVENTS def ismetric(x): return x.startswith(("topdown-", "cpu_core/topdown-", "cpu/topdown-")) resources = ("frontend=", "offcore_rsp=", "ldlat=", "in_tx_cp=", "cycles-ct") def event_to_resource(ev): for j in resources: if remove_qual(j) in ev: return j return "" def resource_split(evlist): r = Counter(map(event_to_resource, evlist)) for j in sorted(r.keys()): if j == "": continue if j == "offcore_rsp=": if r[j] > 2: return True elif r[j] > 1: return True return False def num_generic_counters(evset): # XXX does not handle formulas having different u/k qualifiers, but we would need to fix the # callers to be consistent to handle that return len(evset - set(add_filter(ectx.fixed_events)) - ectx.fixed_events - ectx.outgroup_events - ectx.sched_ignore_events) FORCE_SPLIT = 100 # Force metrics into own group metrics_own_group = True def is_slots(x): return re.match(r'(cpu/|cpu_core/)?slots[,/]', x) is not None def needed_counters(evlist): evset = set(evlist) num = num_generic_counters(evset) metrics = [ismetric(x) for x in evlist] slots = [is_slots(x) for x in evlist] if any(metrics) or any(slots): # slots must be first if metrics are present if any(map(is_slots, evlist)) and not is_slots(evlist[0]): debug_print("split for slots %s" % evlist) return FORCE_SPLIT # force split if there are other events. if metrics_own_group and len(evlist) > sum(metrics) + 1: debug_print("split for other events in topdown %s" % evlist) return FORCE_SPLIT # split if any resource is oversubscribed if resource_split(evlist): debug_print("resource split %s" % evlist) return FORCE_SPLIT evlist = list(compress(evlist, not_list(metrics))) # force split if we overflow fixed or limited l_over, numg = limited_overflow(evlist, num) if l_over: debug_print("split for limited overflow %s " % evlist) return FORCE_SPLIT if limit4_overflow(evlist): debug_print("split for limit4 overflow %s" % evlist) return FORCE_SPLIT return num + numg def event_group(evlist): evlist = add_filter(evlist) l = [] # type: List[str] pgroup = False for is_og, g in groupby(evlist, lambda x: x in ectx.outgroup_events): gl = list(g) slots_or_metric = [ismetric(x) or is_slots(x) for x in gl] # keep the groups for slots or metric because of some kernel # requirements and also some perf versions reorder slots with no group. if is_og or (args.no_group and not any(slots_or_metric)): l += gl else: e = ",".join(gl) e = "{%s}" % e if args.exclusive or args.pinned or args.weak or args.host or args.guest: e += ":" if args.weak: e += "W" if args.exclusive: e += "e" if args.guest: e += "G" if args.host: e += "H" elif args.pinned: if all(slots_or_metric): e += "D" assert pgroup is False assert is_slots(gl[0]) pgroup = True else: assert not any(slots_or_metric) l.append(e) return ",".join(l) def exe_dir(): d = os.path.realpath(sys.argv[0]) d = os.path.dirname(d) if d: return d return "." def add_args(rest, *args): a = [x for x in args if x not in rest] return a + rest def update_arg(arg, flag, sep, newval): i = findprefix(arg, flag, "--") if i >= 0: if arg[i] == flag: arg[i+1] = newval else: arg[i] = flag + sep + newval return True return False def del_arg_val(arg, flag): i = findprefix(arg, flag, "--") del arg[i:i+2 if arg[i] == flag else i+1] def init_args(): p = argparse.ArgumentParser(usage='toplev [options] perf-arguments', description=''' Estimate on which part of the CPU pipeline a workload bottlenecks using the TopDown model. The bottlenecks are expressed as a tree with different levels. Requires a modern Intel CPU. Examples: toplev -l1 --single-thread program measure single threaded program. On hyper threaded systems with Skylake or older the system should be idle. toplev -NB program Measure program showing consolidated bottleneck view and extra information associated with bottlenecks. Note this will multiplex performance counters, so there may be measuring errors. toplev -NB --run-sample program Measure programing showing bottlenecks and extra nodes, and automatically sample for the location of bottlenecks in a second pass. toplev --drilldown --only-bottleneck program Rerun workload with minimal multiplexing until critical bottleneck is found. Only print critical bottleneck toplev -l3 --no-desc -I 100 -x, sleep X measure whole system for X seconds every 100ms, outputting in CSV format. toplev --all --core C0 taskset -c 0,1 program Measure program running on core 0 with all nodes and metrics enables toplev --all --xlsx x.xlsx -a sleep 10 Generate spreadsheet with full system measurement for 10 seconds ''', epilog=''' Other perf arguments allowed (see the perf documentation) After -- perf arguments conflicting with toplev can be used. Some caveats: toplev defaults to measuring the full system and show data for all CPUs. Use taskset to limit the workload to known CPUs if needed. In some cases (idle system, single threaded workload) --single-thread can also be used to get less output. The lower levels of the measurement tree are less reliable than the higher levels. They also rely on counter multi-plexing, and can not run each equation in a single group, which can cause larger measurement errors with non steady state workloads (If you don't understand this terminology; it means measurements in higher levels are less accurate and it works best with programs that primarily do the same thing over and over) If the program is very reproducible -- such as a simple kernel -- it is also possible to use --no-multiplex. In this case the workload is rerun multiple times until all data is collected. Only use with sleep if the workload is running in a steady state. With the --drilldown option toplev can automatically remeasure the workload with only the nodes needed to measure the particular bottlenecks This also requires a reproducible or steady-state workload. toplev needs a new enough perf tool and has specific requirements on the kernel. See http://github.com/andikleen/pmu-tools/wiki/toplev-kernel-support.''', formatter_class=argparse.RawDescriptionHelpFormatter) g = p.add_argument_group('General operation') g.add_argument('--interval', '-I', help='Measure every ms instead of only once', type=int) g.add_argument('--no-multiplex', help='Do not multiplex, but run the workload multiple times as needed. ' 'Requires reproducible workloads.', action='store_true') g.add_argument('--single-thread', '-S', help='Measure workload as single thread. Workload must run single threaded. ' 'In SMT mode other thread must be idle.', action='store_true') g.add_argument('--fast', '-F', help='Skip sanity checks to optimize CPU consumption', action='store_true') g.add_argument('--import', help='Import specified perf stat output file instead of running perf. ' 'Must be for same cpu, same arguments, same /proc/cpuinfo, same topology, unless overriden', dest='import_') g.add_argument('--subset', help="Process only a subset of the input file with --import. " "Valid syntax: a-b. Process from seek offset a to b. b is optional. " "x/n%% process x'th n percent slice. Starts counting at 0. Add - to process to end of input. " "sample:n%% Sample each time stamp in input with n%% (0-100%%) probability. " "toplev will automatically round to the next time stamp boundary.") g.add_argument('--parallel', help="Run toplev --import in parallel in N processes, or the system's number of CPUs if 0 is specified", action='store_true') g.add_argument('--pjobs', type=int, default=0, help='Number of threads to run with parallel. Default is number of CPUs.') g.add_argument('--gen-script', help='Generate script to collect perfmon information for --import later', action='store_true') g.add_argument('--script-record', help='Use perf stat record in script for faster recording or ' 'import generated perf.data (requires new perf)', action='store_true') g.add_argument('--drilldown', help='Automatically rerun to get more details on bottleneck', action='store_true') g.add_argument('--show-cpu', help='Print current CPU type and exit', action='store_true') g = p.add_argument_group('Measurement filtering') g.add_argument('--kernel', help='Only measure kernel code', action='store_true') g.add_argument('--user', help='Only measure user code', action='store_true') g.add_argument('--cpu', '-C', help=argparse.SUPPRESS) g.add_argument('--pid', '-p', help=argparse.SUPPRESS) g.add_argument('--core', help='Limit output to cores. Comma list of Sx-Cx-Tx. All parts optional.') g.add_argument('--no-aggr', '-A', help='Measure every CPU', action='store_true') g.add_argument('--cputype', help='Limit to hybrid cpu type (atom or core)', choices=['atom', 'core']) g = p.add_argument_group('Select events') g.add_argument('--level', '-l', help='Measure upto level N (max 6)', type=int, default=-1) g.add_argument('--metrics', '-m', help="Print extra metrics", action='store_true') g.add_argument('--sw', help="Measure perf Linux metrics", action='store_true') g.add_argument('--no-util', help="Do not measure CPU utilization", action='store_true') g.add_argument('--tsx', help="Measure TSX metrics", action='store_true') g.add_argument('--all', help="Measure everything available", action='store_true') g.add_argument('--frequency', help="Measure frequency", action='store_true') g.add_argument('--power', help='Display power metrics', action='store_true') g.add_argument('--nodes', help='Include or exclude nodes (with + to add, -|^ to remove, ' 'comma separated list, wildcards allowed, add * to include all children/siblings, ' 'add /level to specify highest level node to match, ' 'add ^ to match related siblings and metrics, ' 'start with ! to only include specified nodes)') g.add_argument('--metric-group', help='Add (+) or remove (-|^) metric groups of metrics, ' 'comma separated list from --list-metric-groups.', default=None) g.add_argument('--areas', help='Add specific areas. Comma separate list, wildcards allowed') g.add_argument('--pinned', help='Run topdown metrics (on ICL+) pinned', action='store_true') g.add_argument('--exclusive', help='Use exclusive groups. Requires new kernel and new perf', action='store_true') g.add_argument('--host', action='store_true', help="Count host only") g.add_argument('--guest', action='store_true', help="Count guest only") g.add_argument('--weak', action='store_true', help="Use weak groups to work around scheduling problems") g.add_argument('--thread', help="Enable per thread SMT measurements for pre-ICL, at the cost of more multiplexing.", action='store_true') g.add_argument('--aux', help='Enable auxilliary hierarchy nodes on some models. ' 'Auxiliary nodes offer alternate views of the same bottleneck component, which can impact observed bottleneck percentage totals', action='store_true') g.add_argument('--node-metrics', '-N', help='Add metrics related to selected nodes, but hide when node is not crossing threshold', action='store_true') g.add_argument('--bottlenecks', '-B', help='Show bottlenecks view of Bottleneck metrics. Use -l0 to disable standard topdown view.', action='store_true') g = p.add_argument_group('Model tunables') g.add_argument('--fp16', help='Enable FP16 support in some models', action='store_true') g.add_argument('--hbm-only', help='Enable HBM only mode in some models', action='store_true') g.add_argument('--ret-latency', help='Read JSON file with Retire latencies. Can specify path inside JSON file with :, for example foo.json:6-cores:MEAN') g = p.add_argument_group('Query nodes') g.add_argument('--list-metrics', help='List all metrics. Can be followed by prefixes to limit, ^ for full match', action='store_true') g.add_argument('--list-nodes', help='List all nodes. Can be followed by prefixes to limit, ^ for full match', action='store_true') g.add_argument('--list-metric-groups', help='List metric groups.', action='store_true') g.add_argument('--list-all', help='List every supported node/metric/metricgroup. Can be followed by prefixes to limit, ^ for full match.', action='store_true') g.add_argument('--describe', help='Print full descriptions for listed node prefixes. Add ^ to require full match.', action='store_true') g = p.add_argument_group('Workarounds') g.add_argument('--no-group', help='Dont use groups', action='store_true') g.add_argument('--force-events', help='Assume kernel supports all events. May give wrong results.', action='store_true') g.add_argument('--ignore-errata', help='Do not disable events with errata', action='store_true', default=True) g.add_argument('--handle-errata', help='Disable events with errata', action='store_true') g.add_argument('--reserved-counters', default=0, help='Assume N generic counters are used elsewhere', type=int) g = p.add_argument_group('Filtering output') g.add_argument('--only-bottleneck', help='Only print topdown tree bottleneck and associated metrics (unless overriden with other options like --nodes or --bottleneck)', action='store_true') g.add_argument('--verbose', '-v', help='Print all results even when below threshold or exceeding boundaries. ' 'Note this can result in bogus values, as the TopDown methodology relies on thresholds ' 'to correctly characterize workloads. Values not crossing threshold are marked with <.', action='store_true') g = p.add_argument_group('Output format') g.add_argument('--per-core', help='Aggregate output per core', action='store_true') g.add_argument('--per-socket', help='Aggregate output per socket', action='store_true') g.add_argument('--per-thread', help='Aggregate output per CPU thread', action='store_true') g.add_argument('--global', help='Aggregate output for all CPUs', action='store_true', dest='global_') g.add_argument('--no-desc', help='Do not print event descriptions', action='store_true') g.add_argument('--desc', help='Force event descriptions', action='store_true') g.add_argument('--csv', '-x', help='Enable CSV mode with specified delimeter') g.add_argument('--output', '-o', help='Set output file') g.add_argument('--split-output', help='Generate multiple output files, one for each specified ' 'aggregation option (with -o)', action='store_true') g.add_argument('--graph', help='Automatically graph interval output with tl-barplot.py', action='store_true') g.add_argument("--graph-cpu", help="CPU to graph using --graph") g.add_argument('--title', help='Set title of graph') g.add_argument('-q', '--quiet', help='Avoid unnecessary status output', action='store_true') g.add_argument('--long-desc', help='Print long descriptions instead of abbreviated ones.', action='store_true') g.add_argument('--columns', help='Print CPU output in multiple columns for each node', action='store_true') g.add_argument('--json', help='Print output in JSON format for Chrome about://tracing', action='store_true') g.add_argument('--summary', help='Print summary at the end. Only useful with -I', action='store_true') g.add_argument('--no-area', help='Hide area column', action='store_true') g.add_argument('--perf-output', help='Save perf stat output in specified file') g.add_argument('--perf-summary', help='Save summarized perf stat output in specified file') g.add_argument('--no-perf', help=argparse.SUPPRESS, action='store_true') # noop, for compatibility g.add_argument('--perf', help='Print perf command line', action='store_true') g.add_argument('--print', help="Only print perf command line. Don't run", action='store_true') g.add_argument('--idle-threshold', help="Hide idle CPUs (default <5%% of busiest if not CSV, specify percent)", default=None, type=float) g.add_argument('--no-output', help="Don't print computed output. Does not affect --summary.", action='store_true') g.add_argument('--no-mux', help="Don't print mux statistics", action="store_true") g.add_argument('--abbrev', help="Abbreviate node names in output", action="store_true") g.add_argument('--no-sort', help="Don't sort output by Metric group", action="store_true") g = p.add_argument_group('Environment') g.add_argument('--force-cpu', help='Force CPU type', choices=[x[0] for x in known_cpus]) g.add_argument('--force-topology', metavar='findsysoutput', help='Use specified topology file (find /sys/devices)') g.add_argument('--force-cpuinfo', metavar='cpuinfo', help='Use specified cpuinfo file (/proc/cpuinfo)') g.add_argument('--force-hypervisor', help='Assume running under hypervisor (no uncore, no offcore, no PEBS)', action='store_true') g.add_argument('--no-uncore', help='Disable uncore events', action='store_true') g.add_argument('--no-check', help='Do not check that PMU units exist', action='store_true') g = p.add_argument_group('Additional information') g.add_argument('--print-group', '-g', help='Print event group assignments', action='store_true') g.add_argument('--raw', help="Print raw values", action='store_true') g.add_argument('--valcsv', '-V', help='Write raw counter values into CSV file') g.add_argument('--stats', help='Show statistics on what events counted', action='store_true') g = p.add_argument_group('xlsx output') g.add_argument('--xlsx', help='Generate xlsx spreadsheet output with data for ' 'socket/global/thread/core/summary/raw views with 1s interval. ' 'Add --single-thread to only get program output.') g.add_argument('--set-xlsx', help=argparse.SUPPRESS, action='store_true') # set arguments for xlsx only g.add_argument('--xnormalize', help='Add extra sheets with normalized data in xlsx files', action='store_true') g.add_argument('--xchart', help='Chart data in xlsx files', action='store_true') g.add_argument('--keep', help='Keep temporary files', action='store_true') g.add_argument('--xkeep', dest='keep', action='store_true', help=argparse.SUPPRESS) g = p.add_argument_group('Sampling') g.add_argument('--show-sample', help='Show command line to rerun workload with sampling', action='store_true') g.add_argument('--run-sample', help='Automatically rerun workload with sampling', action='store_true') g.add_argument('--sample-args', help='Extra arguments to pass to perf record for sampling. Use + to specify -', default='-g') g.add_argument('--sample-repeat', help='Repeat measurement and sampling N times. This interleaves counting and sampling. ' 'Useful for background collection with -a sleep X.', type=int) g.add_argument('--sample-basename', help='Base name of sample perf.data files', default="perf.data") g.add_argument('-d', help=argparse.SUPPRESS, action='help') # prevent passing this to perf p.add_argument('--version', help=argparse.SUPPRESS, action='store_true') p.add_argument('--debug', help=argparse.SUPPRESS, action='store_true') # enable scheduler debugging p.add_argument('--dfilter', help=argparse.SUPPRESS, action='append') p.add_argument('--repl', action='store_true', help=argparse.SUPPRESS) # start python repl after initialization p.add_argument('--filterquals', help=argparse.SUPPRESS, action='store_true') # remove events not supported by perf p.add_argument('--setvar', help=argparse.SUPPRESS, action='append') # set env variable (for test suite iterating options) p.add_argument('--tune', nargs='+', help=argparse.SUPPRESS) # override global variables with python expression p.add_argument('--tune-model', nargs='+', help=argparse.SUPPRESS) # override global variables late with python expression p.add_argument('--force-bn', action='append', help=argparse.SUPPRESS) # force bottleneck for testing p.add_argument('--no-json-header', action='store_true', help=argparse.SUPPRESS) # no [ for json p.add_argument('--no-json-footer', action='store_true', help=argparse.SUPPRESS) # no ] for json p.add_argument('--no-csv-header', action='store_true', help=argparse.SUPPRESS) # no header/version for CSV p.add_argument('--no-csv-footer', action='store_true', help=argparse.SUPPRESS) # no version for CSV p.add_argument('--no-version', action='store_true', help="Don't print version") args, rest = p.parse_known_args() io_set_args(args) if args.setvar: for j in args.setvar: l = j.split("=") os.environ[l[0]] = l[1] return args, rest def output_count(): return args.per_core + args.global_ + args.per_thread + args.per_socket def multi_output(): return output_count() > 1 def open_output_files(args): if args.valcsv: try: args.valcsv = flex_open_w(args.valcsv) except IOError as e: sys.exit("Cannot open valcsv file %s: %s" % (args.valcsv, e)) if args.perf_output: try: args.perf_output = flex_open_w(args.perf_output) except IOError as e: sys.exit("Cannot open perf output file %s: %s" % (args.perf_output, e)) def init_xlsx(args): args.set_xlsx = True if args.output: sys.exit("-o / --output not allowed with --xlsx") if args.valcsv: sys.exit("--valcsv not allowed with --xlsx") if args.perf_output: sys.exit("--perf-output not allowed with --xlsx") if args.csv: sys.exit("-c / --csv not allowed with --xlsx") if args.thread: sys.exit("--thread not supported with --xlsx") # XXX if not args.xlsx.endswith(".xlsx"): sys.exit("--xlsx must end in .xlsx") xlsx_base = re.sub(r'\.xlsx$', '.csv', args.xlsx) args.valcsv = re.sub(r'\.csv$', '-valcsv.csv', xlsx_base) args.perf_output = re.sub(r'\.csv$', '-perf.csv', xlsx_base) args.output = xlsx_base if args.xchart: args.xnormalize = True args.verbose = True forced_per_socket = False forced_per_core = False def set_xlsx(args): if not args.interval: args.interval = 1000 args.csv = ',' if args.xlsx: args.summary = True if not args.single_thread: args.per_thread = True args.split_output = True global forced_per_socket if args.per_socket: forced_per_socket = True global forced_per_core if args.per_core: forced_per_core = True args.per_socket = True args.per_core = True args.no_aggr = True args.global_ = True def do_xlsx(env, args): cmd = "%s %s/tl-xlsx.py --valcsv '%s' --perf '%s' --cpuinfo '%s' " % ( sys.executable, exe_dir(), args.valcsv.name, args.perf_output.name, env.cpuinfo if env.cpuinfo else "/proc/cpuinfo") if args.single_thread: names = ["program"] files = [args.output if args.output else "program"] else: names = ((["socket"] if args.per_socket else []) + (["core"] if args.per_core else []) + ["global", "thread"]) files = [tl_output.output_name(args.output, p) for p in names] extrafiles = [] extranames = [] charts = [] if args.xnormalize: for j, n in zip(files, names): nname = j.replace(".csv", "-norm.csv") ncmd = "%s %s/interval-normalize.py --normalize-cpu --error-exit < '%s' > '%s'" % ( sys.executable, exe_dir(), j, nname) if not args.quiet: print(ncmd) ret = os.system(ncmd) if ret: warn("interval-normalize failed: %d" % ret) return ret extrafiles.append(nname) extranames.append("n" + n) if args.xchart: charts.append("n" + n) cmd += " ".join(["--%s '%s'" % (n, f) for n, f in zip(names, files)]) cmd += " " + " ".join(["--add '%s' '%s'" % (f, n) for n, f in zip(extranames, extrafiles)]) cmd += " " + " ".join(["--chart '%s'" % f for f in charts]) cmd += " '%s'" % args.xlsx if not args.quiet: print(cmd) ret = os.system(cmd) if not args.keep: for fn in files + extrafiles: os.remove(fn) return ret def gentmp(o, cpu): o = re.sub(r'\.(xz|gz)', '', o) if o.endswith(".csv"): return o.replace(".csv", "-cpu%d.csv" % cpu) return o + "-cpu%d" % cpu def output_to_tmp(arg, outfn, args): if not args.output or args.xlsx: arg.insert(1, "-o" + outfn) elif update_arg(arg, "--output", "=", outfn): pass elif update_arg(arg, "-o", "", outfn): pass else: # does not handle -o combined with other one letter options sys.exit("Use plain -o / --output argument with --parallel") def merge_files(files, outf, args): for j in files: tl_output.catrmfile(j, outf, args.keep) # run multiple subset toplevs in parallel and merge the results def run_parallel(args, env): procs = [] pofns = [] valfns = [] sums = [] targ = copy(sys.argv) del targ[targ.index("--parallel")] ichunk = os.path.getsize(args.import_) / args.pjobs fileoff = 0 for cpu in range(args.pjobs): arg = copy(targ) if args.xlsx: del_arg_val(arg, "--xlsx") arg = [arg[0], "--set-xlsx", "--perf-output=X", "--valcsv=X"] + arg[1:] outfn = gentmp(args.output if args.output else "toplevo%d" % os.getpid(), cpu) output_to_tmp(arg, outfn, args) if args.perf_output or args.xlsx: pofn = gentmp(args.perf_output if args.perf_output else "toplevp", cpu) update_arg(arg, "--perf-output", "=", pofn) pofns.append(pofn) if args.valcsv or args.xlsx: valfn = gentmp(args.valcsv if args.valcsv else "toplevv", cpu) update_arg(arg, "--valcsv", "=", valfn) valfns.append(valfn) end = "" if cpu < args.pjobs-1: end = "%d" % (fileoff + ichunk) arg.insert(1, ("--subset=%d-" % fileoff) + end) fileoff += ichunk if args.json and args.pjobs > 1: if cpu > 0: arg.insert(1, "--no-json-header") if cpu < args.pjobs - 1 or args.summary: arg.insert(1, "--no-json-footer") sumfn = None if args.summary: del arg[arg.index("--summary")] sumfn = gentmp("toplevs%d" % os.getpid(), cpu) arg.insert(1, "--perf-summary=" + sumfn) sums.append(sumfn) if args.pjobs > 1: if cpu > 0: arg.insert(1, "--no-csv-header") if cpu < args.pjobs - 1 or args.summary: arg.insert(1, "--no-csv-footer") if not args.quiet: print(" ".join(arg)) pp = subprocess.Popen(arg, stdout=subprocess.PIPE, **popentext) # type: ignore procs.append((pp, outfn)) if args.xlsx: init_xlsx(args) set_xlsx(args) logfiles, logf = tl_output.open_all_logfiles(args, args.output) for p in procs: ret = p[0].wait() if ret: sys.exit("Subprocess toplev failed %d" % ret) tl_output.catrmoutput(p[1], logf, logfiles, args.keep) ret = 0 if sums: cmd = [sys.executable, exe_dir() + "/interval-merge.py"] + sums if not args.quiet: print(" ".join(cmd)) inp = subprocess.Popen(cmd, stdout=subprocess.PIPE) outfn = "toplevm%d" % os.getpid() output_to_tmp(targ, outfn, args) if args.xlsx: del_arg_val(targ, "--xlsx") targ.insert(1, "--set-xlsx") if args.perf_output: del_arg_val(targ, "--perf-output") if args.valcsv: del_arg_val(targ, "--valcsv") update_arg(targ, "--import", "=", "/dev/stdin") targ.insert(1, "--no-output") if args.json: targ.insert(1, "--no-json-header") else: targ.insert(1, "--no-csv-header") if not args.quiet: print(" ".join(targ)) outp = subprocess.Popen(targ, stdin=inp.stdout) ret = inp.wait() if ret: sys.exit("interval-merge failed") ret = outp.wait() if ret: sys.exit("summary toplev failed") tl_output.catrmoutput(outfn, logf, logfiles, args.keep) if not args.keep: for j in sums: os.remove(j) open_output_files(args) merge_files(valfns, args.valcsv, args) merge_files(pofns, args.perf_output, args) if args.xlsx: ret = do_xlsx(env, args) # XXX graph return ret def init_idle_threshold(args): if args.idle_threshold: idle_threshold = args.idle_threshold / 100. elif args.csv or args.xlsx or args.set_xlsx: # not for args.graph idle_threshold = 0 # avoid breaking programs that rely on the CSV output else: idle_threshold = 0.05 return idle_threshold ret_latency = None def setup_retlatency(args): global ret_latency if args.ret_latency: try: l = args.ret_latency.split(":") ret_latency = json.load(open(l[0]))["Data"] except IOError: sys.exit("Cannot open %s" % l[0]) except KeyError: sys.exit("retlat file has unparseable format") else: load_default_retlat() def lookup_retlat(event): if ret_latency is None: warn_once("No --ret-latency for %s" % event) return 1.0 try: l = args.ret_latency.split(":") if args.ret_latency else () o = ret_latency[event] return o["MEAN"] for k in l[1:]: if k in o[0] or k.upper() in o[0]: o = o[0][k] # XXX check for remaining arguments break if k not in o[1] and k.upper() in o[1]: k = k.upper() o = o[1][k] if type(o) is list: o = o[0]["MEAN"] return o except KeyError as e: warn_once("bad ret latency key %s" % e) return 1.0 # XXX def gen_cpu_name(cpu): if cpu == "simple": c = event_download.get_cpustr() if not c.startswith("GenuineIntel"): # fix github runner c = "GenuineIntel-6-4E" return c if cpu == "sprmax": cpu = "spr" for j in known_cpus: if cpu == j[0]: if len(j[1]) > 0 and isinstance(j[1][0], tuple): return "GenuineIntel-6-%02X-%d" % j[1][0] else: if len(j[1]) == 0: sys.exit("Cannot generate cpu name for %s" % cpu) if j[1][0] in eventlist_alias: return eventlist_alias[j[1][0]] # type: ignore return "GenuineIntel-6-%02X" % j[1][0] # type: ignore sys.exit("Unknown cpu %s" % cpu) return None def update_args(args, env): if args.force_cpu: env.forcecpu = args.force_cpu cpuname = gen_cpu_name(args.force_cpu) if not os.getenv("EVENTMAP"): os.environ["EVENTMAP"] = cpuname if not os.getenv("UNCORE"): os.environ["UNCORE"] = cpuname if args.force_topology: if not os.getenv("TOPOLOGY"): os.environ["TOPOLOGY"] = args.force_topology ocperf.topology = None # force reread if args.force_cpuinfo: env.cpuinfo = args.force_cpuinfo if args.force_hypervisor: env.hypervisor = True if args.sample_repeat: args.run_sample = True if args.handle_errata: args.ignore_errata = False if args.exclusive and args.pinned: sys.exit("--exclusive and --pinned cannot be combined") def handle_parallel(args, env): if args.parallel: if not args.import_: sys.exit("--parallel requires --import") if args.import_.endswith(".xz") or args.import_.endswith(".gz"): sys.exit("Uncompress input file first") # XXX if args.perf_summary: sys.exit("--parallel does not support --perf-summary") # XXX if args.subset: # XXX support sample sys.exit("--parallel does not support --subset") if args.json and multi_output() and not args.split_output: sys.exit("--parallel does not support multi-output --json without --split-output") if args.graph: sys.exit("--parallel does not support --graph") # XXX if args.pjobs == 0: import multiprocessing args.pjobs = multiprocessing.cpu_count() sys.exit(run_parallel(args, env)) def handle_rest(args, rest): if rest[:1] == ["--"]: rest = rest[1:] if args.cpu: rest = ["--cpu", args.cpu] + rest if args.pid: rest = ["--pid", args.pid] + rest if args.csv and len(args.csv) != 1: sys.exit("--csv/-x argument can be only a single character") if args.xlsx: init_xlsx(args) if args.set_xlsx: set_xlsx(args) return rest def update_args2(args): if args.perf_summary: try: args.perf_summary = flex_open_w(args.perf_summary) except IOError as e: sys.exit("Cannot open perf summary file %s: %s" % (args.perf_summary, e)) # XXX force no_uncore because the resulting file cannot be imported otherwise? if args.all: args.tsx = True args.power = True args.sw = True args.metrics = True args.frequency = True args.level = 6 if args.only_bottleneck: args.quiet = True args.no_version = True def handle_graph(args): graphp = None if args.graph: if not args.interval: args.interval = 100 extra = "" if args.title: title = args.title else: title = "cpu %s" % (args.graph_cpu if args.graph_cpu else 0) extra += '--title "' + title + '" ' if args.split_output: sys.exit("--split-output not allowed with --graph") if args.output: extra += '--output "' + args.output + '" ' if args.graph_cpu: extra += "--cpu " + args.graph_cpu + " " args.csv = ',' cmd = "%s %s/tl-barplot.py %s /dev/stdin" % (sys.executable, exe_dir(), extra) if not args.quiet: print(cmd) graphp = subprocess.Popen(cmd, shell=True, stdin=subprocess.PIPE, **popentext) # type: ignore args.output = graphp.stdin return graphp def init_ring_filter(args): ring_filter = "" if args.kernel and not args.user: ring_filter = 'k' if args.user and not args.kernel: ring_filter = 'u' return ring_filter MAX_ERROR = 0.05 def check_ratio(l): if args.verbose: return True return 0 - MAX_ERROR < l < 1 + MAX_ERROR def update_args_cpu(args, pversion): if args.level < 0: if args.bottlenecks: args.level = 4 else: args.level = 2 if any([x >= 8 for x in cpu.counters.values()]) else 1 if args.show_cpu: print("%s %s %s" % (cpu.true_name, cpu.pmu_name, cpu.name)) sys.exit(0) desired_cpu = args.force_cpu if args.force_cpu else cpu.model if desired_cpu in eventlist_alias: r = eventlist_alias[desired_cpu] if not os.getenv("EVENTMAP"): os.environ["EVENTMAP"] = r if not os.getenv("UNCORE"): os.environ["UNCORE"] = r if cpu.pmu_name and cpu.pmu_name.startswith("generic") and not args.quiet: print("warning: kernel is in architectural mode and might mismeasure events", file=sys.stderr) print("Consider a kernel update. See https://github.com/andikleen/pmu-tools/wiki/toplev-kernel-support", file=sys.stderr) if cpu.cpu in hybrid_cpus: sys.exit("Hybrid %s not supported in architectural mode" % cpu.cpu) if args.xlsx and not forced_per_socket and cpu.sockets == 1: args.per_socket = False if args.xlsx and not forced_per_core and cpu.threads == 1: args.per_core = False if cpu.hypervisor: feat.max_precise = 0 feat.has_max_precise = True if not pversion.has_uncore_expansion: # XXX reenable power args.no_uncore = True if cpu.hypervisor or args.no_uncore: feat.supports_power = False def print_perf(r): if not (args.perf or args.print): return l = ["'" + x + "'" if x.find("{") >= 0 else x for x in r] l = [x.replace(";", "\\;") for x in l] i = l.index('--log-fd') del l[i:i+2] s = " ".join(l) if len(l) > 2: s = s.replace("},", "},\n") print(s) sys.stdout.flush() def gen_script(r): print("#!/bin/sh") print("# Generated from 'toplev " + " ".join(sys.argv) + " for CPU " + cpu.cpu) print("# Show output with toplev.py " + " ".join([x for x in sys.argv if x != "--gen-script"]) + " --import toplev_perf%s --force-cpuinfo toplev_cpuinfo --force-topology toplev_topology --force-cpu " % ( ".data" if args.script_record else ".csv") + cpu.cpu) print("# print until Ctrl-C or run with command on command line (e.g. -a -I 1000 sleep 10)") print("# override output file names with OUT=... script (default toplev_...)") if not args.script_record: print("# enable compression with POSTFIX=.xz script") print("OUT=${OUT:-toplev}") print("PERF=${PERF:-perf}") print("find /sys/devices > ${OUT}_topology") print("cat /proc/cpuinfo > ${OUT}_cpuinfo") r[0] = "$PERF" i = r.index('--log-fd') r[i] = "-o" if args.script_record: r[i + 1] = "${OUT}_perf.data" i = r.index("stat") r[i] = "stat record --quiet" else: r[i + 1] = "${OUT}_perf.csv${POSTFIX}" i = r.index('-x;') if args.script_record: del r[i] else: r[i] = '-x\\;' i = r.index('-e') r[i+1] = "'" + r[i+1] + "'" print(" ".join(r + ['"$@"'])) class PerfRun(object): """Control a perf subprocess.""" def __init__(self): self.skip_to_next_ts = False self.end_seek_offset = None self.sample_prob = None self.skip_line = False self.perf = None self.offset = None def handle_inputsubset(self, f, iss): m = re.match(r'(\d+)-?(\d+)?$', iss) if m: off = int(m.group(1)) f.seek(off) if m.group(2): self.end_seek_offset = int(m.group(2)) + 1 if off: self.skip_to_next_ts = True self.skip_line = True return m = re.match(r'(\d+)/([0-9.]+)%(-)?$', iss) if m: f.seek(0, 2) size = f.tell() chunk = int(size * (float(m.group(2)) / 100.)) nth = int(m.group(1)) if (nth+1)*chunk > size: sys.exit("--subset %s out of range" % iss) f.seek(chunk * nth) if m.group(3) is None: self.end_seek_offset = chunk * (1+nth) + 1 if chunk * nth != 0: self.skip_to_next_ts = True self.skip_line = True return m = re.match('sample:([0-9.]+)%?$', iss) if m: self.sample_prob = float(m.group(1)) / 100. self.random = random.Random() s = os.getenv("TLSEED") if s: self.random.seed(int(s)) self.sampling = False return sys.exit("Unparseable --subset %s" % iss) def skip_input(self): if self.skip_to_next_ts: return True if self.sample_prob: return not self.sampling return False # must be stored before reading the line def store_offset(self): if self.end_seek_offset: self.offset = self.inputf.tell() # type: ignore def skip_first_line(self): if self.skip_line: self.skip_line = False return True return False def next_timestamp(self): if self.end_seek_offset: if self.end_seek_offset <= self.offset: return True self.skip_to_next_ts = False if self.sample_prob: r = self.random.random() self.sampling = r < self.sample_prob return False def execute(self, r): if args.import_: print_perf(r) if args.script_record: self.perf = subprocess.Popen([feat.perf, "stat", "report", "-x;", "-i", args.import_], stderr=subprocess.PIPE, **popentext) # type: ignore return self.perf.stderr self.perf = None try: f = flex_open_r(args.import_) except IOError as e: sys.exit("Cannot open file %s: %s" % (args.import_, e)) if args.subset: try: self.handle_inputsubset(f, args.subset) except OSError: sys.exit("--subset not supported on compressed or unseekable files.") except io.UnsupportedOperation: sys.exit("--subset not supported on compressed or unseekable files.") self.inputf = f return f if args.gen_script: gen_script(r) sys.exit(0) outp, inp = pty.openpty() if 'set_inheritable' in os.__dict__: os.set_inheritable(inp, True) n = r.index("--log-fd") r[n + 1] = "%d" % (inp) print_perf(r) if args.print: sys.exit(0) self.perf = subprocess.Popen(r, close_fds=False) # type: ignore os.close(inp) return os.fdopen(outp, 'r') def wait(self): ret = 0 if self.perf: ret = self.perf.wait() self.perf = None return ret def __del__(self): if self.perf: self.perf.kill() def separator(x): if ":" in x: return "" if x.startswith("cpu"): return "" return ":" def add_filter_event(e): if "/" in e and not e.startswith("cpu"): return e if e == "dummy" or e == "emulation-faults" or e == "duration_time": return e s = separator(e) if not e.endswith(s + args.ring_filter): return e + s + args.ring_filter return e def add_filter(s): if args.ring_filter: s = list(map(add_filter_event, s)) return s def is_cpu_event(s): return re.match(r'cpu(_atom|_core)?/', s) is not None def initialize_event(name, i, e): if "." in name or "_" in name and name not in non_json_events: eo = e.output(noname=not INAME, noexplode=True) ectx.emap.update_event(eo, e) ectx.emap.update_event(remove_qual(eo), e) # XXX if (e.counter not in ectx.standard_counters and not name.startswith("UNC_")): if e.counter.startswith("Fixed"): ectx.limited_counters[i] = int(e.counter.split()[2]) + FIXED_BASE ectx.fixed_events.add(i) elif e.counter.isdecimal() and int(e.counter) >= 32: ectx.limited_counters[i] = int(e.counter) - 32 + FIXED_BASE ectx.fixed_events.add(i) else: # for now use the first counter only to simplify # the assignment. This is sufficient for current # CPUs ectx.limited_counters[i] = int(e.counter.split(",")[0]) ectx.limited_set.add(i) nameu = e.name.upper() if nameu in ectx.constraint_fixes: e.counter = ectx.constraint_fixes[nameu] for k, v in ectx.constraint_patterns: if nameu.startswith(k): e.counter = v if e.counter == ectx.limit4_counters or eo.endswith("/p"): ectx.limit4_events.add(i) if e.errata and e.errata != "0" and e.errata != "null": if e.errata not in ectx.errata_whitelist: ectx.errata_events[name] = e.errata else: ectx.errata_warn_events[name] = e.errata if ('pebs' in e.__dict__ and e.pebs == 2) or name.startswith("FRONTEND_"): ectx.require_pebs_events.add(name) else: non_json_events.add(i) valid_events.add_event(i) if INAME: valid_events.add_event(ocperf.gen_name(re.sub(r':.*','', name), False)) # XXX sup, handle :... uniquely if not is_cpu_event(i) and i not in ectx.fixed_events: if not i.startswith("uncore"): valid_events.add_event(i) if i.startswith("msr/"): ectx.sched_ignore_events.add(i) else: ectx.outgroup_events.add(add_filter_event(i)) def raw_event(i, name="", period=False, initialize=False): e = None orig_i = i if i == "cycles" and (cpu.cpu in hybrid_cpus or cached_exists("/sys/devices/cpu_core")): i = "cpu_clk_unhalted.thread" if "." in i or "_" in i and i not in non_json_events: if not cpu.ht: i = i.replace(":percore", "") extramsg = [] # type: List[str] e = ectx.emap.getevent(i, nocheck=event_nocheck(), extramsg=extramsg) if e is None: if i not in ectx.notfound_cache: ectx.notfound_cache[i] = extramsg[0] if not args.quiet: print("%s %s" % (i, extramsg[0]), file=sys.stderr) return "dummy" if has(e, 'perfqual') and not cached_exists("/sys/devices/%s/format/%s" % (ectx.emap.pmu, e.perfqual)): if not args.quiet: print("%s event not supported in hypervisor or architectural mode" % i, file=sys.stderr) return "dummy" if re.match("^[0-9]", name): name = "T" + name if args.filterquals: e.filter_qual() i = e.output(noname=not INAME, name=name, period=period, noexplode=True) if not ectx.force_metrics: m = re.search(r'(topdown-[a-z-]+)', i) if m and not cached_exists("/sys/devices/%s/events/%s" % (ectx.emap.pmu, m.group(1))): if not args.quiet: print("%s event not supported in sysfs" % m.group(1)) i = "dummy" if initialize: initialize_event(orig_i, i, e) return i # generate list of converted raw events from events string def raw_events(evlist, initialize=False): return [raw_event(x, initialize=initialize) for x in evlist] def mark_fixed(s): r = raw_event(s) if r in ectx.fixed_events: return "%s[F]" % s return s def pwrap(s, linelen=70, indent=""): print(indent + ("\n" + indent).join( textwrap.wrap(s, linelen, break_long_words=False)), file=sys.stderr) def pwrap_not_quiet(s, linelen=70, indent=""): if not args.quiet: pwrap(s, linelen, indent) def perf_args(evstr, rest): add = [] if args.interval: add += ['-I', str(args.interval)] return [feat.perf, "stat", "-x;", "--log-fd", "X"] + add + ["-e", evstr] + rest def setup_perf(evstr, rest): prun = PerfRun() inf = prun.execute(perf_args(evstr, rest)) return inf, prun class Stat(object): def __init__(self): self.total = 0 self.errors = Counter() # type: typing.Counter[str] def print_not(a, count, msg, j): print(("%s %s %s %.2f%% in %d measurements" % (j, j, msg, # XXX rmap again with ectx 100.0 * (float(count) / float(a.total)), a.total)), file=sys.stderr) # XXX need to get real ratios from perf def print_account(ad): total = Counter() # type: typing.Counter[str] for j in ad: a = ad[j] for e in a.errors: if args.stats: print_not(a, a.errors[e], e, j) total[e] += 1 if sum(total.values()) > 0 and not args.quiet: print(", ".join(["%d events %s" % (num, e) for e, num in total.items()]), file=sys.stderr) class ValidEvents(object): def update(self): self.string = "|".join(self.valid_events) def __init__(self): self.valid_events = [r"cpu(_core|_atom)?/.*?/", "uncore.*?/.*?/", "ref-cycles", "power.*", r"msr.*", "emulation-faults", r"r[0-9a-fA-F]+", "cycles", "instructions", "dummy", "slots", r"topdown-(fe-bound|be-bound|retiring|bad-spec|heavy-ops|br-mispredict|fetch-lat|mem-bound)"] self.update() def add_event(self, ev): if re.match(self.string, ev): return # add first to overwrite more generic regexprs list r... self.valid_events.insert(0, ev) self.update() valid_events = ValidEvents() def is_event(l, n): if len(l) <= n: return False # use static string to make regexpr caching work return re.match(valid_events.string, l[n], re.I) def set_interval(env, d, interval): env['interval-ns'] = d * 1e9 env['interval-ms'] = d * 1e3 env['interval-s'] = d env['interval'] = interval def key_to_coreid(k): try: x = cpu.cputocore[int(k)] return x[0] * 1000 + x[1] except ValueError: return 0 def key_to_socketid(k): try: return cpu.cputocore[int(k)][0] except ValueError: return 0 def core_fmt(core): if cpu.sockets > 1: return "S%d-C%d" % (core / 1000, core % 1000,) return "C%d" % (core % 1000,) def socket_fmt(j): return "S%d" % cpu.cputocore[j][0] def thread_fmt(j): return core_fmt(key_to_coreid(j)) + ("-T%d" % cpu.cputothread[int(j)]) def display_core(cpunum, ignore_thread=False): for match in args.core.split(","): m = re.match(r'(?P<socket>S\d+)?-?(?P<core>C\d+)?-?(?P<thread>T\d+)?', match, re.I) if not m or not any((m.group('core'), m.group('socket'),)): sys.exit("Bad core match %s" % match) def matching(name, mapping): return mapping[cpunum] == int(m.group(name)[1:]) # type: ignore if m.group('socket') and not matching('socket', cpu.cputosocket): continue if m.group('core') and cpu.cputocore[cpunum][1] != int(m.group('core')[1:]): continue if not ignore_thread and m.group('thread') and not matching('thread', cpu.cputothread): continue return True return False OUTPUT_CORE_THREAD = 0 OUTPUT_CORE = 1 OUTPUT_THREAD = 2 OUTPUT_SOCKET = 3 OUTPUT_GLOBAL = 4 def display_keys(runner, keys, mode, post=""): allowed_threads = runner.cpu_list if mode == OUTPUT_GLOBAL: return ("",) if len(keys) > 1 and smt_mode: if mode == OUTPUT_SOCKET: all_cpus = dedup(map(socket_fmt, allowed_threads)) else: cores = [key_to_coreid(x) for x in keys if int(x) in allowed_threads] if mode != OUTPUT_CORE: threads = [thread_fmt(x) + post for x in allowed_threads] else: threads = [] all_cpus = [core_fmt(x)+post for x in cores] + threads else: all_cpus = [x + post for x in keys] if any(map(package_node, runner.olist)): all_cpus += ["S%d" % x for x in range(cpu.sockets)] return all_cpus def verify_rev(rev, cpus): for k in cpus: for ind, o in enumerate(rev[k]): assert o == rev[cpus[0]][ind] assert len(rev[k]) == len(rev[cpus[0]]) def is_cycles(ev): return re.match(r'cycles|slots|r20003c|cpu(_atom|_core)?/slots|cpu(_atom|_core)?/event=0x(3c|a4),umask=0x(0|1)', ev) is not None def find_cycles(rev): for l in rev.values(): for idle_ev in l: if is_cycles(idle_ev): return idle_ev return "" def find_idle_keys(res, rev, idle_thresh): if sum([len(res[k]) for k in res.keys()]) == 0: return set() if len(res.keys()) == 1: return set() idle_ev = find_cycles(rev) if idle_ev == "": warn_once("no idle detection because no cycle event found") return set() cycles = { k: max([0] + [val for val, ev in zip(res[k], rev[k]) if ev == idle_ev]) for k in res.keys() } if sum(cycles.values()) == 0 and not args.quiet: print_once("no idle detection because cycles counts are zero") return set() max_cycles = max(cycles.values()) return {k for k in cycles.keys() if cycles[k] < max_cycles * idle_thresh} def is_idle(cpus, idle_keys): return all([("%d" % c) in idle_keys for c in cpus]) def idle_core(core, idle_keys): return is_idle(cpu.coreids[core], idle_keys) def idle_socket(socket, idle_keys): return is_idle(cpu.sockettocpus[socket], idle_keys) # from https://stackoverflow.com/questions/4836710/does-python-have-a-built-in-function-for-string-natural-sort def num_key(s): return [int(t) if t.isdigit() else t for t in re.split(r'(\d+)', s)] def invalid_res(res, key, nothing): if len(res) == 0: if isinstance(key, list): for j in key: nothing.add(j) else: nothing.add(key) return True return False def runner_name(r): if r.pmu is None: return "cpu" return r.pmu.replace("cpu_", "") default_compute_iter = 3 # override how often to recompute to converge all the thresholds COMPUTE_ITER = None def print_keys(runner, res, rev, valstats, out, interval, env, mode, runner_list): nothing = set() # type: Set[str] allowed_threads = runner.cpu_list def filtered(j): return j != "" and j.isdecimal() and int(j) not in allowed_threads core_node = lambda obj: safe_ref(obj, 'domain') in runner.ectx.core_domains thread_node = lambda obj: not (core_node(obj) or package_node(obj)) idle_keys = find_idle_keys(res, rev, runner.idle_threshold) idle_mark_keys = find_idle_keys(res, rev, IDLE_MARKER_THRESHOLD) printer = runner.printer hidden_keys = set() stat = runner.stat keys = sorted(res.keys(), key=num_key) post = "" if runner.pmu != "cpu": if len(res.keys()) > 1: post += "-" post += runner_name(runner) out.set_cpus(display_keys(runner, keys, mode, post)) runner.printer.numprint = 0 if smt_mode: printed_cores = set() printed_sockets = set() for j in list(keys): if j != "" and int(j) not in cpu.cputocore: warn_once("Warning: input cpu %s not in cpuinfo." % j) del res[j] keys = sorted(res.keys(), key=num_key) for j in keys: if filtered(j): continue sid = key_to_socketid(j) core = key_to_coreid(j) if mode == OUTPUT_CORE and core in printed_cores: continue if mode == OUTPUT_SOCKET and sid in printed_sockets: continue if j in idle_keys: hidden_keys.add(j) continue runner.reset_thresh() if mode == OUTPUT_GLOBAL: cpus = keys elif mode == OUTPUT_SOCKET: cpus = [x for x in keys if key_to_socketid(x) == sid and not filtered(x)] else: cpus = [x for x in keys if key_to_coreid(x) == core and not filtered(x)] combined_res = list(zip(*[res[x] for x in cpus])) combined_st = [combine_valstat(z) for z in zip(*[valstats[x] for x in cpus])] env['num_merged'] = len(cpus) if mode in (OUTPUT_CORE,OUTPUT_SOCKET,OUTPUT_GLOBAL): merged_res = combined_res merged_st = combined_st else: merged_res = res[j] merged_st = valstats[j] if invalid_res(merged_res, cpus, nothing): continue # may need to repeat to get stable threshold values # in case of mutual dependencies between SMT and non SMT # but don't loop forever (?) used_stat = stat onemore = False iterations = COMPUTE_ITER if COMPUTE_ITER else default_compute_iter for _ in range(iterations): env['num_merged'] = 1 changed = runner.compute(merged_res, rev[j], merged_st, env, thread_node, used_stat, runner_list) verify_rev(rev, cpus) env['num_merged'] = len(cpus) changed += runner.compute(combined_res, rev[cpus[0]], combined_st, env, core_node, used_stat, runner_list) if changed == 0 and COMPUTE_ITER is None: # do always one more so that any thresholds depending on a later node are caught if not onemore: onemore = True continue break used_stat = None # find bottleneck bn = find_bn(runner.olist, not_package_node) if mode == OUTPUT_GLOBAL: printer.print_res(runner.olist, out, interval, "", not_package_node, bn) break if mode == OUTPUT_SOCKET: printer.print_res(runner.olist, out, interval, socket_fmt(int(j)), not_package_node, bn, idle_socket(sid, idle_mark_keys)) printed_sockets.add(sid) continue if mode == OUTPUT_THREAD: runner.compute(res[j], rev[j], valstats[j], env, package_node, stat, runner_list) printer.print_res(runner.olist, out, interval, thread_fmt(int(j))+post, any_node, bn, j in idle_mark_keys) continue # per core or mixed core/thread mode # print the SMT aware nodes if core not in printed_cores: printer.print_res(runner.olist, out, interval, core_fmt(core)+post, core_node, bn, idle_core(core, idle_mark_keys)) printed_cores.add(core) # print the non SMT nodes if mode == OUTPUT_CORE: fmt = core_fmt(core) idle = idle_core(core, idle_mark_keys) else: fmt = thread_fmt(int(j)) idle = j in idle_mark_keys printer.print_res(runner.olist, out, interval, fmt+post, thread_node, bn, idle) elif mode != OUTPUT_GLOBAL: env['num_merged'] = 1 for j in keys: if filtered(j): continue if j in idle_keys: hidden_keys.add(j) continue if invalid_res(res[j], j, nothing): continue runner.reset_thresh() runner.compute(res[j], rev[j], valstats[j], env, not_package_node, stat, runner_list) bn = find_bn(runner.olist, not_package_node) printer.print_res(runner.olist, out, interval, j+post, not_package_node, bn, j in idle_mark_keys) if mode == OUTPUT_GLOBAL: env['num_merged'] = 1 cpus = [x for x in keys if not filtered(x)] if cpus: combined_res = [sum([res[j][i] for j in cpus]) for i in range(len(res[cpus[0]]))] combined_st = [combine_valstat([valstats[j][i] for j in cpus]) for i in range(len(valstats[cpus[0]]))] if smt_mode: nodeselect = package_node else: nodeselect = any_node if not invalid_res(combined_res, cpus, nothing): runner.reset_thresh() runner.compute(combined_res, rev[cpus[0]] if len(cpus) > 0 else [], combined_st, env, nodeselect, stat, runner_list) bn = find_bn(runner.olist, lambda x: True) printer.print_res(runner.olist, out, interval, "", nodeselect, bn, False) elif mode != OUTPUT_THREAD: packages = set() for j in keys: if j == "": continue if j.isdecimal(): if filtered(j): continue p_id = cpu.cputosocket[int(j)] if p_id in packages: continue packages.add(p_id) jname = "S%d" % p_id else: jname = j if j in idle_keys: hidden_keys.add(j) continue runner.reset_thresh() if invalid_res(res[j], j, nothing): continue runner.compute(res[j], rev[j], valstats[j], env, package_node, stat, runner_list) printer.print_res(runner.olist, out, interval, jname, package_node, None, j in idle_mark_keys) # no bottlenecks from package nodes for now out.flush() if not FUZZYINPUT: stat.referenced_check(res, runner.sched.evnum) stat.compute_errors() runner.idle_keys |= hidden_keys if nothing and not args.quiet: print("%s: Nothing measured%s" % (runner.pmu, " for " if len(nothing) > 0 and "" not in nothing else ""), " ".join(sorted(nothing)), file=sys.stderr) if runner.printer.numprint == 0 and not args.quiet and runner.olist: print("No node %scrossed threshold" % ( "for %s " % runner_name(runner) if runner.pmu != "cpu" else ""), file=sys.stderr) def print_and_split_keys(runner, res, rev, valstats, out, interval, env, rlist): if multi_output(): if args.per_thread: out.remark("Per thread") out.reset("thread") print_keys(runner, res, rev, valstats, out, interval, env, OUTPUT_THREAD, rlist) if args.per_core: out.remark("Per core") out.reset("core") print_keys(runner, res, rev, valstats, out, interval, env, OUTPUT_CORE, rlist) if args.per_socket: out.remark("Per socket") out.reset("socket") print_keys(runner, res, rev, valstats, out, interval, env, OUTPUT_SOCKET, rlist) if args.global_: out.remark("Global") out.reset("global") print_keys(runner, res, rev, valstats, out, interval, env, OUTPUT_GLOBAL, rlist) else: if args.split_output: sys.exit("--split-output needs --per-thread / --global / --per-socket / --per-core") mode = OUTPUT_CORE_THREAD if args.per_thread: mode = OUTPUT_THREAD elif args.per_core: mode = OUTPUT_CORE elif args.per_socket: mode = OUTPUT_SOCKET elif args.global_: mode = OUTPUT_GLOBAL print_keys(runner, res, rev, valstats, out, interval, env, mode, rlist) def print_check_keys(runner, res, rev, valstats, out, interval, env, rlist): if res and all([sum(res[k]) == 0.0 and len(res[k]) > 0 for k in res.keys()]) and cpu.cpu == cpu.realcpu: if args.subset: return if runner.pmu == "cpu": sys.exit("All measured values 0. perf broken?") else: if not args.quiet: print("Measured values for %s all 0" % runner_name(runner), file=sys.stderr) return if args.interval and interval is None: interval = float('nan') if not args.no_output: print_and_split_keys(runner, res, rev, valstats, out, interval, env, rlist) def print_summary(summary, out, runner_list, full_system): if args.perf_summary: p = summary.summary_perf for sv in zip_longest(*p.values()): for ind, title in enumerate(p.keys()): r = sv[ind] l = [] if args.interval: l.append("\tSUMMARY") if full_system: l.append(("CPU" + title) if re.match(r'\d+$', title) else title) if output_numcpus: l.append("0") # XXX if r is None: continue if title.isdecimal(): cpunum = int(title) if (r[2].startswith("uncore") or r[2].startswith("power")) and ( cpunum != cpu.sockettocpus[cpu.cputosocket[cpunum]][0]): continue if r[2].startswith("duration_time") and cpunum != 0 and not args.cpu and not args.core: continue args.perf_summary.write(";".join(l + ["%f" % r[0], r[1], r[2], "%f" % r[3], "%.2f" % r[4], "", ""]) + "\n") if not args.summary: return for runner, res, rev in runner_split(runner_list, summary.res, summary.rev): print_and_split_keys(runner, res, rev, summary.valstats, out, float('nan'), summary.env, runner_list) def is_outgroup(x): return set(x) - ectx.outgroup_events == set() class SaveContext(object): """Save (some) environment context, in this case stdin seek offset to make < file work when we reexecute the workload multiple times.""" def __init__(self): try: self.startoffset = sys.stdin.tell() # type: int | None except OSError: self.startoffset = None except IOError: self.startoffset = None def restore(self): if self.startoffset is not None: sys.stdin.seek(self.startoffset) def execute_no_multiplex(runner_list, out, rest, summary): results = [] # type: List[Any] groups = [] num_outg = 0 for runner in runner_list: groups += [g.evnum for g in runner.sched.evgroups] num_outg += sum([g.outgroup for g in runner.sched.evgroups]) num_runs = len(groups) - num_outg outg = [] n = 0 ctx = SaveContext() resoff = Counter() # type: typing.Counter[str] RES, REV, INTERVAL, VALSTATS, ENV = range(5) ret = 0 # runs could be further reduced by tweaking # the scheduler to avoid any duplicated events for runner in runner_list: groups = [g.evnum for g in runner.sched.evgroups] for g, gg in zip(groups, runner.sched.evgroups): if gg.outgroup: outg.append(g) continue print("RUN #%d of %d%s: %s" % (n + 1, num_runs, " for %s" % runner_name(runner) if len(runner_list) > 1 else "", " ".join([quote(o.name) for o in gg.objl]))) # becomes results for first iteration lresults = results if n == 0 else [] res = None events = outg + [g] runner.set_ectx() evstr = group_join(events) flat_events = flatten(events) flat_rmap = [event_rmap(e, runner_list) for e in flat_events] runner.clear_ectx() for nret, res, rev, interval, valstats, env in do_execute( [runner], summary, evstr, flat_rmap, out, rest, resoff, flat_events): ret = max(ret, nret) lresults.append([res, rev, interval, valstats, env]) if res: for t in res.keys(): resoff[t] += len(res[t]) if n > 0: if len(lresults) != len(results): if not args.quiet: print("Original run had %d intervals, this run has %d. " "Workload run time not stable?" % (len(lresults), len(results)), file=sys.stderr) if len(lresults) > len(results): # throw away excessive intervals lresults = lresults[:len(results)] else: # fill the missing intervals with dummy data v = lresults[0] for ind, _ in enumerate(results): if ind >= len(lresults): lresults.append([dummy_dict(v[RES]), v[REV], v[INTERVAL], dummy_dict(v[RES], ValStat(0,0)), v[ENV]]) assert len(lresults) == len(results) i = 0 for r, lr in zip(results, lresults): for j in (RES, REV, VALSTATS): append_dict(r[j], lr[j]) i += 1 ctx.restore() outg = [] n += 1 assert num_runs == n for res, rev, interval, valstats, env in results: if summary: summary.add(res, rev, valstats, env) for runner, res, rev in runner_split(runner_list, res, rev): print_check_keys(runner, res, rev, valstats, out, interval, env, runner_list) return ret def runner_split(runner_list, res, rev): for r in runner_list: if len(res.keys()) == 1 and "" in res: off = r.sched.offset end = off + len(r.sched.evnum) yield r, { "": res[""][off:end]}, { "": rev[""][off:end] } elif r.cpu_list: d = defaultdict(list) # type: DefaultDict[str,list] d.update({ "%d" % k: res["%d" % k] for k in r.cpu_list }) yield r, d, rev else: yield r, res, rev def execute(runner_list, out, rest, summary): evstr, flat_events, flat_rmap = "", [], [] for runner in runner_list: new_events = [x.evnum for x in runner.sched.evgroups if len(x.evnum) > 0] if len(new_events) == 0: continue runner.set_ectx() if evstr: evstr += "," evstr += group_join(new_events) new_flat_events = flatten(new_events) flat_events += new_flat_events flat_rmap += [event_rmap(e, runner_list) for e in new_flat_events] runner.clear_ectx() ctx = SaveContext() for ret, res, rev, interval, valstats, env in do_execute( runner_list, summary, evstr, flat_rmap, out, rest, Counter(), None): if summary: summary.add(res, rev, valstats, env) for runner, res, rev in runner_split(runner_list, res, rev): print_check_keys(runner, res, rev, valstats, out, interval, env, runner_list) ctx.restore() return ret def find_group(num, runner_list): offset = 0 for runner in runner_list: if num - offset < len(runner.sched.evnum): break offset += len(runner.sched.evnum) num -= offset groups = runner.sched.evgroups g = groups[bisect.bisect_right(groups, GroupCmp(num)) - 1] if g.base <= num < g.base + len(g.evnum): return g warn("group for event %d not found" % num) return None def dump_raw(valcsv, interval, title, event, ename, val, index, stddev, multiplex, runner_list): if index < 0: return g = find_group(index, runner_list) if g is None: return nodes = " ".join(sorted([o.name.replace(" ", "_") for o in g.objl if event in o.evnum])) if args.raw: print("raw", title, "event", event, "val", val, "ename", ename, "index", index, "group", g.num, "nodes", nodes) if args.valcsv: valcsv.writerow((interval, title, g.num, ename, val, event, index, stddev, multiplex, nodes)) def group_join(events): e = "" last = None sep = "" for j in events: e += sep # add dummy event to separate identical events to avoid merging # in perf stat if last == j[0] and sep: e += "emulation-faults," e += event_group(j) sep = "," last = j[-1] return e def update_perf_summary(summary, off, title, val, event, unit, multiplex): if title not in summary.summary_perf: summary.summary_perf[title] = [] if len(summary.summary_perf[title]) <= off: summary.summary_perf[title].append([val, unit, event, 0, multiplex]) else: r = summary.summary_perf[title][off] r[0] += val assert r[1] == unit assert r[2] == event or event == "dummy" r[3] = min(r[3], multiplex) def find_runner(rlist, off, title, event): if len(rlist) == 1 and rlist[0].pmu == "cpu": return rlist[0], off for r in rlist: if title == "": if r.sched.offset <= off < r.sched.offset+len(r.sched.evnum): return r, off - r.sched.offset elif r.cpu_list: # in the per cpu case each hybrid cpu has its own line, so no offsets for the runners # but need to handle leaks below if int(title) in r.cpu_list: # For hybrid, non cpu events like msr/tsc/ get expanded over all CPUs. # and leak into the other runner who doesn't know anything about them. # XXX check does not handle dups if not event.startswith("cpu") and (off >= len(r.sched.evnum) or event != r.sched.evnum[off]): return None, 0 return r, off else: return r, off return None, 0 def perf_name(e): m = re.search(r'name=([^,/]+)', e) if m: return m.group(1) return None FINE = 0 SKIP = 1 FUZZY = 2 def check_event(rlist, event, off, title, prev_interval, l, revnum, linenum, last_linenum): r, off = find_runner(rlist, off, title, event) if r is None: return r, FINE, event # likely event expanded over all CPUs if event.startswith("cpu") and not event.startswith(r.pmu): if args.debug: print("event wrong pmu", event, title, r.pmu) return None, FINE, event # cannot check because it's an event that needs to be expanded first if not event.startswith("cpu") and title.isdecimal() and int(title) not in r.cpu_list: return r, FINE, event if revnum is None: revnum = r.sched.evnum if event.startswith("uncore"): event = re.sub(r'_[0-9]+', '', event) try: expected_ev = remove_qual(revnum[off]) except IndexError: if FUZZYINPUT: expected_ev = event else: sys.exit("Out of range event %s offset %d (len %d). %s" % (event, off, len(revnum), "Mismatch in toplev arguments from recording?" if args.import_ else "")) if event != expected_ev: en = perf_name(expected_ev) if en == event: return r, FINE, expected_ev # work around perf bug that incorrectly expands uncore events in some versions if off > 0 and event == remove_qual(revnum[off - 1]): return None, FINE, expected_ev if FUZZYINPUT: return r, FUZZY, expected_ev # some perf version don't output <not counted/supported due to dd15480a3d67 # if the event is expected within a small window assume it's not counted # and reuse the value for the next near = revnum[off:off+PERF_SKIP_WINDOW] for j in near: if event == remove_qual(j): if args.debug: print("skipping, expected", expected_ev, "got", event, "off", off, "title", title, "context", revnum[off:off+PERF_SKIP_WINDOW]) if linenum == last_linenum[0]: # avoid endless loop return r, FINE, expected_ev last_linenum[0] = linenum return r, SKIP, expected_ev print("Event in input does not match schedule (%s vs expected %s [pmu:%s/ind:%d/tit:%s/int:%f+%d])." % ( event, expected_ev, r.pmu, off, title, prev_interval, linenum), file=sys.stderr) sys.stdout.write(l) if args.import_: sys.exit("Different arguments than original toplev?") sys.exit("Input corruption") return r, FINE, event def update_missing(res, rev, valstats, fallback): for k in rev.keys(): for ind, event in enumerate(rev[k]): if not isnan(res[k][ind]): continue key = (k, event) if key in fallback: if args.debug: print("updating fuzzy event", k, event, fallback[key][0]) res[k][ind] = fallback[key][0] valstats[k][ind] = fallback[key][1] continue pn = perf_name(event) if pn: key = (k, pn) if key in fallback: res[k][ind] = fallback[key][0] valstats[k][ind] = fallback[key][1] if args.debug: print("updating", k, rev[k][ind]) continue if not args.quiet: print("Cannot find value for", k, rev[k][ind], pn, "in input") res[k][ind] = float("nan") assert not any([x is None for x in res[k]]) def do_execute(rlist, summary, evstr, flat_rmap, out, rest, resoff, revnum): res = defaultdict(list) # type: DefaultDict[str,List[float]] rev = defaultdict(list) # type: DefaultDict[str, List[str]] valstats = defaultdict(list) # type: DefaultDict[str,List[ValStat]] env = {} # type: Dict[str,str] account = defaultdict(Stat) # type: DefaultDict[str,Stat] inf, prun = setup_perf(evstr, rest) prev_interval = 0.0 interval = None interval_dur = 0.0 linenum = 1 skip = False last_linenum = [0] fallback = {} # type: Dict[Tuple[str,str], Tuple[float, ValStat]] need_fallback = False if not args.import_ and not args.interval: start = time.time() while True: if skip: # when skipping reuse line from last iteration pass else: try: prun.store_offset() l = inf.readline() origl = l if not l: break # some perf versions break CSV output lines incorrectly for power events if l.endswith("Joules"): l2 = inf.readline() l = l + l2.strip() if l.startswith("#") or l.strip() == "": linenum += 1 skip = False continue except OSError: # handle pty EIO break except IOError: break except KeyboardInterrupt: continue if re.match(r'^(Timestamp|Value|Location)', l): # header generated by toplev in import mode. ignore. linenum += 1 skip = False continue if prun.skip_first_line(): skip = False continue origl = l if args.interval: m = re.match(r"\s*([0-9.]{9,}|SUMMARY);(.*)", l) if m: interval = float(m.group(1)) if m.group(1) != "SUMMARY" else 0.0 l = m.group(2) if interval != prev_interval: linenum = 1 # skip the first because we can't tell when it started if prev_interval != 0.0 and prun.next_timestamp(): interval_dur = interval - prev_interval interval = prev_interval break if res: interval_dur = interval - prev_interval set_interval(env, interval_dur, prev_interval) if need_fallback: update_missing(res, rev, valstats, fallback) fallback = {} need_fallback = False yield 0, res, rev, prev_interval, valstats, env res = defaultdict(list) rev = defaultdict(list) valstats = defaultdict(list) prev_interval = interval start = interval elif not l[:1].isspace(): # these are likely bogus summary lines printed by v5.8 perf stat # just ignore skip = False continue if prun.skip_input(): continue if args.perf_output: args.perf_output.write(origl.rstrip() + "\n") n = l.split(";") # filter out the empty unit field added by 3.14 n = [x for x in n if x not in ('', 'Joules', 'ns')] # timestamp is already removed # -a --per-socket socket,numcpus,count,event,... # -a --per-core core,numcpus,count,event,... # -a -A cpu,count,event,... # count,event,... if is_event(n, 1): title, count, event, off = "", n[0], n[1], 2 elif is_event(n, 3): title, count, event, off = n[0], n[2], n[3], 4 elif is_event(n, 2): title, count, event, off = n[0], n[1], n[2], 3 else: if not FUZZYINPUT: warn("unparseable perf output\n%s" % origl.rstrip()) linenum += 1 skip = False continue # dummy event used as separator to avoid merging problems if event.startswith("emulation-faults"): linenum += 1 skip = False continue title = title.replace("CPU", "") # code later relies on stripping ku flags event = remove_qual(event) event = re.sub(r'\s+\[.*\]', '', event) # duplicated duration_time in perf ~6.5. was already added from the first. if event == "duration_time" and count == "<not counted>": linenum += 1 skip = False continue skip = False origevent = event runner, action, event = check_event(rlist, event, len(res[title]), title, prev_interval, origl, revnum, linenum, last_linenum) if runner is None: linenum += 1 continue if action == SKIP: l = origl skip = True if action == FUZZY: need_fallback = True multiplex = float('nan') event = event.rstrip() if re.match(r"\s*[0-9.]+", count): val = float(count) elif re.match(r"\s*<", count): account[event].errors[count.replace("<","").replace(">","")] += 1 multiplex = 0. val = 0 else: warn("unparseable perf count\n%s" % l.rstrip()) linenum += 1 continue # post fixes: # ,xxx% -> -rXXX stddev stddev = 0. if len(n) > off and n[off].endswith("%") and not skip: stddev = (float(n[off].replace("%", "").replace(",", ".")) / 100.) * val off += 1 # ,xxx,yyy -> multiplexing in newer perf if len(n) > off + 1 and not skip: multiplex = float(n[off + 1].replace(",", ".")) off += 2 st = ValStat(stddev=stddev, multiplex=multiplex) account[event].total += 1 def ignored_cpu(num): return num not in runner.cpu_list and (num not in cpu.cputocore or not any( [k in runner.cpu_list for k in cpu.coreids[cpu.cputocore[num]]])) def add(t): if runner.cpu_list and t.isdecimal() and ignored_cpu(int(t)): return if skip: res[t].append(0.0) elif action == FUZZY: res[t].append(float("nan")) else: res[t].append(val) fallback[(t, origevent)] = (val, st) rev[t].append(event) valstats[t].append(st) if args.perf_summary: # XXX add unit, enabled, num-cpus assert len(res[t]) == len(rev[t]) update_perf_summary(summary, resoff[t] + len(res[t]) - 1, t, val, event, "", multiplex) def dup_val(l): for j in l: if j in runner.cpu_list: add("%d" % j) def uncore_event(event): return re.match(r'power|uncore', event) # power/uncore events are only output once for every socket if ((uncore_event(event) or uncore_event(origevent)) and title.isdecimal() and (not ((args.core or args.cpu) and not args.single_thread))): cpunum = int(title) socket = cpu.cputosocket[cpunum] dup_val(cpu.sockettocpus[socket]) elif re.match(r'(S\d+-)?(D\d+-)?C\d+', title) and (smt_mode or args.no_aggr): m = re.match(r'(?:S(\d+)-)?(?:D(\d+)-)?C(\d+)', title) assert m is not None if m.group(2): # XXX warn_once("die topology not supported currently") socket, core = int(m.group(1)), int(m.group(3)) dup_val(cpu.coreids[(socket, core)]) # duration time is only output once, except with --cpu/-C (???) # except perf 6.2+ outputs it with -A on all cpus, but not counting except the first elif ((event.startswith("duration_time") or origevent.startswith("duration_time")) and title.isdecimal() and not args.cpu and not args.core): dup_val(runner.cpu_list) else: add(title) if skip: continue linenum += 1 if args.raw or args.valcsv: dump_raw(out.valcsv, interval if args.interval else "", title, event, event_rmap(event, rlist), val if val or not re.match(r"\s*<", count) else count, len(res[title]) - 1, stddev, multiplex, rlist) inf.close() if not args.import_ and not args.interval: set_interval(env, time.time() - start, start) elif args.interval: set_interval(env, interval_dur if interval_dur else args.interval/1000., interval if interval else float('NaN')) else: warn_no_assert("cannot determine time duration. Per second metrics may be wrong. Use -Ixxx.") set_interval(env, 0, 0) ret = prun.wait() print_account(account) if need_fallback: update_missing(res, rev, valstats, fallback) yield ret, res, rev, interval, valstats, env run_l1_parallel = False # disabled for now until we can fix the perf scheduler def adjust_ev(ev, level): # use the programmable slots for non L1 so that level 1 # can (mostly) run in parallel with other groups. # this also helps for old or non ICL kernels # XXX this creates groups without slots leader if isinstance(ev, str) and ev.startswith("TOPDOWN.SLOTS") and ((run_l1_parallel and level != 1) or not ectx.slots_available): ev = ev.replace("TOPDOWN.SLOTS", "TOPDOWN.SLOTS_P") return ev def ev_collect(ev, level, obj): if isinstance(ev, types.LambdaType): return ev(lambda ev, level: ev_collect(ev, level, obj), level) if ev == "mux": return DummyArith() if ev.startswith("interval-"): if not feat.supports_duration_time: return DummyArith() ev = "duration_time" ev = adjust_ev(ev, level) if level == 999: if ret_latency is None: setup_retlatency(args) return DummyArith() key = (ev, level, obj.name) if key not in obj.evlevels: if ev.startswith(("TOPDOWN.SLOTS", "PERF_METRICS.")): ind = [x[1] == level for x in obj.evlevels] ins = ind.index(True) if any(ind) else 0 obj.evlevels.insert(ins + (0 if ev.startswith("TOPDOWN.SLOTS") else 1), key) else: obj.evlevels.append(key) if safe_ref(obj, 'nogroup') or ev == "duration_time": ectx.outgroup_events.add(ev.lower()) return DummyArith() def canon_event(e): m = re.match(r"(.*?):(.*)", e) if m: e = m.group(1) return e.lower() def find_runner_by_pmu(pmu, runner_list): for r in runner_list: if r.pmu == pmu: return r return None def event_pmu(ev): m = re.match(r'(.*?)/', ev) if m: return m.group(1) return None def event_ectx(ev, runner_list): pmu = event_pmu(ev) if pmu and pmu.startswith("cpu"): r = find_runner_by_pmu(pmu, runner_list) if r: return r.ectx # this works for now because the atom model doesn't use any uncore events # may need to fix later once it does return ectx if ectx else runner_list[0].ectx def do_event_rmap(e, ectx_): n = ectx_.emap.getperf(e) if ectx_.emap.getevent(n, nocheck=event_nocheck()): return n if e in non_json_events: return e if e.startswith("uncore"): warn_no_assert("rmap: cannot find %s, using dummy" % e) else: warn("rmap: cannot find %s, using dummy" % e) return "dummy" def event_rmap(e, runner_list): ectx_ = event_ectx(e, runner_list) if e in ectx_.rmap_cache: return ectx_.rmap_cache[e] n = do_event_rmap(e, ectx_) ectx_.rmap_cache[e] = n return n cycles_aliases = frozenset(("cycles", "cpu_clk_unhalted.thread", "cpu_clk_unhalted.core", "cpu_core/event=0x3c,umask=0x0/", "cpu_atom/event=0x3c,umask=0x0/", "cpu/event=0x3c,umask=0x0/")) # compare events to handle name aliases def compare_event(aname, bname): # XXX this should be handled in ocperf if aname in cycles_aliases and bname in cycles_aliases: return True a = ectx.emap.getevent(aname, nocheck=event_nocheck()) if a is None: return False b = ectx.emap.getevent(bname, nocheck=event_nocheck()) if b is None: return False fields = ('val','event','cmask','edge','inv') return map_fields(a, fields) == map_fields(b, fields) def is_hybrid(): return ocperf.file_exists("/sys/devices/cpu/format/any") def lookup_res(res, rev, ev, obj, env, level, referenced, cpuoff, st, runner_list): """get measurement result, possibly wrapping in UVal""" if level == 999: return lookup_retlat(ev) ev = adjust_ev(ev, level) if isinstance(ev, str) and ev.startswith("interval") and feat.supports_duration_time: scale = { "interval-s": 1e9, "interval-ns": 1, "interval-ms": 1e6 }[ev] return lookup_res(res, rev, "duration_time", obj, env, level, referenced, cpuoff, st, runner_list)/scale if ev in env: return env[ev] if ev == "mux": return min([s.multiplex for s in st]) # # when the model passed in a lambda run the function for each logical cpu # (by resolving its EVs to only that CPU) # and then sum up. This is needed for the workarounds to make various # per thread counters at least as big as unhalted cycles. # # otherwise we always sum up. # if isinstance(ev, types.LambdaType): return sum([ev(lambda ev, level: lookup_res(res, rev, ev, obj, env, level, referenced, off, st, runner_list), level) for off in range(env['num_merged'])]) index = obj.res_map[(ev, level, obj.name)] referenced.add(index) #print((ev, level, obj.name), "->", index) if not args.fast: try: r = rev[index] except IndexError: warn_once_no_assert("Not enough lines in perf output for rev (%d vs %d for %s) at %s, event %s" % (index, len(rev), obj.name, env['interval'], ev)) return 0 rmap_ev = event_rmap(r, runner_list).lower() ev = ev.lower() assert (rmap_ev == canon_event(ev).replace("/k", "/") or compare_event(rmap_ev, ev) or rmap_ev == "dummy" or (rmap_ev.endswith("_any") and not is_hybrid())), "event rmap mismatch %s vs %s" % (rmap_ev, ev) try: vv = res[index] except IndexError: warn_once("Not enough lines in perf output for res (%d vs %d for %s) at %s" % (index, len(res), obj.name, env['interval'])) return 0.0 if isinstance(vv, tuple): if cpuoff == -1: vv = sum(vv) else: try: vv = vv[cpuoff] except IndexError: warn_once("Partial CPU thread data from perf for %s" % obj.name) return 0.0 if st[index].stddev or st[index].multiplex != 100.0: return UVal(name=ev, value=vv, stddev=st[index].stddev, mux=st[index].multiplex) return vv class BadEvent(Exception): def __init__(self, name): super(Exception, self).__init__() self.event = name # XXX check for errata def sample_event(e, emap): ev = emap.getevent(e, nocheck=event_nocheck()) if not ev: raise BadEvent(e) postfix = args.ring_filter if postfix: postfix = ":" + postfix return ev.name + postfix def sample_desc(s, emap): try: return " ".join([sample_event(x, emap) for x in s]) except BadEvent as e: warn_once_no_assert("Unknown sample event %s" % (e.event)) return "" def get_level(x): return x[1] def get_levels(evlev): return [x[1] for x in evlev] def get_names(evlev): return [x[0] for x in evlev] def full_name(obj): name = obj.name while 'parent' in obj.__dict__ and obj.parent: obj = obj.parent name = obj.name + "." + name return name def package_node(obj): return safe_ref(obj, 'domain') in ("Package", "SystemMetric") def not_package_node(obj): return not package_node(obj) def core_node(obj): return safe_ref(obj, 'domain') in ectx.core_domains def thread_node(obj): if package_node(obj): return False if core_node(obj): return False return True def any_node(obj): return True def obj_domain(obj): return obj.domain.replace("Estimated", "est").replace("Calculated", "calc") def metric_unit(obj): if has(obj, 'unit'): return obj.unit if has(obj, 'domain'): return obj_domain(obj).replace("SystemMetric", "SysMetric") return "Metric" def obj_desc(obj, sep="\n\t"): desc = obj.desc[1:].replace("\n", sep) # by default limit to first sentence if not args.long_desc and "." in desc: desc = desc[:desc.find(".") + 1] + ".." return desc def get_mg(obj): return ref_or(obj, 'metricgroup', frozenset([])) # only check direct children, the rest are handled recursively def children_over(l, obj): n = [o.thresh for o in l if 'parent' in o.__dict__ and o.parent == obj] return any(n) def bottleneck_related(obj, bn): if obj == bn: return True if (get_mg(bn) & get_mg(obj)) - tma_mgroups: return True return False def obj_desc_runtime(obj, rest, bn): # hide description if children are also printed if children_over(rest, obj) or (not args.desc and not bottleneck_related(obj, bn)): desc = "" else: desc = obj_desc(obj) if 'htoff' in obj.__dict__ and obj.htoff and obj.thresh and cpu.ht and not args.single_thread: desc += """ Warning: Hyper Threading may lead to incorrect measurements for this node. Suggest to re-measure with HT off (run cputop.py "thread == 1" offline | sh).""" return desc def node_filter(obj, default, sibmatch, mgroups): if args.nodes: fname = full_name(obj) name = obj.name def _match(m): return (fnmatch(name, m) or fnmatch(fname, m) or fnmatch(fname, "*." + m)) def match(m, checklevel=True): if m.endswith("^"): m = m[:-1] r = re.match("(.*)/([0-9]+)", m) if r: level = int(r.group(2)) if checklevel and obj.level > level: return False m = r.group(1) return _match(m) def has_siblings(j, obj): return j.endswith("^") and 'sibling' in obj.__dict__ and obj.sibling def has_mg(j, obj): return j.endswith("^") and get_mg(obj) nodes = args.nodes if nodes[0] == '!': default = False nodes = nodes[1:] for j in nodes.split(","): j = j.strip() if j == "": continue i = 0 if j[0] == '^' or j[0] == '-': if match(j[1:]): return False continue elif j[0] == '+': i += 1 if match(j[i:], True): if has_siblings(j, obj): sibmatch |= set(obj.sibling) if has_mg(j, obj): mgroups |= obj.metricgroup obj.forced = True return True if has_siblings(j, obj): for sib in obj.sibling: fname = full_name(sib) name = sib.name if match(j[i:], False): sibmatch.add(obj) return True return default def _find_bn(bn, level): siblings = sorted([x for x in bn if x.level - 1 == level], key=lambda x: x.val, reverse=True) if len(siblings) == 0: return None # remove overlap nodes siblings = [x for x in siblings if not (has(x, 'overlap') and x.overlap)] # ambigious if level > 0 and len(siblings) > 1 and siblings[0].val - siblings[1].val <= SIB_THRESH: return None n = _find_bn([x for x in bn if full_name(x).startswith(full_name(siblings[0]))], level + 1) if n is None: return siblings[0] return n def find_bn(olist, match): if args.force_bn: bn = sorted([o for o in olist if o.name in args.force_bn], key=lambda x: x.level, reverse=True) if bn: return bn[0] bn = [o for o in olist if match(o) and not o.metric and o.thresh] if not bn: return None return _find_bn(bn, 0) pmu_does_not_exist = set() # XXX check if PMU can be accessed from current user def missing_pmu(e): if event_nocheck(): return False m = re.match(r"([a-z0-9_]+)/", e) if m: pmu = m.group(1) if pmu in pmu_does_not_exist: return True if not os.path.isdir("/sys/devices/%s" % pmu): pmu_does_not_exist.add(pmu) return True return False def query_errata(obj, errata_events, errata_nodes, errata_names): errata = [errata_events[x] for x in obj.evlist if x in errata_events] if any(errata): errata_nodes.add(obj) errata_names |= set(errata) def olist_by_metricgroup(l, mg): if args.no_sort: return l valid = set(l) visited = set() ml = [] for obj in l: if obj in visited: continue def add(obj): if obj not in visited: ml.append(obj) visited.add(obj) if has(obj, 'metricgroup'): for g in sorted(obj.metricgroup): for j in mg[g]: if j in valid: add(j) add(obj) return ml def node_unit(obj): return (" " + obj_domain(obj)) if has(obj, 'domain') else "" def node_below(obj): return not obj.thresh class Summary(object): """Accumulate counts for summary.""" def __init__(self): self.res = defaultdict(list) self.rev = defaultdict(list) self.env = Counter() # type: typing.Counter[str] self.valstats = defaultdict(list) self.summary_perf = OrderedDict() def add(self, res, rev, valstats, env): for j in sorted(res.keys()): for ind, val in enumerate(res[j]): if ind < len(self.res[j]): self.res[j][ind] += val self.valstats[j][ind] = combine_valstat([self.valstats[j][ind], valstats[j][ind]]) else: self.res[j].append(val) self.valstats[j].append(valstats[j][ind]) if len(rev.keys()) == 1: append_dict(self.rev, rev) else: self.rev.update(rev) for j in env.keys(): self.env[j] += env[j] def parse_metric_group(l, mg): if l is None: return [], [] add, rem = [], [] for n in l.split(","): if n[0:1] == '-' or n[0:1] == '^': if n[1:] not in mg: print("metric group", n[1:], "not found", file=sys.stderr) continue rem += [x.name for x in mg[n[1:]]] continue if n[0:1] == '+': n = n[1:] if n not in mg: print("metric group", n, "not found", file=sys.stderr) continue add += [x.name for x in mg[n]] return add, rem def obj_area(obj): return obj.area if has(obj, 'area') and not args.no_area else None def get_parents(obj): def get_par(obj): return obj.parent if 'parent' in obj.__dict__ else None p = get_par(obj) l = [] while p: l.append(p) p = get_par(p) return l def quote(s): if " " in s: return '"' + s + '"' return s class Group(object): def __init__(self, evnum, objl, num, outgroup=False): self.evnum = evnum self.base = -1 self.objl = set(objl) self.outgroup = outgroup self.num = num class GroupCmp(object): def __init__(self, v): self.v = v def __lt__(self, g): return self.v < g.base # Control whether even unrelated groups can be merged any_merge = True # Interleave uncore events between CPU groups distribute_uncore = False def grab_group(l): if needed_counters(l) <= ectx.counters: return len(l) n = 1 while needed_counters(l[:n]) < ectx.counters and n < len(l): n += 1 if needed_counters(l[:n]) > ectx.counters and n > 0: n -= 1 assert needed_counters(l[:n]) <= ectx.counters return n def update_group_map(evnum, obj, group): for lev in obj.evlevels: r = raw_event(lev[0]) # can happen during splitting # the update of the other level will fix it if r in evnum and lev not in obj.group_map: obj.group_map[lev] = (group, evnum.index(r)) def do_distribute_uncore(evgroups): cg = [g for g in evgroups if not g.outgroup] og = [g for g in evgroups if g.outgroup] return [x for x in chain(*zip_longest(cg, og)) if x is not None] def gen_res_map(solist): for obj in solist: for k in obj.group_map.keys(): gr = obj.group_map[k] obj.res_map[k] = gr[0].base + gr[1] def print_group(g): evkeys = [k for o in g.objl for k in o.group_map.keys() if o.group_map[k][0] == g] objnames = {("%s" % quote(x[2])) + ("[%d]" % x[1] if x[1] else "") for x in evkeys} if len(objnames) == 0: return evnames = {mark_fixed(x[0]) for x in evkeys} pwrap(" ".join(objnames) + ":", 78) pwrap(" ".join(evnames).lower() + (" [%d counters]" % needed_counters(g.evnum)) + (" [%d]" % g.base if args.debug else ""), 75, " ") def match_patlist(l, s): for x in l.split(","): if fnmatch(s, x): return True return False class Scheduler(object): """Schedule events into groups.""" def __init__(self): self.evnum = [] # flat global list self.evgroups = [] # of Group self.og_groups = {} # list of groups that still have generic counters, for faster # duplicate checks self.evgroups_nf = [] self.nextgnum = 0 self.event_to_group = {} # should avoid adding those in the first place instead def dummy_unreferenced(self, olist): refs = defaultdict(set) for o in olist: for g, ind in o.group_map.values(): refs[g].add(ind) for g in self.evgroups: ref = refs[g] if len(ref) < len(g.evnum): for i in range(len(g.evnum)): if i not in ref: debug_print("unreferenced %s [%d] %s" % (g.evnum[i], i, " ".join([o.name for o in g.objl]))) g.evnum[i] = "dummy" def split_groups(self, obj, evlev): for lev, evl in groupby(sorted(evlev, key=get_level), get_level): evlev = list(evl) evnum = [raw_event(x[0]) for x in evlev] while evlev: n = grab_group(evnum) self.add(obj, evnum[:n], None) evlev = evlev[n:] evnum = evnum[n:] # may modify evnum def add_duplicate(self, evnum, obj): evset = set(evnum) num_gen = num_generic_counters(evset) full = set() if ((has(obj, 'area') and match_patlist(DEDUP_AREA, obj.area)) or match_patlist(DEDUP_NODE, obj.name)): # reuse any previous event independent of group subsets # for bottleneck nodes which are too large for the usual # heuristics duped = [] for ind, e in enumerate(evnum): if ismetric(e) or is_slots(e): continue if e in self.event_to_group: g = self.event_to_group[e] debug_print("dedup %s %s to %s" % (obj.name, e, " ".join([x.name for x in g.objl]))) g.objl.add(obj) update_group_map(g.evnum, obj, g) duped.append(ind) # need to remove in place so that caller sees it # remove backwards so that indexes stay valid for ind in reversed(duped): del evnum[ind] if len(evnum) == 0: debug_print("%s fully deduped" % obj.name) return True for g in reversed(self.evgroups_nf if num_gen else self.evgroups): if g.outgroup: continue # # In principle we should only merge if there is any overlap, # otherwise completely unrelated nodes get merged. But the perf # scheduler isn't very good at handling smaller groups, and # with eventual exclusive use we would like as big groups as # possible. Still keep it as a --tune option to play around. if ((any_merge or not evset.isdisjoint(g.evnum)) and needed_counters(cat_unique(g.evnum, evnum)) <= ectx.counters): obj_debug_print(obj, "add_duplicate %s in %s obj %s to group %d" % ( " ".join(evnum), " ".join(g.evnum), obj.name, g.num)) for k in evnum: if k not in g.evnum: g.evnum.append(k) if k not in self.event_to_group: self.event_to_group[k] = g g.objl.add(obj) update_group_map(g.evnum, obj, g) return True # memorize already full groups elif num_generic_counters(set(g.evnum)) >= ectx.counters: full.add(g) if full: self.evgroups_nf = [g for g in self.evgroups_nf if g not in full] return False def add(self, obj, evnum, evlev): # does not fit into a group. if needed_counters(evnum) > ectx.counters: self.split_groups(obj, evlev) return evnum = dedup(evnum) if not self.add_duplicate(evnum, obj): g = Group(evnum, [obj], self.nextgnum) obj_debug_print(obj, "add %s to group %d" % (evnum, g.num)) for k in evnum: if k not in self.event_to_group: self.event_to_group[k] = g self.nextgnum += 1 self.evgroups.append(g) self.evgroups_nf.append(g) update_group_map(evnum, obj, g) def add_outgroup(self, obj, evnum): obj_debug_print(obj, "add_outgroup %s" % evnum) for ev in evnum: if ev in self.og_groups: g = self.og_groups[ev] g.objl.add(obj) else: g = Group([ev], [obj], self.nextgnum, True) self.nextgnum += 1 self.og_groups[ev] = g self.evgroups.append(g) self.evgroups_nf.append(g) update_group_map([ev], obj, g) def allocate_bases(self): base = 0 for g in self.evgroups: g.base = base self.evnum += g.evnum base += len(g.evnum) def print_group_summary(self, olist): num_groups = len([g for g in self.evgroups if not g.outgroup]) print("%d cpu groups, %d outgroups with %d events total (%d unique) for %d objects, %d dummies" % ( num_groups, len(self.evgroups) - num_groups, len(self.evnum), len(set(self.evnum)), len(olist), self.evnum.count("dummy")), file=sys.stderr) # fit events into available counters def schedule(self, olist): # sort objects by level and inside each level by num-counters solist = sorted(olist, key=lambda x: (x.level, x.nc)) # try to fit each objects events into groups # that fit into the available CPU counters for obj in solist: obj_debug_print(obj, "schedule %s " % obj.name) evnum = obj.evnum evlevels = obj.evlevels oe = [e in ectx.outgroup_events for e in obj.evnum] if any(oe): # add events outside group separately og_evnum = list(compress(obj.evnum, oe)) self.add_outgroup(obj, og_evnum) if all(oe): continue # keep other events ie = not_list(oe) evlevels = list(compress(obj.evlevels, ie)) evnum = list(compress(obj.evnum, ie)) self.add(obj, evnum, evlevels) if args.no_multiplex or distribute_uncore: self.evgroups = do_distribute_uncore(self.evgroups) if not KEEP_UNREF: self.dummy_unreferenced(olist) self.allocate_bases() if args.print_group: for g in self.evgroups: print_group(g) gen_res_map(olist) if args.print_group: self.print_group_summary(olist) def should_print_obj(obj, match, thresh_mg, bn): assert not isinstance(obj.val, DummyArith) if obj.val is None: return False if obj.thresh or obj.metric or args.verbose: if not match(obj): return False elif args.only_bottleneck and obj != bn: if args.node_metrics and 'group_select' in obj.__dict__ and get_mg(obj) & get_mg(bn): return True if args.bottlenecks and 'area' in obj.__dict__ and obj.area == "Bottleneck": return True # XXX handle more explicit options like metrics? if 'forced' in obj.__dict__ and obj.forced: return True elif obj.metric: if args.node_metrics and 'group_select' in obj.__dict__ and not (get_mg(obj) & thresh_mg): return False if args.verbose or (obj.metric and obj.thresh and obj.val != 0.0): return True elif check_ratio(obj.val): # somewhat redundant thresh_mg |= get_mg(obj) - tma_mgroups return True return False def get_uval(ob): u = ob.val if isinstance(ob.val, UVal) else UVal(ob.name, ob.val) u.name = ob.name return u # pre compute column lengths def compute_column_lengths(olist, out): for obj in olist: out.set_hdr(full_name(obj), obj_area(obj)) if obj.metric: out.set_unit(metric_unit(obj)) else: out.set_unit(node_unit(obj)) out.set_below(node_below(obj)) class Printer(object): """Print measurements while accumulating some metadata.""" def __init__(self, metricgroups): self.sample_obj = set() self.bottlenecks = set() self.numprint = 0 self.metricgroups = metricgroups def print_res(self, olist, out, timestamp, title, match, bn, idlemark=False): if bn: self.bottlenecks.add(bn) out.logf.flush() # determine all objects to print thresh_mg = set() # type: Set[str] olist = [o for o in olist if should_print_obj(o, match, thresh_mg, bn)] # sort by metric group olist = olist_by_metricgroup(olist, self.metricgroups) compute_column_lengths(olist, out) # step 3: print for i, obj in enumerate(olist): val = get_uval(obj) if has(obj, 'maxval') and obj.maxval is not None and obj.maxval != 0: maxval = UVal(obj.name, obj.maxval) val = min(val, maxval) desc = obj_desc_runtime(obj, olist[i + 1:], bn) if obj.metric: out.metric(obj_area(obj), obj.name, val, timestamp, desc, title, metric_unit(obj), idlemark) else: out.ratio(obj_area(obj), full_name(obj), val, timestamp, "%" + node_unit(obj), desc, title, sample_desc(obj.sample, self.emap) if has(obj, 'sample') else None, "<==" if obj == bn else "", node_below(obj), idlemark) if obj.thresh or args.verbose: self.sample_obj.add(obj) self.numprint += 1 def init_emap(self, emap): self.emap = emap # check nodes argument for typos def check_nodes(runner_list, nodesarg): def opt_obj_name(s): if s[:1] in ('+', '^', '-'): s = s[1:] if "/" in s: s = s[:s.index("/")] if s.endswith("^"): s = s[:-1] return s if nodesarg[:1] == "!": nodesarg = nodesarg[1:] options = [opt_obj_name(s) for s in nodesarg.split(",")] def valid_node(s): if s == "": return True for r in runner_list: if s in r.odict: return True for k in r.olist: if fnmatch(k.name, s) or fnmatch(full_name(k), s): return True return False valid = list(map(valid_node, options)) if not all(valid): sys.exit("Unknown node(s) in --nodes: " + " ".join([o for o, v in zip(options, valid) if not v])) class Runner(object): """Handle measurements of event groups. Map events to groups.""" def reset(self): self.stat = ComputeStat(args.quiet) self.olist = [] self.idle_keys = set() self.sched = Scheduler() self.printer = Printer(self.metricgroups) def __init__(self, max_level, idle_threshold, kernel_version, pmu=None): # always needs to be filtered by olist: self.metricgroups = defaultdict(list) self.reset() self.odict = {} self.max_level = max_level self.max_node_level = 0 self.idle_threshold = idle_threshold self.ectx = EventContext(pmu) self.pmu = pmu self.full_olist = [] self.cpu_list = [] # type: List[int] self.kernel_version = kernel_version def set_ectx(self): global ectx ectx = self.ectx def clear_ectx(self): # confuses the type checker #global ectx #ectx = None pass def do_run(self, obj): obj.res = None obj.res_map = {} obj.group_map = {} self.olist.append(obj) self.full_olist.append(obj) self.odict[obj.name] = obj if has(obj, 'metricgroup'): for j in sorted(obj.metricgroup): self.metricgroups[j].append(obj) self.max_node_level = max(self.max_node_level, obj.level) # remove unwanted nodes after their parent relationship has been set up def filter_nodes(self): add_met, remove_met = parse_metric_group(args.metric_group, self.metricgroups) add_obj = {self.odict[x] for x in add_met} parents = [get_parents(x) for x in add_obj] if add_obj: for o in self.olist: if safe_ref(o, 'sibling') is None: continue m = set(o.sibling) & add_obj for s in m: parents.append(s) parents += get_parents(s) self.sibmatch = set() # type: Set[Any] mgroups = set() # type: Set[str] def want_node(obj, mgroups, tma_mgroups): area = safe_ref(obj, 'area') if args.no_uncore and area == "Info.System": return False if args.areas and area and any([fnmatch(area, p) for p in args.areas.split(",")]): return True if args.bottlenecks and area == "Info.Bottleneck": return True want = ((obj.metric and args.metrics) or (('force_metric' in obj.__dict__) and obj.force_metric) or obj.name in add_met or obj in parents) and obj.name not in remove_met if not obj.metric and obj.level <= self.max_level: want = True want = node_filter(obj, want, self.sibmatch, mgroups) mg = get_mg(obj) tma_mgroups |= set([x for x in mg if x.startswith("Tma")]) if args.node_metrics and want and not obj.metric: mgroups |= set(mg) - tma_mgroups return want for x in self.olist: if 'forced' in x.__dict__: x.forced = False # this updates sibmatch fmatch = [want_node(x, mgroups, tma_mgroups) for x in self.olist] def select_node(obj): if obj in self.sibmatch: return True if get_mg(obj) & mgroups: obj.group_select = True return True return False # now keep what is both in fmatch and sibmatch and mgroups # assume that mgroups matches do not need propagation self.olist = [obj for obj, fil in zip(self.olist, fmatch) if fil or select_node(obj)] def setup_children(self): for obj in self.olist: if not obj.metric and 'parent' in obj.__dict__ and obj.parent: obj.parent.children.append(obj) def reset_thresh(self): for obj in self.olist: obj.thresh = Undef def run(self, obj): obj.thresh = False obj.metric = False obj.children = [] self.do_run(obj) def metric(self, obj): obj.thresh = Undef obj.metric = True obj.level = 0 obj.sibling = None self.do_run(obj) def force_metric(self, obj): obj.force_metric = True self.metric(obj) # collect the events by pre-computing the equation def collect(self): self.set_ectx() bad_nodes = set() bad_events = set() unsup_nodes = set() errata_nodes = set() # type: Set[Any] errata_warn_nodes = set() # type: Set[Any] errata_names = set() # type: Set[str] errata_warn_names = set() # type: Set[str] min_kernel = [] # type: List[int] for obj in self.olist: obj.evlevels = [] obj.compute(lambda ev, level: ev_collect(ev, level, obj)) obj.val = None obj.evlist = [x[0] for x in obj.evlevels] obj.evnum = raw_events(obj.evlist, initialize=True) obj.nc = needed_counters(dedup(obj.evnum)) # work arounds for lots of different problems unsup = [x for x in obj.evlist if unsup_event(x, unsup_events, self.kernel_version, min_kernel)] if any(unsup): bad_nodes.add(obj) bad_events |= set(unsup) unsup = [x for x in obj.evlist if missing_pmu(x)] if any(unsup): unsup_nodes.add(obj) query_errata(obj, ectx.errata_events, errata_nodes, errata_names) query_errata(obj, ectx.errata_warn_events, errata_warn_nodes, errata_warn_names) if bad_nodes: if args.force_events: pwrap_not_quiet("warning: Using --force-events. Nodes: " + " ".join([x.name for x in bad_nodes]) + " may be unreliable") else: if not args.quiet: pwrap("warning: removing " + " ".join([x.name for x in bad_nodes]) + " due to unsupported events in kernel: " + " ".join(sorted(bad_events)), 80, "") if min_kernel: print("Fixed in kernel %d" % (sorted(min_kernel, key=kv_to_key, reverse=True)[0]), file=sys.stderr) print("Use --force-events to override (may result in wrong measurements)", file=sys.stderr) self.olist = [x for x in self.olist if x not in bad_nodes] if unsup_nodes: pwrap_not_quiet("Nodes " + " ".join(x.name for x in unsup_nodes) + " has unsupported PMUs") self.olist = [x for x in self.olist if x not in unsup_nodes] if errata_nodes and not args.ignore_errata: pwrap_not_quiet("Nodes " + " ".join(x.name for x in errata_nodes) + " have errata " + " ".join(errata_names) + " and were disabled. " + "Override with --ignore-errata") self.olist = [x for x in self.olist if x in errata_nodes] if errata_warn_nodes and not args.ignore_errata: pwrap_not_quiet("Nodes " + " ".join(x.name for x in errata_warn_nodes) + " have errata " + " ".join(errata_warn_names)) self.clear_ectx() def propagate_siblings(self): changed = [0] def propagate(k, changed, srco): if args.debug: print("propagate", srco.name, "->", k.name) if not k.thresh: k.thresh = True changed[0] += 1 for obj in self.olist: if obj in self.sibmatch: propagate(obj, changed, obj) if obj.thresh and obj.sibling: if isinstance(obj.sibling, (list, tuple)): for k in obj.sibling: propagate(k, changed, obj) else: propagate(obj.sibling, changed, obj) return changed[0] def compute(self, res, rev, valstats, env, match, stat, runner_list): self.set_ectx() changed = 0 # step 1: compute for obj in self.olist: obj.errcount = 0 if not match(obj): continue ref = set() # type: Set[int] oldthresh = obj.thresh if 'parent' in obj.__dict__ and obj.parent and obj.parent not in self.olist: obj.parent.thresh = True obj.compute(lambda e, level: lookup_res(res, rev, e, obj, env, level, ref, -1, valstats, runner_list)) # compatibility for models that don't set thresh for metrics if isinstance(obj.thresh, UVal) and obj.name == "Undef": obj.thresh = True if args.force_bn and obj.name in args.force_bn: obj.thresh = True if obj.thresh != oldthresh and oldthresh != Undef: changed += 1 if stat: stat.referenced |= ref if not obj.res_map and not all([x in env for x in obj.evnum]) and not args.quiet: print("%s not measured" % (obj.__class__.__name__,), file=sys.stderr) if not obj.metric and not check_ratio(obj.val): obj.thresh = False if stat: stat.mismeasured.add(obj.name) if stat and has(obj, 'errcount') and obj.errcount > 0: if obj.name not in stat.errors: stat.errcount += obj.errcount stat.errors.add(obj.name) stat.referenced |= set(obj.res_map.values()) # step 2: propagate siblings changed += self.propagate_siblings() self.clear_ectx() return changed def list_metric_groups(self): if not args.quiet: print("MetricGroups:") mg = sorted(self.metricgroups.keys()) if args.quiet or args.csv: pre = "" else: pre = "\t" for j in mg: print(pre + j) def list_nodes(self, title, filt, rest): def match(rest, n, fn): return not rest or any([n.startswith(x) or fn.startswith(x) if not x.endswith("^") else n == x[:-1] or fn == x[:-1] for x in rest]) if title and not args.quiet: print("%s:" % title) for obj in self.olist: fn = full_name(obj) if args.csv: pre, sep, dsep = "", "\n", "" elif args.quiet: pre, sep, dsep = "", "\n", "\n" else: pre, sep, dsep = "\t", "\n", "\n\t" if filt(obj) and match(rest, obj.name, fn): print(fn, end=sep) if not args.no_desc: print(pre + obj_desc(obj, sep=dsep)) def filter_per_core(self, single_thread, rest): if ("Slots" not in self.ectx.core_domains and cpu.ht and not single_thread and has_core_node(self)): if not feat.supports_percore: self.olist = filternot(lambda obj: safe_ref(obj, 'domain') in self.ectx.core_domains, self.olist) else: rest = add_args(rest, "--percore-show-thread") return rest def runner_restart(runner, offset): emap = runner.printer.emap runner.reset() runner.printer.emap = emap runner.olist = list(runner.full_olist) for o in runner.olist: o.group_map = {} o.res_map = {} runner.filter_nodes() runner.collect() runner.set_ectx() runner.sched.schedule(runner.olist) runner.clear_ectx() runner.sched.offset = offset offset += len(runner.sched.evnum) return offset def runner_init(runner): runner.setup_children() runner.filter_nodes() runner.collect() def supports_pebs(): if feat.has_max_precise: return feat.max_precise > 0 return not cpu.hypervisor def remove_pp(s): if s.endswith(":pp"): return s[:-3] return s def clean_event(e): return remove_pp(e).replace(".", "_").replace(":", "_").replace('=','') def do_sample(sample_obj, rest, count, ret, kernel_version): samples = [("cycles:pp", "Precise cycles", )] for obj in sample_obj: for s in obj.sample: samples.append((s, obj.name)) # first dedup samples = dedup(samples) # now merge objects with the same sample event into one def sample_event(x): return x[0] samples = sorted(samples, key=sample_event) samples = [(k, "_".join([x[1] for x in g])) for k, g in groupby(samples, key=sample_event)] # find unsupported events nsamp = [x for x in samples if not unsup_event(x[0], unsup_events, kernel_version)] nsamp = [(remove_pp(x[0]), x[1]) if unsup_event(x[0], unsup_pebs, kernel_version) else x for x in nsamp] if nsamp != samples: missing = [x[0] for x in set(samples) - set(nsamp)] warn("Update kernel to handle sample events:" + "\n".join(missing)) def force_pebs(ev): return ev in ectx.require_pebs_events no_pebs = not supports_pebs() if no_pebs: for j in nsamp: # initialize ectx.require_pebs_events raw_event(j[0], initialize=True) nnopebs = {x[0] for x in nsamp if force_pebs(x[0])} if nnopebs and not args.quiet: for o in nnopebs: warn_no_assert("sample event %s not (currently) supported in virtualization" % o[0]) nsamp = [x for x in nsamp if x[0] not in nnopebs] sl = [raw_event(s[0], s[1] + "_" + clean_event(s[0]), period=True) for s in nsamp] sl = add_filter(sl) sample = ",".join([x for x in sl if x]) if no_pebs: sample = re.sub(r'/p+', '/', sample) sample = re.sub(r':p+', '', sample) if not args.quiet: print("Sampling:") extra_args = args.sample_args.replace("+", "-").split() perf_data = args.sample_basename if count is not None: perf_data += ".%d" % count sperf = ([feat.perf, "record"] + extra_args + ["-e", sample, "-o", perf_data] + [x for x in rest if x not in ("-A", "--percore-show-thread")]) cmd = " ".join(sperf) if not (args.run_sample and args.quiet): print(cmd) if args.run_sample and ret == 0: ret = os.system(cmd) if ret: print("Sampling failed") sys.exit(1) if not args.quiet: print("Run `" + feat.perf + " report%s%s' to show the sampling results" % ( (" -i %s" % perf_data) if perf_data != "perf.data" else "", " --no-branch-history" if "-b" in extra_args else "")) def suggest_bottlenecks(runner): def gen_bn(o): if o.children: return "+%s*/%d" % (o.name, o.level + BOTTLENECK_LEVEL_INC) if 'sibling' in o.__dict__ and o.sibling: return "+%s^" % full_name(o) return None printer = runner.printer children = [gen_bn(o) for o in printer.bottlenecks] measured = set([x.name for x in runner.olist]) children = [x for x in children if x and x not in measured] parents = [] for b in printer.bottlenecks: parents += ["+" + full_name(o) for o in get_parents(b) if o.name not in measured] if args.nodes: children = [x for x in children if x[1:-1] not in args.nodes] parents = [x for x in parents if x[1:] not in args.nodes] if children: mux = ",+MUX" if not (args.metrics or args.all) and (args.nodes is None or "MUX" not in args.nodes) else "" if not args.quiet: print("Add%s --nodes '!%s%s' for breakdown." % ( "ing" if args.drilldown else "", ",".join(children + parents), mux)) if args.drilldown: if runner.pmu != "cpu" and not args.quiet: print("Please make sure workload does not move between core types for drilldown", file=sys.stderr) if args.nodes: args.nodes += "," else: args.nodes = "" if args.nodes == "" or args.nodes[0] != '!': args.nodes = "!" + args.nodes args.nodes += ",".join(children) + mux return True return False def suggest_desc(runner): def nummatch(n): return sum([x.name.startswith(n) for x in runner.olist]) printer = runner.printer print("Run toplev --describe %s to get more information on bottleneck%s%s" % ( " ".join([full_name(x) + "^" if nummatch(x.name) > 1 else x.name + "^" for x in printer.bottlenecks]), "s" if len(printer.bottlenecks) > 1 else "", (" for " + runner.pmu.replace("cpu_", "")) if runner.pmu and runner.pmu != "cpu" else ""), file=sys.stderr) if not args.run_sample: print_once("Add --run-sample to find locations") def sysctl(name): try: with open("/proc/sys/" + name.replace(".","/"), "r") as f: val = int(f.readline()) except IOError: return 0 return val def update_cpu(args, cpu): # check nmi watchdog # XXX need to get this state from CSV import if sysctl("kernel.nmi_watchdog") != 0 or os.getenv("FORCE_NMI_WATCHDOG"): # XXX should probe if nmi watchdog runs on fixed or generic counter for j in cpu.counters.keys(): cpu.counters[j] -= 1 # FIXME if not args.quiet and not args.import_: print("Consider disabling nmi watchdog to minimize multiplexing", file=sys.stderr) print("(echo 0 | sudo tee /proc/sys/kernel/nmi_watchdog or\n echo kernel.nmi_watchdog=0 >> /etc/sysctl.conf ; sysctl -p as root)", file=sys.stderr) for j in cpu.counters.keys(): cpu.counters[j] -= args.reserved_counters if cpu.cpu is None: sys.exit("Unsupported CPU model %s %d" % (cpu.vendor, cpu.model,)) def get_kernel(): kv = os.getenv("KERNEL_VERSION") if not kv: kv = platform.release() return kv_to_key(list(map(int, kv.split(".")[:2]))) def check_exclusive(args, kernel_version): if args.exclusive: if kernel_version < 510: sys.exit("--exclusive needs kernel 5.10+") global metrics_own_group metrics_own_group = False global run_l1_parallel run_l1_parallel = False def ht_warning(): if cpu.ht and not args.quiet: print("WARNING: HT enabled", file=sys.stderr) print("Measuring multiple processes/threads on the same core may is not reliable.", file=sys.stderr) def setup_metrics(model, pmu): fmenv = os.getenv("FORCEMETRICS") ectx.force_metrics = fmenv is not None if ectx.force_metrics: try: ectx.metrics_override = True if int(fmenv if fmenv else "0") else False except ValueError: ectx.metrics_override = False if ectx.force_metrics: model.topdown_use_fixed = ectx.metrics_override else: model.topdown_use_fixed = os.path.exists("/sys/devices/%s/events/topdown-fe-bound" % pmu) if args.no_group: model.topdown_use_fixed = False ectx.core_domains = ectx.core_domains - set(["Slots"]) if ectx.force_metrics: ectx.slots_available = ectx.metrics_override else: ectx.slots_available = os.path.exists("/sys/devices/%s/events/slots" % pmu) def parse_cpu_list(s): l = [] for j in s.split(","): m = re.match(r'(\d+)(-\d+)?', j) if m is None: continue if m.group(2): for k in range(int(m.group(1)), int(m.group(2)[1:])+1): l.append(k) else: l.append(int(m.group(1))) return l def read_cpus(base): with open(base + "/cpus") as cpus: return parse_cpu_list(cpus.readline()) return [] def use_cpu(cpu): if args.core: return display_core(cpu, True) if args.cpu: return cpu in parse_cpu_list(args.cpu) return True def get_cpu_list(fn): return [k for k in read_cpus(fn) if use_cpu(k)] def init_runner_list(kernel_version): idle_threshold = init_idle_threshold(args) runner_list = [] hybrid_pmus = [] hybrid_pmus = glob.glob("/sys/devices/cpu_*") if args.force_cpu and args.force_cpu not in hybrid_cpus: hybrid_pmus = hybrid_pmus[:1] # real hybrid if hybrid_pmus and cpu.cpu in hybrid_cpus: for j in hybrid_pmus: pmuname = os.path.basename(j).replace("cpu_", "") if args.cputype and pmuname != args.cputype: continue cpu_list = get_cpu_list(j) if len(cpu_list) == 0: continue r = Runner(args.level, idle_threshold, kernel_version, pmu=os.path.basename(j)) runner_list.append(r) r.cpu_list = cpu_list # hybrid, but faking non hybrid cpu elif hybrid_pmus: runner_list = [Runner(args.level, idle_threshold, kernel_version, pmu= "cpu_atom" if cpu.cpu in atom_hybrid_cpus else "cpu_core")] runner_list[0].cpu_list = get_cpu_list("/sys/devices/cpu_core") if len(runner_list[0].cpu_list) == 0: sys.exit("cpu_core fallback has no cpus") # no hybrid else: r = Runner(args.level, idle_threshold, kernel_version, pmu="cpu") r.cpu_list = [] runner_list = [r] if args.all: assert all([ru.max_level <= args.level for ru in runner_list]) return runner_list def handle_more_options(args): if args.single_thread: cpu.ht = False if args.quiet: if not args.desc: args.no_desc = True args.no_util = True def tune_model(model): if args.tune_model: for t in args.tune_model: exec(t) def init_model(model, runner, pe): version = model.version model.print_error = pe model.check_event = lambda ev: ectx.emap.getevent(ev) is not None model.Setup(runner) model.num_cores = len(cpu.coreids) # includes sockets if cpu.ht: model.num_threads = 2 if "Errata_Whitelist" in model.__dict__: ectx.errata_whitelist += model.Errata_Whitelist.split(";") if "base_frequency" in model.__dict__: model.base_frequency = cpu.freq * 1000 if "model" in model.__dict__: model.model = cpu.modelid if "Num_CPUs" in model.__dict__: model.Num_CPUs = lambda a, b, c: len(cpu.allcpus) if args.fp16: if "FP16" in model.__dict__: model.FP16 = lambda a, b, c: True else: sys.exit("--fp16 option but no support in model") if args.hbm_only: if "HBM_Only" in model.__dict__: model.HBM_Only = lambda a, b, c: True else: sys.exit("--hbm-only option but no support in model") tune_model(model) return version def legacy_smt_setup(model): global smt_mode if args.thread: model.ebs_mode = cpu.ht return model.smt_enabled = cpu.ht smt_mode |= cpu.ht def load_default_retlat(): global ret_latency if ret_latency is None: name = cpu.cpu.split("-")[0] fn = os.path.dirname(os.path.realpath(__file__)) + ("/%s-retlat.json" % name) try: ret_latency = json.load(open(fn))["Data"] except IOError: sys.exit("Cannot find default ret latency file %s\n" % fn + "Please generate with representative workload using genretlat -o %s workload" % fn) def model_setup(runner, cpuname, pe, kernel_version): global smt_mode if cpuname == "ivb": import ivb_client_ratios model = ivb_client_ratios legacy_smt_setup(model) elif cpuname == "ivt": import ivb_server_ratios model = ivb_server_ratios legacy_smt_setup(model) elif cpuname == "snb": import snb_client_ratios model = snb_client_ratios legacy_smt_setup(model) elif cpuname == "jkt": import jkt_server_ratios model = jkt_server_ratios legacy_smt_setup(model) elif cpuname == "hsw": import hsw_client_ratios model = hsw_client_ratios legacy_smt_setup(model) elif cpuname == "hsx": import hsx_server_ratios model = hsx_server_ratios legacy_smt_setup(model) elif cpuname == "bdw": import bdw_client_ratios model = bdw_client_ratios legacy_smt_setup(model) elif cpuname == "bdx": import bdx_server_ratios model = bdx_server_ratios legacy_smt_setup(model) elif cpuname == "skl": import skl_client_ratios model = skl_client_ratios legacy_smt_setup(model) elif cpuname == "skx": import skx_server_ratios model = skx_server_ratios legacy_smt_setup(model) elif cpuname == "clx": import clx_server_ratios model = clx_server_ratios legacy_smt_setup(model) elif cpuname == "icx": import icx_server_ratios icx_server_ratios.smt_enabled = cpu.ht model = icx_server_ratios setup_metrics(model, runner.pmu) # work around kernel constraint table bug in some kernel versions if kernel_version < 510: ectx.constraint_fixes["CYCLE_ACTIVITY.STALLS_MEM_ANY"] = "0,1,2,3" smt_mode = cpu.ht elif cpu.cpu == "spr": import spr_server_ratios spr_server_ratios.smt_enabled = cpu.ht model = spr_server_ratios setup_metrics(model, runner.pmu) smt_mode = cpu.ht if kernel_version < 670: # expect to be fixed in 6.7 # kernel incorrectly schedules ocr on 0-3 only ectx.constraint_patterns.append(("OCR.", "0,1,2,3", )) elif cpu.cpu == "sprmax": import spr_max_server_ratios spr_max_server_ratios.smt_enabled = cpu.ht model = spr_max_server_ratios setup_metrics(model, runner.pmu) smt_mode = cpu.ht if kernel_version < 670: # expect to be fixed in 6.7 # kernel incorrectly schedules ocr on 0-3 only ectx.constraint_patterns.append(("OCR.", "0,1,2,3", )) elif cpuname == "icl": import icl_client_ratios icl_client_ratios.smt_enabled = cpu.ht model = icl_client_ratios setup_metrics(model, runner.pmu) # work around kernel constraint table bug in some kernel versions if kernel_version < 510: ectx.constraint_fixes["CYCLE_ACTIVITY.STALLS_MEM_ANY"] = "0,1,2,3" smt_mode = cpu.ht elif cpuname == "tgl": import icl_client_ratios icl_client_ratios.smt_enabled = cpu.ht model = icl_client_ratios setup_metrics(model, runner.pmu) if kernel_version < 510: ectx.constraint_fixes["CYCLE_ACTIVITY.STALLS_MEM_ANY"] = "0,1,2,3" smt_mode = cpu.ht elif (cpuname == "adl" and runner.pmu in ("cpu_core", "cpu")) or cpuname == "adl-glc": import adl_glc_ratios setup_metrics(adl_glc_ratios, runner.pmu) adl_glc_ratios.smt_enabled = cpu.ht model = adl_glc_ratios smt_mode = cpu.ht if kernel_version < 670: # expect to be fixed in 6.7 # kernel incorrectly schedules ocr on 0-3 only ectx.constraint_patterns.append(("OCR.", "0,1,2,3", )) elif (cpuname == "adl" and runner.pmu == "cpu_atom") or cpuname == "adl-grt": import adl_grt_ratios model = adl_grt_ratios model.use_aux = args.aux elif (cpuname == "mtl" and runner.pmu in ("cpu_core", "cpu")) or cpuname == "mtl-rwc": import mtl_rwc_ratios setup_metrics(mtl_rwc_ratios, runner.pmu) mtl_rwc_ratios.smt_enabled = cpu.ht model = mtl_rwc_ratios ectx.constraint_patterns.append(("OCR.", "0,1,2,3", )) elif (cpuname == "mtl" and runner.pmu == "cpu_atom") or cpuname == "mtl-cmt": import mtl_cmt_ratios model = mtl_cmt_ratios model.use_aux = args.aux elif cpuname == "slm": import slm_ratios model = slm_ratios elif cpuname == "knl": import knl_ratios knl_ratios.smt_enabled = smt_mode = cpu.ht model = knl_ratios elif cpuname == "ehl": import ehl_ratios model = ehl_ratios else: if not args.quiet: print("Warning: Unknown CPU model number, falling back to simple model") ht_warning() import simple_ratios model = simple_ratios return init_model(model, runner, pe) def runner_emaps(pe, runner_list): version = "" for runner in runner_list: runner.set_ectx() emap = ocperf.find_emap(pmu=runner.pmu if runner.pmu else "cpu") if not emap: ocperf.ocverbose = True ocperf.find_emap() sys.exit("Unknown CPU or CPU event map not found (EVENTMAP:%s, model:%d)" % (os.environ["EVENTMAP"] if "EVENTMAP" in os.environ else "?", cpu.model)) runner.ectx.init_emap(emap) runner.printer.init_emap(emap) if version: version += ", " version += model_setup(runner, cpu.cpu, pe, runner.kernel_version) runner.clear_ectx() return version def setup_pe(): pe = lambda x: None if args.debug: printed_error = set() def print_err(x): if x not in printed_error: print(x) printed_error.add(x) pe = lambda e: print_err(e) return pe def handle_misc_options(args, version): if args.version: print("toplev, CPU: %s, TMA version: %s" % (cpu.cpu, version)) sys.exit(0) if args.gen_script: args.quiet = True if args.subset: if not args.import_: sys.exit("--subset requires --import mode") if args.script_record: sys.exit("--subset cannot be used with --script-record. Generate temp file with perf stat report -x\\;") def handle_cmd(args, runner_list, rest): if args.describe: args.long_desc = True if not rest: sys.exit("No nodes to describe") for r in runner_list: r.list_nodes(None, any_node, rest) if args.list_metrics or args.list_all: for r in runner_list: r.list_nodes("Metrics", lambda obj: obj.metric, rest) if args.list_nodes or args.list_all: for r in runner_list: r.list_nodes("Nodes", lambda obj: not obj.metric, rest) if args.list_metric_groups or args.list_all: for r in runner_list: r.list_metric_groups() if args.list_metric_groups or args.list_metrics or args.list_nodes or args.list_all or args.describe: if any([x.startswith("-") for x in rest]): sys.exit("Unknown arguments for --list*/--describe") sys.exit(0) def has_core_node(runner): res = False runner.set_ectx() for o in runner.olist: if core_node(o): res = True break runner.clear_ectx() return res def any_core_node(runner_list): for r in runner_list: if has_core_node(r): return True return False def check_root(): if not (os.geteuid() == 0 or sysctl("kernel.perf_event_paranoid") == -1) and not args.quiet: print("Warning: Needs root or echo -1 > /proc/sys/kernel/perf_event_paranoid", file=sys.stderr) def extra_setup_once(runner, rest): if not args.no_util: import perf_metrics perf_metrics.Setup(runner) if args.power and feat.supports_power: import power_metrics power_metrics.Setup(runner) if args.sw: import linux_metrics linux_metrics.Setup(runner) if args.power and feat.supports_power: if not args.quiet and not args.import_ and not args.print: print("Running with --power. Will measure complete system.") if args.single_thread: print("--single-thread conflicts with --power") check_root() rest = add_args(rest, "-a") return rest def extra_setup(runner): if args.tsx and cpu.has_tsx and cpu.cpu in tsx_cpus and runner.pmu in ("cpu", "cpu_core"): import tsx_metrics tsx_metrics.Setup(runner) if args.frequency: import frequency frequency.SetupCPU(runner, cpu) def runner_extra_init(args, rest, runner_list): rest = extra_setup_once(runner_list[0], rest) for r in runner_list: extra_setup(r) if args.nodes: check_nodes(runner_list, args.nodes) for r in runner_list: r.setup_children() return rest def runner_filter(args, rest, runner_list): for r in runner_list: rest = r.filter_per_core(args.single_thread, rest) return rest def update_smt(args, rest): if not smt_mode and not args.single_thread and not args.no_aggr: hybrid = cpu.cpu in hybrid_cpus multi = output_count() if multi > 0: rest = add_args(rest, "-a") if (multi > 1 or args.per_thread) and not hybrid: args.no_aggr = True if args.per_socket and multi == 1 and not hybrid: rest = add_args(rest, "--per-socket") if args.per_core and multi == 1 and not hybrid: rest = add_args(rest, "--per-core") return rest def runner_node_filter(runner_list): for r in runner_list: r.filter_nodes() def update_smt_mode(runner_list): if smt_mode and not os.getenv('FORCEHT'): # do not need SMT mode if no objects have Core scope if not any_core_node(runner_list): return False return smt_mode def check_full_system(args, rest): full_system = False if not args.single_thread and smt_mode: if not args.quiet and not args.import_: print("Will measure complete system.") if smt_mode: if args.cpu: print("Warning: --cpu/-C mode with HyperThread must specify all core thread pairs!", file=sys.stderr) if args.pid: sys.exit("-p/--pid mode not compatible with SMT. Use sleep in global mode.") check_root() rest = add_args(rest, "-a") args.no_aggr = True full_system = True else: full_system = args.no_aggr or "--per-core" in rest or "--per-socket" in rest if args.no_aggr: rest = add_args(rest, "-A") return full_system, rest output_numcpus = False def init_perf_output(args, rest, full_system): if (args.perf_output or args.perf_summary) and not args.no_csv_header: ph = [] if args.interval: ph.append("Timestamp") if full_system: ph.append("Location") if ("--per-socket" in rest or "--per-core" in rest) and not args.no_aggr: ph.append("Num-CPUs") global output_numcpus output_numcpus = True ph += ["Value", "Unit", "Event", "Run-Time", "Enabled", "", ""] if args.perf_output: args.perf_output.write(";".join(ph) + "\n") if args.perf_summary: args.perf_summary.write(";".join(ph) + "\n") def setup_cpus(args, rest, cpu, runner_list): if args.cpu: allcpus = parse_cpu_list(args.cpu) else: allcpus = cpu.allcpus if args.core: allowed_threads = [x for x in allcpus if display_core(x, False)] allowed_cores = [x for x in allcpus if display_core(x, True)] rest = ["-C", ",".join(["%d" % x for x in allowed_cores])] + rest else: allowed_threads = allcpus if len(runner_list) > 1 and args.no_aggr and not runner_list[0].cpu_list: # XXX cores = list(sorted(cpu.coreids.keys())) part = len(cores)//len(runner_list) start = 0 for r in runner_list: r.cpu_list = sorted(flatten([cpu.coreids[x] for x in cores[start:start+part]])) start += part else: for r in runner_list: if r.cpu_list: r.cpu_list = sorted(list(set(r.cpu_list) & set(allowed_threads))) else: r.cpu_list = list(allowed_threads) return rest def init_output(args, version): if args.json: if args.csv: sys.exit("Cannot combine --csv with --json") if args.columns: sys.exit("Cannot combine --columns with --json") out = tl_output.OutputJSON(args.output, args.csv, args, version, cpu) # type: tl_output.Output elif args.csv: if args.columns: out = tl_output.OutputColumnsCSV(args.output, args.csv, args, version, cpu) else: out = tl_output.OutputCSV(args.output, args.csv, args, version, cpu) elif args.columns: out = tl_output.OutputColumns(args.output, args, version, cpu) else: out = tl_output.OutputHuman(args.output, args, version, cpu) return out def init_valcsv(out, args): if args.valcsv: out.valcsv = csv.writer(args.valcsv, lineterminator='\n', delimiter=';') if not args.no_csv_header: out.valcsv.writerow(("Timestamp", "CPU", "Group", "Event", "Value", "Perf-event", "Index", "STDDEV", "MULTI", "Nodes")) # XXX use runner_restart def runner_first_init(args, runner_list): nnodes = 0 for r in runner_list: runner_init(r) nnodes += len(r.olist) if nnodes == 0: sys.exit("No nodes enabled") if args.nodes: check_nodes(runner_list, args.nodes) offset = 0 for r in runner_list: r.set_ectx() r.sched.schedule(r.olist) r.sched.offset = offset offset += len(r.sched.evnum) r.clear_ectx() def suggest(runner): printer = runner.printer if printer.bottlenecks and not args.quiet: suggest_desc(runner) if args.level < runner.max_node_level and printer.bottlenecks: return suggest_bottlenecks(runner) return False def measure_and_sample(runner_list, count, out, orig_smt_mode, rest, full_system): rrest = rest while True: summary = Summary() try: if args.no_multiplex and not args.import_: ret = execute_no_multiplex(runner_list, out, rrest, summary) else: ret = execute(runner_list, out, rrest, summary) except KeyboardInterrupt: ret = 1 print_summary(summary, out, runner_list, full_system) repeat = False for runner in runner_list: runner.stat.compute_errors() repeat |= suggest(runner) if (args.show_sample or args.run_sample) and ret == 0: for runner in runner_list: runner.set_ectx() do_sample(runner.printer.sample_obj, rest, count, ret, runner.kernel_version) runner.clear_ectx() if 100 <= ret <= 200 and repeat: print("Perf or workload appears to have failed with error %d. Not drilling down" % ret, file=sys.stderr) break if count is not None: count += 1 if repeat: if not args.quiet: print("Rerunning workload", file=sys.stderr) offset = 0 nnodes = 0 for r in runner_list: offset = runner_restart(r, offset) nnodes += len(r.olist) global smt_mode smt_mode = orig_smt_mode if smt_mode and not os.getenv('FORCEHT'): if not any_core_node(runner_list): smt_mode = False # XXX do all checks for incompatible arguments like top level if smt_mode and not args.single_thread: check_root() rrest = add_args(rrest, "-a") rrest = add_args(rrest, "-A") full_system = True if nnodes == 0: sys.exit("No nodes enabled") else: break return ret, count, full_system def report_idle(runner_list): ik = set() for r in runner_list: ik |= r.idle_keys if ik and not args.quiet: print("Idle CPUs %s may have been hidden. Override with --idle-threshold 100" % idle_range_list(ik), file=sys.stderr) def report_not_supported(runner_list): notfound_caches = {} for r in runner_list: notfound_caches.update(r.ectx.notfound_cache) if notfound_caches and any(["not supported" not in x for x in notfound_caches.values()]) and not args.quiet: print("Some events not found. Consider running event_download to update event lists", file=sys.stderr) def measure(out, orig_smt_mode, rest, runner_list, full_system): if args.sample_repeat: cnt = 1 for j in range(args.sample_repeat): ret, cnt, full_system = measure_and_sample(runner_list, cnt, out, orig_smt_mode, rest, full_system) if ret: break else: ret, count, full_system = measure_and_sample(runner_list, 0 if args.drilldown else None, out, orig_smt_mode, rest, full_system) return ret def idle_range_list(l): if all([x.isdigit() for x in l]): # adapted from https://stackoverflow.com/questions/2154249/identify-groups-of-continuous-numbers-in-a-list def get_range(g): group = [x[1] for x in g] if len(group) == 1: return "%d" % group[0] return "%d-%d" % (group[0], group[-1]) l = [get_range(g) for k, g in groupby(enumerate(sorted([int(x) for x in l])), lambda x: x[0] - x[1])] return ",".join(l) def finish_graph(graphp): if args.graph: args.output.close() graphp.wait() def main(args, rest, feat, env, cpu): pversion = ocperf.PerfVersion() handle_parallel(args, env) rest = handle_rest(args, rest) open_output_files(args) update_args2(args) graphp = handle_graph(args) args.ring_filter = init_ring_filter(args) update_args_cpu(args, pversion) update_cpu(args, cpu) kernel_version = get_kernel() check_exclusive(args, kernel_version) runner_list = init_runner_list(kernel_version) global KEEP_UNREF if len(runner_list) > 1 and isinstance(KEEP_UNREF, bool): KEEP_UNREF = True # for now -- dummy can get assigned to wrong runner global INAME global FUZZYINPUT if len(runner_list) > 1 and (INAME or FUZZYINPUT): sys.exit("INAME and FUZZYINPUT do not support hybrid") handle_more_options(args) version = runner_emaps(setup_pe(), runner_list) handle_misc_options(args, version) handle_cmd(args, runner_list, rest) rest = runner_extra_init(args, rest, runner_list) rest = runner_filter(args, rest, runner_list) rest = update_smt(args, rest) runner_node_filter(runner_list) global smt_mode orig_smt_mode = smt_mode smt_mode = update_smt_mode(runner_list) full_system, rest = check_full_system(args, rest) init_perf_output(args, rest, full_system) rest = setup_cpus(args, rest, cpu, runner_list) if args.pinned: run_l1_parallel = True out = init_output(args, version) init_valcsv(out, args) runner_first_init(args, runner_list) if args.repl: import code code.interact(banner='toplev repl', local=locals()) sys.exit(0) ret = measure(out, orig_smt_mode, rest, runner_list, full_system) out.print_footer() out.flushfiles() if args.xlsx and ret == 0: ret = do_xlsx(env, args) report_idle(runner_list) report_not_supported(runner_list) finish_graph(graphp) sys.exit(ret) if __name__ == '__main__': # these are top level to avoid globals, which break the type checker # alternative would be to pass them everywhere, but that would be tedious args, rest_ = init_args() feat = PerfFeatures(args) ectx = EventContextBase() # only for type checker # allow tune to override toplevel without global if args.tune: for t in args.tune: exec(t) env_ = tl_cpu.Env() update_args(args, env_) # XXX move into ectx cpu = tl_cpu.CPU(known_cpus, nocheck=event_nocheck(), env=env_) main(args, rest_, feat, env_, cpu)
175,261
Python
.py
4,179
32.340512
193
0.574306
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,890
mtl_rwc_ratios.py
andikleen_pmu-tools/mtl_rwc_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel 14th gen Core (code name Meteor Lake) with Redwood Cove # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 topdown_use_fixed = False def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 12 Mem_L2_Store_Cost = 10 Mem_STLB_Hit_Cost = 7 MS_Switches_Cost = 3 Avg_Assist_Cost = ( 99 *3 + 63 + 30 ) / 5 Pipeline_Width = 6 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 61 PERF_METRICS_MSR = 1 DS = 0 # Aux. formulas def Br_DoI_Jumps(self, EV, level): return EV("BR_INST_RETIRED.NEAR_TAKEN", level) - EV("BR_INST_RETIRED.COND_TAKEN", level) - 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) def Branching_Retired(self, EV, level): return (EV("BR_INST_RETIRED.ALL_BRANCHES", level) + 2 * EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("INST_RETIRED.NOP", level)) / SLOTS(self, EV, level) def Serialize_Core(self, EV, level): return self.Core_Bound.compute(EV) * (self.Serializing_Operation.compute(EV) + EV("RS.EMPTY:u1", level) / CLKS(self, EV, level) * self.Ports_Utilized_0.compute(EV)) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV)) def Umisp(self, EV, level): return 10 * self.Microcode_Sequencer.compute(EV) * self.Other_Mispredicts.compute(EV) / self.Branch_Mispredicts.compute(EV) def Assist(self, EV, level): return (self.Microcode_Sequencer.compute(EV) / (self.Microcode_Sequencer.compute(EV) + self.Few_Uops_Instructions.compute(EV))) * (self.Assists.compute(EV) / self.Microcode_Sequencer.compute(EV)) def Assist_Frontend(self, EV, level): return (1 - EV("INST_RETIRED.REP_ITERATION", level) / EV("UOPS_RETIRED.MS:c1", level)) * (self.Fetch_Latency.compute(EV) * (self.MS_Switches.compute(EV) + self.Branch_Resteers.compute(EV) * (self.Clears_Resteers.compute(EV) + self.Mispredicts_Resteers.compute(EV) * self.Other_Mispredicts.compute(EV) / self.Branch_Mispredicts.compute(EV)) / (self.Clears_Resteers.compute(EV) + self.Unknown_Branches.compute(EV) + self.Mispredicts_Resteers.compute(EV))) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) def Assist_Retired(self, EV, level): return Assist(self, EV, level) * self.Heavy_Operations.compute(EV) def Core_Bound_Cycles(self, EV, level): return self.Ports_Utilized_0.compute(EV) * CLKS(self, EV, level) + Few_Uops_Executed_Threshold(self, EV, level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE_CYCLES_GE_1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.THREAD:c1", level) # factor used for metrics associating fixed costs for FB Hits - according to probability theory if all FB Hits come at a random rate in original L1_Miss cost interval then the average cost for each one is 0.5 of the fixed cost def FB_Factor(self, EV, level): return 1 + FBHit_per_L1Miss(self, EV, level) / 2 def FBHit_per_L1Miss(self, EV, level): return EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("MEM_LOAD_RETIRED.L1_MISS", level) def Fetched_Uops(self, EV, level): return EV("UOPS_ISSUED.ANY", level) def Few_Uops_Executed_Threshold(self, EV, level): return EV("EXE_ACTIVITY.1_PORTS_UTIL", level) + self.Retiring.compute(EV) * EV("EXE_ACTIVITY.2_PORTS_UTIL:u0xc", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) + 2 * EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 * EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_ARITH_INST_RETIRED.VECTOR", level) def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def Light_Ops_Sum(self, EV, level): return self.FP_Arith.compute(EV) + self.Int_Operations.compute(EV) + self.Memory_Operations.compute(EV) + self.Fused_Instructions.compute(EV) + self.Non_Fused_Branches.compute(EV) def MEM_Bound_Ratio(self, EV, level): return EV("MEMORY_ACTIVITY.STALLS_L3_MISS", level) / CLKS(self, EV, level) def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_INST_RETIRED.LOCK_LOADS", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) def Mispred_Clears_Fraction(self, EV, level): return self.Branch_Mispredicts.compute(EV) / self.Bad_Speculation.compute(EV) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.DATA_RD:c4", level)) , level ) def Store_L2_Hit_Cycles(self, EV, level): return EV("MEM_STORE_RETIRED.L2_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def True_XSNP_HitM_Fraction(self, EV, level): return EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM", level) / (EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HITM", level) + EV("OCR.DEMAND_DATA_RD.L3_HIT.SNOOP_HIT_WITH_FWD", level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 28 * Core_Frequency(self, EV, level) def Mem_XSNP_Hit_Cost(self, EV, level): return 27 * Core_Frequency(self, EV, level) def Mem_XSNP_None_Cost(self, EV, level): return 12 * Core_Frequency(self, EV, level) def Mem_L2_Hit_Cost(self, EV, level): return 3 * Core_Frequency(self, EV, level) def PERF_METRICS_SUM(self, EV, level): return (EV("PERF_METRICS.FRONTEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BAD_SPECULATION", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.RETIRING", level) / EV("TOPDOWN.SLOTS", level)) + (EV("PERF_METRICS.BACKEND_BOUND", level) / EV("TOPDOWN.SLOTS", level)) def Retire_Fraction(self, EV, level): return EV("UOPS_RETIRED.SLOTS", level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return self.Retiring.compute(EV) * SLOTS(self, EV, level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return num_cores * num_threads if num_cores else(8 + 16 /(2 - smt_enabled)) # A system parameter for dependent-loads (pointer chasing like access pattern) of the workload. An integer fraction in range from 0 (no dependent loads) to 100 (all loads are dependent loads) def Dependent_Loads_Weight(self, EV, level): return 20 # Total pipeline cost of Branch Misprediction related bottlenecks def Mispredictions(self, EV, level): val = 100 *(1 - Umisp(self, EV, level)) * (self.Branch_Mispredicts.compute(EV) + self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses) def Big_Code(self, EV, level): val = 100 * self.Fetch_Latency.compute(EV) * (self.ITLB_Misses.compute(EV) + self.ICache_Misses.compute(EV) + self.Unknown_Branches.compute(EV)) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) self.thresh = (val > 20) return val # Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end) def Instruction_Fetch_BW(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) - (1 - Umisp(self, EV, level)) * self.Fetch_Latency.compute(EV) * self.Mispredicts_Resteers.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) - Assist_Frontend(self, EV, level)) - Big_Code(self, EV, level) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks def Cache_Memory_Bandwidth(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.MEM_Bandwidth.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.SQ_Full.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.FB_Full.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of external Memory- or Cache-Latency related bottlenecks def Cache_Memory_Latency(self, EV, level): val = 100 *((self.Memory_Bound.compute(EV) * (self.DRAM_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.MEM_Latency.compute(EV) / (self.MEM_Latency.compute(EV) + self.MEM_Bandwidth.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.L3_Hit_Latency.compute(EV) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * self.L2_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.Store_Latency.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)))) + (self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.L1_Hit_Latency.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs) def Memory_Data_TLBs(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * (self.L1_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.DTLB_Load.compute(EV) / (self.Store_Fwd_Blk.compute(EV) + self.L1_Hit_Latency.compute(EV) + self.DTLB_Load.compute(EV) + self.Lock_Latency.compute(EV) + self.Split_Loads.compute(EV) + self.FB_Full.compute(EV))) + (self.Memory_Bound.compute(EV) * (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.DTLB_Store.compute(EV) / (self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors) def Memory_Synchronization(self, EV, level): val = 100 *(self.Memory_Bound.compute(EV) * ((self.L3_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * (self.Contested_Accesses.compute(EV) + self.Data_Sharing.compute(EV)) / (self.L3_Hit_Latency.compute(EV) + self.Contested_Accesses.compute(EV) + self.SQ_Full.compute(EV) + self.Data_Sharing.compute(EV)) + (self.Store_Bound.compute(EV) / (self.L1_Bound.compute(EV) + self.L3_Bound.compute(EV) + self.DRAM_Bound.compute(EV) + self.Store_Bound.compute(EV) + self.L2_Bound.compute(EV))) * self.False_Sharing.compute(EV) / ((self.Split_Stores.compute(EV) + self.DTLB_Store.compute(EV) + self.Streaming_Stores.compute(EV) + self.Store_Latency.compute(EV) + self.False_Sharing.compute(EV)) - self.Store_Latency.compute(EV))) + self.Machine_Clears.compute(EV) * (1 - self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV)))) self.thresh = (val > 10) return val # Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy. def Compute_Bound_Est(self, EV, level): val = 100 *((self.Core_Bound.compute(EV) * self.Divider.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV))) + (self.Core_Bound.compute(EV) * (self.Ports_Utilization.compute(EV) / (self.Serializing_Operation.compute(EV) + self.Ports_Utilization.compute(EV) + self.Divider.compute(EV))) * (self.Ports_Utilized_3m.compute(EV) / (self.Ports_Utilized_0.compute(EV) + self.Ports_Utilized_1.compute(EV) + self.Ports_Utilized_2.compute(EV) + self.Ports_Utilized_3m.compute(EV))))) self.thresh = (val > 20) return val # Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments) def Irregular_Overhead(self, EV, level): val = 100 *(Assist_Frontend(self, EV, level) + Umisp(self, EV, level) * self.Branch_Mispredicts.compute(EV) + (self.Machine_Clears.compute(EV) * self.Other_Nukes.compute(EV) / (self.Other_Nukes.compute(EV))) + Serialize_Core(self, EV, level) + Assist_Retired(self, EV, level)) self.thresh = (val > 10) return val # Total pipeline cost of remaining bottlenecks in the back-end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls. def Other_Bottlenecks(self, EV, level): val = 100 -(Big_Code(self, EV, level) + Instruction_Fetch_BW(self, EV, level) + Mispredictions(self, EV, level) + Cache_Memory_Bandwidth(self, EV, level) + Cache_Memory_Latency(self, EV, level) + Memory_Data_TLBs(self, EV, level) + Memory_Synchronization(self, EV, level) + Compute_Bound_Est(self, EV, level) + Irregular_Overhead(self, EV, level) + Branching_Overhead(self, EV, level) + Useful_Work(self, EV, level)) self.thresh = (val > 20) return val # Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations def Branching_Overhead(self, EV, level): val = 100 * Branching_Retired(self, EV, level) self.thresh = (val > 5) return val # Total pipeline cost of "useful operations" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead. def Useful_Work(self, EV, level): val = 100 *(self.Retiring.compute(EV) - Branching_Retired(self, EV, level) - Assist_Retired(self, EV, level)) self.thresh = (val > 20) return val # Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled def Core_Bound_Likely(self, EV, level): val = 100 *(1 - self.Core_Bound.compute(EV) / self.Ports_Utilization.compute(EV) if self.Core_Bound.compute(EV)< self.Ports_Utilization.compute(EV) else 1) if SMT_2T_Utilization(self, EV, level)> 0.5 else 0 self.thresh = (val > 0.5) return val # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return EV("TOPDOWN.SLOTS", level) if topdown_use_fixed else EV("TOPDOWN.SLOTS", level) # Fraction of Physical Core issue-slots utilized by this Logical Processor def Slots_Utilization(self, EV, level): return SLOTS(self, EV, level) / (EV("TOPDOWN.SLOTS:percore", level) / 2) if smt_enabled else 1 # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common. def FP_Arith_Utilization(self, EV, level): return (EV("FP_ARITH_DISPATCHED.PORT_0", level) + EV("FP_ARITH_DISPATCHED.PORT_1", level) + EV("FP_ARITH_DISPATCHED.PORT_5", level)) / (2 * CORE_CLKS(self, EV, level)) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # uops Executed per Cycle def EPC(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / CLKS(self, EV, level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.DISTRIBUTED", level) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_INST_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408 def IpFLOP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / FLOP_Count(self, EV, level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_SP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_SINGLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_DP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX128(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX256(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per PAUSE (lower number means higher occurrence rate) def IpPause(self, EV, level): return Instructions(self, EV, level) / EV("CPU_CLK_UNHALTED.PAUSE_INST", level) # Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate) def IpSWPF(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("SW_PREFETCH_ACCESS.T0:u0xF", level) self.thresh = (val < 100) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.SLOTS:c1", level) # Estimated fraction of retirement-cycles dealing with repeat instructions def Strings_Cycles(self, EV, level): val = EV("INST_RETIRED.REP_ITERATION", level) / EV("UOPS_RETIRED.SLOTS:c1", level) self.thresh = (val > 0.1) return val # Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate) def IpAssist(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("ASSISTS.ANY", level) self.thresh = (val < 100000) return val def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Average number of uops fetched from LSD per cycle def Fetch_LSD(self, EV, level): return EV("LSD.UOPS", level) / EV("LSD.CYCLES_ACTIVE", level) # Average number of uops fetched from DSB per cycle def Fetch_DSB(self, EV, level): return EV("IDQ.DSB_UOPS", level) / EV("IDQ.DSB_CYCLES_ANY", level) # Average number of uops fetched from MITE per cycle def Fetch_MITE(self, EV, level): return EV("IDQ.MITE_UOPS", level) / EV("IDQ.MITE_CYCLES_ANY", level) # Average number of Uops issued by front-end when it issued something def Fetch_UpC(self, EV, level): return EV("UOPS_ISSUED.ANY", level) / EV("UOPS_ISSUED.ANY:c1", level) # Fraction of Uops delivered by the LSD (Loop Stream Detector; aka Loop Cache) def LSD_Coverage(self, EV, level): return EV("LSD.UOPS", level) / Fetched_Uops(self, EV, level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Average number of cycles the front-end was delayed due to an Unknown Branch detection. See Unknown_Branches node. def Unknown_Branch_Cost(self, EV, level): return EV("INT_MISC.UNKNOWN_BRANCH_CYCLES", level) / EV("INT_MISC.UNKNOWN_BRANCH_CYCLES:c1:e1", level) # Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details. def DSB_Switch_Cost(self, EV, level): return EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", level) / EV("DSB2MITE_SWITCHES.PENALTY_CYCLES:c1:e1", level) # Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.DSB_Switches.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV)) + self.Fetch_Bandwidth.compute(EV) * self.MITE.compute(EV) / (self.LSD.compute(EV) + self.MITE.compute(EV) + self.DSB.compute(EV))) self.thresh = (val > 10) return val # Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck. def DSB_Bandwidth(self, EV, level): val = 100 *(self.Frontend_Bound.compute(EV) * (self.Fetch_Bandwidth.compute(EV) / (self.Fetch_Bandwidth.compute(EV) + self.Fetch_Latency.compute(EV))) * (self.DSB.compute(EV) / (self.LSD.compute(EV) + self.MITE.compute(EV) + self.DSB.compute(EV)))) self.thresh = (val > 10) return val # Average Latency for L1 instruction cache misses def ICache_Miss_Latency(self, EV, level): return EV("ICACHE_DATA.STALLS", level) / EV("ICACHE_DATA.STALLS:c1:e1", level) # Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck. def IC_Misses(self, EV, level): val = 100 *(self.Fetch_Latency.compute(EV) * self.ICache_Misses.compute(EV) / (self.LCP.compute(EV) + self.ICache_Misses.compute(EV) + self.DSB_Switches.compute(EV) + self.Branch_Resteers.compute(EV) + self.MS_Switches.compute(EV) + self.ITLB_Misses.compute(EV))) self.thresh = (val > 5) return val # Instructions per non-speculative DSB miss (lower number means higher occurrence rate) def IpDSB_Miss_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FRONTEND_RETIRED.ANY_DSB_MISS", level) self.thresh = (val < 50) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # L2 cache true code cacheline misses per kilo instruction def L2MPKI_Code(self, EV, level): return 1000 * EV("FRONTEND_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache speculative code cacheline misses per kilo instruction def L2MPKI_Code_All(self, EV, level): return 1000 * EV("L2_RQSTS.CODE_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional non-taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Ntaken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_NTAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate). def IpMisp_Cond_Taken(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.COND_TAKEN", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate). def IpMisp_Ret(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.RET", level) self.thresh = (val < 500) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.INDIRECT", level) self.thresh = (val < 1000) return val # Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear) def Branch_Misprediction_Cost(self, EV, level): return Mispredictions(self, EV, level) * SLOTS(self, EV, level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / 100 # Speculative to Retired ratio of all clears (covering Mispredicts and nukes) def Spec_Clears_Ratio(self, EV, level): return EV("INT_MISC.CLEARS_COUNT", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) # Fraction of branches that are non-taken conditionals def Cond_NT(self, EV, level): return EV("BR_INST_RETIRED.COND_NTAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are taken conditionals def Cond_TK(self, EV, level): return EV("BR_INST_RETIRED.COND_TAKEN", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are CALL or RET def CallRet(self, EV, level): return (EV("BR_INST_RETIRED.NEAR_CALL", level) + EV("BR_INST_RETIRED.NEAR_RETURN", level)) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches that are unconditional (direct or indirect) jumps def Jump(self, EV, level): return Br_DoI_Jumps(self, EV, level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) # Fraction of branches of other types (not individually covered by other metrics in Info.Branches group) def Other_Branches(self, EV, level): return 1 -(Cond_NT(self, EV, level) + Cond_TK(self, EV, level) + CallRet(self, EV, level) + Jump(self, EV, level)) # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("MEM_LOAD_COMPLETED.L1_MISS_ANY", level) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L1 cache true misses per kilo instruction for all demand loads (including speculative) def L1MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.ALL_DEMAND_DATA_RD", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for all request types (including speculative) def L2MPKI_All(self, EV, level): return 1000 * EV("L2_RQSTS.MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for all demand loads (including speculative) def L2MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("L2_RQSTS.RFO_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all request types (including speculative) def L2HPKI_All(self, EV, level): return 1000 *(EV("L2_RQSTS.REFERENCES", level) - EV("L2_RQSTS.MISS", level)) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all demand loads (including speculative) def L2HPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_HIT", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.L3_MISS", level) / EV("INST_RETIRED.ANY", level) # Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss-handling entries) def FB_HPKI(self, EV, level): return 1000 * EV("MEM_LOAD_RETIRED.FB_HIT", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) def L3_Cache_Access_BW(self, EV, level): return 64 * EV("OFFCORE_REQUESTS.ALL_REQUESTS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_PENDING", level) + EV("DTLB_LOAD_MISSES.WALK_PENDING", level) + EV("DTLB_STORE_MISSES.WALK_PENDING", level)) / (4 * CORE_CLKS(self, EV, level)) self.thresh = (val > 0.5) return val # STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Code_STLB_MPKI(self, EV, level): return 1000 * EV("ITLB_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Load_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_LOAD_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk) def Store_STLB_MPKI(self, EV, level): return 1000 * EV("DTLB_STORE_MISSES.WALK_COMPLETED", level) / EV("INST_RETIRED.ANY", level) # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average per-core data access bandwidth to the L3 cache [GB / sec] def L3_Cache_Access_BW_2T(self, EV, level): return L3_Cache_Access_BW(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Latency for L3 cache miss demand Loads def Load_L3_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.L3_MISS_DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.L3_MISS_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD:c1", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Un-cacheable retired load per kilo instruction def UC_Load_PKI(self, EV, level): return 1000 * EV("MEM_LOAD_MISC_RETIRED.UC", level) / EV("INST_RETIRED.ANY", level) # "Bus lock" per kilo instruction def Bus_Lock_PKI(self, EV, level): return 1000 * EV("SQ_MISC.BUS_LOCK", level) / EV("INST_RETIRED.ANY", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_DISTRIBUTED", level) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Fraction of cycles the processor is waiting yet unhalted; covering legacy PAUSE instruction, as well as C0.1 / C0.2 power-performance optimized states. Sample code of TPAUSE: https://github.com/torvalds/linux/blob/master/arch/x86/lib/delay.c#L105. If running on Linux, please check the power control interface: https://github.com/torvalds/linux/blob/master/arch/x86/kernel/cpu/umwait.c and https://github.com/torvalds/linux/blob/master/Documentation/ABI/testing/sysfs-devices-system-cpu#L587 def C0_Wait(self, EV, level): val = EV("CPU_CLK_UNHALTED.C0_WAIT", level) / CLKS(self, EV, level) self.thresh = (val > 0.05) return val # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return 64 *(EV("UNC_HAC_ARB_TRK_REQUESTS.ALL", level) + EV("UNC_HAC_ARB_COH_TRK_REQUESTS.ALL", level)) / OneBillion / Time(self, EV, level) # Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches def MEM_Parallel_Reads(self, EV, level): return EV("UNC_ARB_DAT_OCCUPANCY.RD", level) / EV("UNC_ARB_DAT_OCCUPANCY.RD:c1", level) # Total package Power in Watts def Power(self, EV, level): return EV("UNC_PKG_ENERGY_STATUS", level) * Energy_Unit /(Time(self, EV, level) * OneMillion ) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_CLOCK.SOCKET", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_4:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.FRONTEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) - EV("INT_MISC.UOP_DROPPING", 1) / SLOTS(self, EV, 1) if topdown_use_fixed else(EV("IDQ_BUBBLES.CORE", 1) - EV("INT_MISC.UOP_DROPPING", 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_16:pp', 'FRONTEND_RETIRED.LATENCY_GE_8:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = ((EV("PERF_METRICS.FETCH_LATENCY", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) - EV("INT_MISC.UOP_DROPPING", 2) / SLOTS(self, EV, 2)) if topdown_use_fixed else(EV("IDQ_BUBBLES.CYCLES_0_UOPS_DELIV.CORE", 2) * Pipeline_Width - EV("INT_MISC.UOP_DROPPING", 2)) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.L2_MISS:pp', 'FRONTEND_RETIRED.L1I_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_DATA.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.STLB_MISS:pp', 'FRONTEND_RETIRED.ITLB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE_TAG.STALLS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.CLEAR_RESTEER_CYCLES", 3) / CLKS(self, EV, 3) + self.Unknown_Branches.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class Mispredicts_Resteers: name = "Mispredicts_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 4) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Mispredicts_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage.""" class Clears_Resteers: name = "Clears_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['INT_MISC.CLEAR_RESTEER_CYCLES'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'MachineClears']) maxval = None def compute(self, EV): try: self.val = (1 - Mispred_Clears_Fraction(self, EV, 4)) * EV("INT_MISC.CLEAR_RESTEER_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Clears_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears.""" class Unknown_Branches: name = "Unknown_Branches" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['FRONTEND_RETIRED.UNKNOWN_BRANCH'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("INT_MISC.UNKNOWN_BRANCH_CYCLES", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Unknown_Branches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (e.g. first time the branch is fetched or hitting BTB capacity limit) hence called Unknown Branches""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.MS_FLOWS'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("UOPS_RETIRED.MS:c1:e1", 3) / Retire_Fraction(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DECODE.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.DSB_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['FRONTEND_RETIRED.LATENCY_GE_2_BUBBLES_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_1:pp', 'FRONTEND_RETIRED.LATENCY_GE_2:pp'] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = ['FRONTEND_RETIRED.ANY_DSB_MISS'] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.MITE_CYCLES_ANY", 3) - EV("IDQ.MITE_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Decoder0_Alone: name = "Decoder0_Alone" domain = "Slots_Estimated" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("INST_DECODED.DECODERS:c1", 4) - EV("INST_DECODED.DECODERS:c2", 4)) / CORE_CLKS(self, EV, 4) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Decoder0_Alone zero division") return self.val desc = """ This metric represents fraction of cycles where decoder-0 was the only active decoder""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.DSB_CYCLES_ANY", 3) - EV("IDQ.DSB_CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class LSD: name = "LSD" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'LSD']) maxval = None def compute(self, EV): try: self.val = (EV("LSD.CYCLES_ACTIVE", 3) - EV("LSD.CYCLES_OK", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LSD zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to LSD (Loop Stream Detector) unit. LSD typically does well sustaining Uop supply. However; in some rare cases; optimal uop-delivery could not be reached for small loops whose size (in terms of number of uops) does not suit well the LSD structure.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = max(1 -(self.Frontend_Bound.compute(EV) + self.Backend_Bound.compute(EV) + self.Retiring.compute(EV)) , 0 ) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['TOPDOWN.BR_MISPREDICT_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.BRANCH_MISPREDICTS", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("TOPDOWN.BR_MISPREDICT_SLOTS", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Other_Mispredicts: name = "Other_Mispredicts" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'BrMispredicts']) maxval = None def compute(self, EV): try: self.val = max(self.Branch_Mispredicts.compute(EV) * (1 - EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) / (EV("INT_MISC.CLEARS_COUNT", 3) - EV("MACHINE_CLEARS.COUNT", 3))) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Mispredicts zero division") return self.val desc = """ This metric estimates fraction of slots the CPU was stalled due to other cases of misprediction (non-retired x86 branches or other types).""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Other_Nukes: name = "Other_Nukes" domain = "Slots" area = "BAD" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'Machine_Clears']) maxval = None def compute(self, EV): try: self.val = max(self.Machine_Clears.compute(EV) * (1 - EV("MACHINE_CLEARS.MEMORY_ORDERING", 3) / EV("MACHINE_CLEARS.COUNT", 3)) , 0.0001 ) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Nukes zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Nukes (Machine Clears) not related to memory ordering.""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = ['TOPDOWN.BACKEND_BOUND_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.BACKEND_BOUND", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) if topdown_use_fixed else EV("TOPDOWN.BACKEND_BOUND_SLOTS", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.MEMORY_BOUND", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("TOPDOWN.MEMORY_BOUND_SLOTS", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT:pp', 'MEM_LOAD_RETIRED.FB_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3) - EV("MEMORY_ACTIVITY.STALLS_L1D_MISS", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.STLB_HIT_LOADS", 4) * min(EV("MEM_INST_RETIRED.STLB_HIT_LOADS", 999) , Mem_STLB_Hit_Cost) / CLKS(self, EV, 4) + self.Load_STLB_Miss.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Load_STLB_Hit: name = "Load_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = max(0 , self.DTLB_Load.compute(EV) - self.Load_STLB_Miss.compute(EV)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the (first level) DTLB was missed by load accesses, that later on hit in second-level TLB (STLB)""" class Load_STLB_Miss: name = "Load_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_LOAD_MISSES.WALK_ACTIVE", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Load_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the Second-level TLB (STLB) was missed by load accesses, performing a hardware page walk""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class L1_Hit_Latency: name = "L1_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L1_HIT'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = min(2 *(EV("MEM_INST_RETIRED.ALL_LOADS", 4) - EV("MEM_LOAD_RETIRED.FB_HIT", 4) - EV("MEM_LOAD_RETIRED.L1_MISS", 4)) * Dependent_Loads_Weight(self, EV, 4) / 100 , max(EV("CYCLE_ACTIVITY.CYCLES_MEM_ANY", 4) - EV("MEMORY_ACTIVITY.CYCLES_L1D_MISS", 4) , 0)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Hit_Latency zero division") return self.val desc = """ This metric roughly estimates fraction of cycles with demand load accesses that hit the L1 cache. The short latency of the L1 data cache may be exposed in pointer-chasing memory access patterns as an example.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.LOCK_LOADS'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.LOCK_LOADS", 4) * EV("MEM_INST_RETIRED.LOCK_LOADS", 999) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.SPLIT_LOADS", 4) * min(EV("MEM_INST_RETIRED.SPLIT_LOADS", 999) , Load_Miss_Real_Latency(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = EV("L1D_PEND_MISS.FB_FULL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("MEMORY_ACTIVITY.STALLS_L1D_MISS", 3) - EV("MEMORY_ACTIVITY.STALLS_L2_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("MEMORY_ACTIVITY.STALLS_L2_MISS", 3) - EV("MEMORY_ACTIVITY.STALLS_L3_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD', 'MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", 4) * min(EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_MISS", 999) , Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) + EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", 4) * min(EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", 999) , Mem_XSNP_HitM_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * True_XSNP_HitM_Fraction(self, EV, 4)) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", 4) * min(EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_NO_FWD", 999) , Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) + EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", 4) * min(EV("MEM_LOAD_L3_HIT_RETIRED.XSNP_FWD", 999) , Mem_XSNP_Hit_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * (1 - True_XSNP_HitM_Fraction(self, EV, 4))) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = EV("MEM_LOAD_RETIRED.L3_HIT", 4) * min(EV("MEM_LOAD_RETIRED.L3_HIT", 999) , Mem_XSNP_None_Cost(self, EV, 4) - Mem_L2_Hit_Cost(self, EV, 4)) * FB_Factor(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = (EV("XQ.FULL_CYCLES", 4) + EV("L1D_PEND_MISS.L2_STALLS", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_RETIRED.L3_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = MEM_Bound_Ratio(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_INST_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.BOUND_ON_STORES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OCR.DEMAND_RFO.L3_HIT.SNOOP_HITM", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.SPLIT_STORES", 4) * min(EV("MEM_INST_RETIRED.SPLIT_STORES", 999) , 1) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class Streaming_Stores: name = "Streaming_Stores" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['OCR.STREAMING_WR.ANY_RESPONSE'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBW', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = 9 * EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) / CLKS(self, EV, 4) if DS else 0 EV("OCR.STREAMING_WR.ANY_RESPONSE", 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Streaming_Stores zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to Streaming store memory accesses; Streaming store optimize out a read request required by RFO stores. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should Streaming stores be a bottleneck.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_INST_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("MEM_INST_RETIRED.STLB_HIT_STORES", 4) * min(EV("MEM_INST_RETIRED.STLB_HIT_STORES", 999) , Mem_STLB_Hit_Cost) / CLKS(self, EV, 4) + self.Store_STLB_Miss.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Store_STLB_Hit: name = "Store_STLB_Hit" domain = "Clocks_Estimated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = max(0 , self.DTLB_Store.compute(EV) - self.Store_STLB_Miss.compute(EV)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Hit zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the TLB was missed by store accesses, hitting in the second- level TLB (STLB)""" class Store_STLB_Miss: name = "Store_STLB_Miss" domain = "Clocks_Calculated" area = "BE/Mem" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("DTLB_STORE_MISSES.WALK_ACTIVE", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_STLB_Miss zero division") return self.val desc = """ This metric estimates the fraction of cycles where the STLB was missed by store accesses, performing a hardware page walk""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.DIV_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.DIV_ACTIVE", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Serializing_Operation: name = "Serializing_Operation" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['RESOURCE_STALLS.SCOREBOARD'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SCOREBOARD", 3) / CLKS(self, EV, 3) + self.C02_WAIT.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Serializing_Operation zero division") return self.val desc = """ This metric represents fraction of cycles the CPU issue- pipeline was stalled due to serializing operations. Instructions like CPUID; WRMSR or LFENCE serialize the out- of-order execution which may limit performance.""" class Slow_Pause: name = "Slow_Pause" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['CPU_CLK_UNHALTED.PAUSE_INST'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.PAUSE", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Slow_Pause zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to PAUSE Instructions.""" class C01_WAIT: name = "C01_WAIT" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['C0Wait']) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.C01", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "C01_WAIT zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due staying in C0.1 power-performance optimized state (Faster wakeup time; Smaller power savings).""" class C02_WAIT: name = "C02_WAIT" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['C0Wait']) maxval = None def compute(self, EV): try: self.val = EV("CPU_CLK_UNHALTED.C02", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "C02_WAIT zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due staying in C0.2 power-performance optimized state (Slower wakeup time; Larger power savings).""" class Memory_Fence: name = "Memory_Fence" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("MISC2_RETIRED.LFENCE", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Fence zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to LFENCE Instructions.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Core_Bound_Cycles(self, EV, 3) / CLKS(self, EV, 3) if (EV("ARITH.DIV_ACTIVE", 3)<(EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) - EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3))) else Few_Uops_Executed_Threshold(self, EV, 3) / CLKS(self, EV, 3) EV("EXE_ACTIVITY.BOUND_ON_LOADS", 3) EV("CYCLE_ACTIVITY.STALLS_TOTAL", 3) EV("ARITH.DIV_ACTIVE", 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = max((EV("EXE_ACTIVITY.EXE_BOUND_0_PORTS", 4) + max(EV("RS.EMPTY_RESOURCE", 4) - EV("RESOURCE_STALLS.SCOREBOARD", 4) , 0)) / CLKS(self, EV, 4) , 1) * (EV("CYCLE_ACTIVITY.STALLS_TOTAL", 4) - EV("EXE_ACTIVITY.BOUND_ON_LOADS", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Mixing_Vectors: name = "Mixing_Vectors" domain = "Clocks" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 160 * EV("ASSISTS.SSE_AVX_MIX", 5) / CLKS(self, EV, 5) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error(self, "Mixing_Vectors zero division") return self.val desc = """ This metric estimates penalty in terms of percentage of cycles. Usually a Mixing_Vectors over 5% is worth investigating. Read more in Appendix B1 of the Optimizations Guide for this topic.""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.1_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.1_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['EXE_ACTIVITY.2_PORTS_UTIL'] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("EXE_ACTIVITY.2_PORTS_UTIL", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = ['UOPS_EXECUTED.CYCLES_GE_3'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_EXECUTED.CYCLES_GE_3", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_0", 5) + EV("UOPS_DISPATCHED.PORT_1", 5) + EV("UOPS_DISPATCHED.PORT_5_11", 5) + EV("UOPS_DISPATCHED.PORT_6", 5)) / (5 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU and 2nd branch""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_6: name = "Port_6" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED.PORT_6'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_6", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_6 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 Primary Branch and simple ALU""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_2_3_10'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED.PORT_2_3_10", 5) / (3 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = ['UOPS_DISPATCHED.PORT_7_8'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED.PORT_4_9", 5) + EV("UOPS_DISPATCHED.PORT_7_8", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.RETIRING", 1) / EV("TOPDOWN.SLOTS", 1)) / PERF_METRICS_SUM(self, EV, 1) if topdown_use_fixed else EV("UOPS_RETIRED.SLOTS", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. .. Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) * EV("UOPS_EXECUTED.X87", 4) / EV("UOPS_EXECUTED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class Int_Operations: name = "Int_Operations" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Int_Vector_128b.compute(EV) + self.Int_Vector_256b.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Operations zero division") return self.val desc = """ This metric represents overall Integer (Int) select operations fraction the CPU has executed (retired). Vector/Matrix Int operations and shuffles are counted. Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain.""" class Int_Vector_128b: name = "Int_Vector_128b" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'IntVector', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = (EV("INT_VEC_RETIRED.ADD_128", 4) + EV("INT_VEC_RETIRED.VNNI_128", 4)) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Vector_128b zero division") return self.val desc = """ This metric represents 128-bit vector Integer ADD/SUB/SAD or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.""" class Int_Vector_256b: name = "Int_Vector_256b" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'IntVector', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = (EV("INT_VEC_RETIRED.ADD_256", 4) + EV("INT_VEC_RETIRED.MUL_256", 4) + EV("INT_VEC_RETIRED.VNNI_256", 4)) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Int_Vector_256b zero division") return self.val desc = """ This metric represents 256-bit vector Integer ADD/SUB/SAD/MUL or VNNI (Vector Neural Network Instructions) uops fraction the CPU has retired.""" class Memory_Operations: name = "Memory_Operations" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("MEM_UOP_RETIRED.ANY", 3) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring memory operations -- uops for memory load or store accesses.""" class Fused_Instructions: name = "Fused_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Branches', 'BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INST_RETIRED.MACRO_FUSED", 3) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fused_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring fused instructions -- where one uop can represent multiple contiguous instructions. CMP+JCC or DEC+JCC are common examples of legacy fusions. { Note new MOV+OP and Load+OP fusions appear under Other_Light_Ops in MTL!}. See section 'Optimizing for Macro-fusion' in Optimization Manual:""" class Non_Fused_Branches: name = "Non_Fused_Branches" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Branches', 'BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * (EV("BR_INST_RETIRED.ALL_BRANCHES", 3) - EV("INST_RETIRED.MACRO_FUSED", 3)) / Retired_Slots(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Non_Fused_Branches zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring branch instructions that were not fused. Non- conditional branches like direct JMP or CALL would count here. Can be used to examine fusible conditional jumps that were not fused.""" class Other_Light_Ops: name = "Other_Light_Ops" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Pipeline']) maxval = None def compute(self, EV): try: self.val = max(0 , self.Light_Operations.compute(EV) - Light_Ops_Sum(self, EV, 3)) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Other_Light_Ops zero division") return self.val desc = """ This metric represents the remaining light uops fraction the CPU has executed - remaining means not covered by other sibling nodes. May undercount due to FMA double counting""" class Nop_Instructions: name = "Nop_Instructions" domain = "Slots" area = "RET" level = 4 htoff = False sample = ['INST_RETIRED.NOP'] errcount = 0 sibling = None metricgroup = frozenset(['BvBO', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INST_RETIRED.NOP", 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Nop_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring NOP (no op) instructions. Compilers often use NOPs for certain address alignments - e.g. start address of a function or loop body.. Improve Codegen by correctly placing NOPs outside hot sections (e.g. outside loop body).""" class Shuffles_256b: name = "Shuffles_256b" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC', 'Pipeline']) maxval = None def compute(self, EV): try: self.val = self.Light_Operations.compute(EV) * EV("INT_VEC_RETIRED.SHUFFLES", 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Shuffles_256b zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring Shuffle operations of 256-bit vector size (FP or Integer). Shuffles may incur slow cross \"vector lane\" data transfers.""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['UOPS_RETIRED.HEAVY'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = (EV("PERF_METRICS.HEAVY_OPERATIONS", 2) / EV("TOPDOWN.SLOTS", 2)) / PERF_METRICS_SUM(self, EV, 2) if topdown_use_fixed else EV("UOPS_RETIRED.HEAVY", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences. .""" class Few_Uops_Instructions: name = "Few_Uops_Instructions" domain = "Slots" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Heavy_Operations.compute(EV) - self.Microcode_Sequencer.compute(EV)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Few_Uops_Instructions zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring instructions that that are decoder into two or up to five uops. This highly-correlates with the number of uops in such instructions.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['UOPS_RETIRED.MS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_RETIRED.MS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['ASSISTS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("ASSISTS.ANY", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class Page_Faults: name = "Page_Faults" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 99 * EV("ASSISTS.PAGE_FAULT", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error(self, "Page_Faults zero division") return self.val desc = """ This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Page Faults. A Page Fault may apply on first application access to a memory page. Note operating system handling of page faults accounts for the majority of its cost.""" class FP_Assists: name = "FP_Assists" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = 30 * EV("ASSISTS.FP", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "FP_Assists zero division") return self.val desc = """ This metric roughly estimates fraction of slots the CPU retired uops as a result of handing Floating Point (FP) Assists. FP Assist may apply when working with very small floating point values (so-called Denormals).. Consider DAZ (Denormals Are Zero) and/or FTZ (Flush To Zero) options in your compiler; \"-ffast-math\" with -O2 in GCC for example. This option may improve performance if the denormal values are not critical in your application. Also note that the DAZ and FTZ modes are not compatible with the IEEE Standard 754.. https://www.intel.com/content/www/us/en/develop/docume ntation/vtune-help/top/reference/cpu-metrics-reference/bad- speculation-back-end-bound-pipeline-slots/fp-assists.html""" class AVX_Assists: name = "AVX_Assists" domain = "Slots_Estimated" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = 63 * EV("ASSISTS.SSE_AVX_MIX", 5) / SLOTS(self, EV, 5) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "AVX_Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops as a result of handing SSE to AVX* or AVX* to SSE transition Assists.""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = ['FRONTEND_RETIRED.MS_FLOWS'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_Mispredictions: name = "Mispredictions" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts', 'BvMP']) sibling = None def compute(self, EV): try: self.val = Mispredictions(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Mispredictions zero division") desc = """ Total pipeline cost of Branch Misprediction related bottlenecks""" class Metric_Big_Code: name = "Big_Code" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBC', 'BigFootprint', 'Fed', 'Frontend', 'IcMiss', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Big_Code(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Big_Code zero division") desc = """ Total pipeline cost of instruction fetch related bottlenecks by large code footprint programs (i-side cache; TLB and BTB misses)""" class Metric_Instruction_Fetch_BW: name = "Instruction_Fetch_BW" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvFB', 'Fed', 'FetchBW', 'Frontend']) sibling = None def compute(self, EV): try: self.val = Instruction_Fetch_BW(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Instruction_Fetch_BW zero division") desc = """ Total pipeline cost of instruction fetch bandwidth related bottlenecks (when the front-end could not sustain operations delivery to the back-end)""" class Metric_Cache_Memory_Bandwidth: name = "Cache_Memory_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMB', 'Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Bandwidth(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Bandwidth zero division") desc = """ Total pipeline cost of external Memory- or Cache-Bandwidth related bottlenecks""" class Metric_Cache_Memory_Latency: name = "Cache_Memory_Latency" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvML', 'Mem', 'MemoryLat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Cache_Memory_Latency(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Cache_Memory_Latency zero division") desc = """ Total pipeline cost of external Memory- or Cache-Latency related bottlenecks""" class Metric_Memory_Data_TLBs: name = "Memory_Data_TLBs" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMT', 'Mem', 'MemoryTLB', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Data_TLBs(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Memory_Data_TLBs zero division") desc = """ Total pipeline cost of Memory Address Translation related bottlenecks (data-side TLBs)""" class Metric_Memory_Synchronization: name = "Memory_Synchronization" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvMS', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Memory_Synchronization(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Memory_Synchronization zero division") desc = """ Total pipeline cost of Memory Synchronization related bottlenecks (data transfers and coherency updates across processors)""" class Metric_Compute_Bound_Est: name = "Compute_Bound_Est" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvCB', 'Cor']) sibling = None def compute(self, EV): try: self.val = Compute_Bound_Est(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Compute_Bound_Est zero division") desc = """ Total pipeline cost when the execution is compute-bound - an estimation. Covers Core Bound when High ILP as well as when long-latency execution units are busy.""" class Metric_Irregular_Overhead: name = "Irregular_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['Bad', 'BvIO', 'Cor', 'Ret']) sibling = None def compute(self, EV): try: self.val = Irregular_Overhead(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "Irregular_Overhead zero division") desc = """ Total pipeline cost of irregular execution (e.g. FP-assists in HPC, Wait time with work imbalance multithreaded workloads, overhead in system services or virtualized environments)""" class Metric_Other_Bottlenecks: name = "Other_Bottlenecks" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvOB', 'Cor', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Other_Bottlenecks(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Other_Bottlenecks zero division") desc = """ Total pipeline cost of remaining bottlenecks in the back- end. Examples include data-dependencies (Core Bound when Low ILP) and other unlisted memory-related stalls.""" class Metric_Branching_Overhead: name = "Branching_Overhead" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvBO', 'Ret']) sibling = None def compute(self, EV): try: self.val = Branching_Overhead(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "Branching_Overhead zero division") desc = """ Total pipeline cost of instructions used for program control-flow - a subset of the Retiring category in TMA. Examples include function calls; loops and alignments. (A lower bound). Consider Loop Unrolling or function inlining optimizations""" class Metric_Useful_Work: name = "Useful_Work" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Bottleneck" metricgroup = frozenset(['BvUW', 'Ret']) sibling = None def compute(self, EV): try: self.val = Useful_Work(self, EV, 0) self.thresh = (self.val > 20) except ZeroDivisionError: handle_error_metric(self, "Useful_Work zero division") desc = """ Total pipeline cost of \"useful operations\" - the portion of Retiring category not covered by Branching_Overhead nor Irregular_Overhead.""" class Metric_Core_Bound_Likely: name = "Core_Bound_Likely" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Botlnk.L0" metricgroup = frozenset(['Cor', 'SMT']) sibling = None def compute(self, EV): try: self.val = Core_Bound_Likely(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Core_Bound_Likely zero division") desc = """ Probability of Core Bound bottleneck hidden by SMT-profiling artifacts. Tip: consider analysis with SMT disabled""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Slots_Utilization: name = "Slots_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Slots_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Slots_Utilization zero division") desc = """ Fraction of Physical Core issue-slots utilized by this Logical Processor""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_FP_Arith_Utilization: name = "FP_Arith_Utilization" domain = "Core_Metric" maxval = 2.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = FP_Arith_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FP_Arith_Utilization zero division") desc = """ Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add use all of ADD/MUL/FMA in Scalar or 128/256-bit vectors - less common.""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_EPC: name = "EPC" domain = "Metric" maxval = 20.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = EPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "EPC zero division") desc = """ uops Executed per Cycle""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpFLOP: name = "IpFLOP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpFLOP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpFLOP zero division") desc = """ Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_IpArith_Scalar_SP: name = "IpArith_Scalar_SP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_SP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_SP zero division") desc = """ Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_DP: name = "IpArith_Scalar_DP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_DP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_DP zero division") desc = """ Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX128: name = "IpArith_AVX128" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX128(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX128 zero division") desc = """ Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX256: name = "IpArith_AVX256" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX256(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX256 zero division") desc = """ Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpPause: name = "IpPause" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpPause(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpPause zero division") desc = """ Instructions per PAUSE (lower number means higher occurrence rate)""" class Metric_IpSWPF: name = "IpSWPF" domain = "Inst_Metric" maxval = 1000 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Prefetches']) sibling = None def compute(self, EV): try: self.val = IpSWPF(self, EV, 0) self.thresh = (self.val < 100) except ZeroDivisionError: handle_error_metric(self, "IpSWPF zero division") desc = """ Instructions per Software prefetch instruction (of any type: NTA/T0/T1/T2/Prefetch) (lower number means higher occurrence rate)""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_Strings_Cycles: name = "Strings_Cycles" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['MicroSeq', 'Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Strings_Cycles(self, EV, 0) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error_metric(self, "Strings_Cycles zero division") desc = """ Estimated fraction of retirement-cycles dealing with repeat instructions""" class Metric_IpAssist: name = "IpAssist" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['MicroSeq', 'Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = IpAssist(self, EV, 0) self.thresh = (self.val < 100000) except ZeroDivisionError: handle_error_metric(self, "IpAssist zero division") desc = """ Instructions per a microcode Assist invocation. See Assists tree node for details (lower number means higher occurrence rate)""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_Fetch_LSD: name = "Fetch_LSD" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_LSD(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_LSD zero division") desc = """ Average number of uops fetched from LSD per cycle""" class Metric_Fetch_DSB: name = "Fetch_DSB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_DSB(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_DSB zero division") desc = """ Average number of uops fetched from DSB per cycle""" class Metric_Fetch_MITE: name = "Fetch_MITE" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_MITE(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_MITE zero division") desc = """ Average number of uops fetched from MITE per cycle""" class Metric_Fetch_UpC: name = "Fetch_UpC" domain = "Metric" maxval = 6.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = Fetch_UpC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Fetch_UpC zero division") desc = """ Average number of Uops issued by front-end when it issued something""" class Metric_LSD_Coverage: name = "LSD_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'LSD']) sibling = None def compute(self, EV): try: self.val = LSD_Coverage(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "LSD_Coverage zero division") desc = """ Fraction of Uops delivered by the LSD (Loop Stream Detector; aka Loop Cache)""" class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_Unknown_Branch_Cost: name = "Unknown_Branch_Cost" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = Unknown_Branch_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Unknown_Branch_Cost zero division") desc = """ Average number of cycles the front-end was delayed due to an Unknown Branch detection. See Unknown_Branches node.""" class Metric_DSB_Switch_Cost: name = "DSB_Switch_Cost" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss']) sibling = None def compute(self, EV): try: self.val = DSB_Switch_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DSB_Switch_Cost zero division") desc = """ Average number of cycles of a switch from the DSB fetch-unit to MITE fetch unit - see DSB_Switches tree node for details.""" class Metric_DSB_Misses: name = "DSB_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = DSB_Misses(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Misses zero division") desc = """ Total pipeline cost of DSB (uop cache) misses - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_DSB_Bandwidth: name = "DSB_Bandwidth" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['DSB', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Bandwidth(self, EV, 0) self.thresh = (self.val > 10) except ZeroDivisionError: handle_error_metric(self, "DSB_Bandwidth zero division") desc = """ Total pipeline cost of DSB (uop cache) hits - subset of the Instruction_Fetch_BW Bottleneck.""" class Metric_ICache_Miss_Latency: name = "ICache_Miss_Latency" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = ICache_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ICache_Miss_Latency zero division") desc = """ Average Latency for L1 instruction cache misses""" class Metric_IC_Misses: name = "IC_Misses" domain = "Scaled_Slots" maxval = 0 errcount = 0 area = "Info.Botlnk.L2" metricgroup = frozenset(['Fed', 'FetchLat', 'IcMiss']) sibling = None def compute(self, EV): try: self.val = IC_Misses(self, EV, 0) self.thresh = (self.val > 5) except ZeroDivisionError: handle_error_metric(self, "IC_Misses zero division") desc = """ Total pipeline cost of Instruction Cache misses - subset of the Big_Code Bottleneck.""" class Metric_IpDSB_Miss_Ret: name = "IpDSB_Miss_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSBmiss', 'Fed']) sibling = None def compute(self, EV): try: self.val = IpDSB_Miss_Ret(self, EV, 0) self.thresh = (self.val < 50) except ZeroDivisionError: handle_error_metric(self, "IpDSB_Miss_Ret zero division") desc = """ Instructions per non-speculative DSB miss (lower number means higher occurrence rate)""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_L2MPKI_Code: name = "L2MPKI_Code" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code zero division") desc = """ L2 cache true code cacheline misses per kilo instruction""" class Metric_L2MPKI_Code_All: name = "L2MPKI_Code_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['IcMiss']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Code_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Code_All zero division") desc = """ L2 cache speculative code cacheline misses per kilo instruction""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Cond_Ntaken: name = "IpMisp_Cond_Ntaken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Ntaken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Ntaken zero division") desc = """ Instructions per retired Mispredicts for conditional non- taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Cond_Taken: name = "IpMisp_Cond_Taken" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Cond_Taken(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Cond_Taken zero division") desc = """ Instructions per retired Mispredicts for conditional taken branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Ret: name = "IpMisp_Ret" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Ret(self, EV, 0) self.thresh = (self.val < 500) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Ret zero division") desc = """ Instructions per retired Mispredicts for return branches (lower number means higher occurrence rate).""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Branch_Misprediction_Cost: name = "Branch_Misprediction_Cost" domain = "Core_Metric" maxval = 300 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Branch_Misprediction_Cost(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Branch_Misprediction_Cost zero division") desc = """ Branch Misprediction Cost: Fraction of TMA slots wasted per non-speculative branch misprediction (retired JEClear)""" class Metric_Spec_Clears_Ratio: name = "Spec_Clears_Ratio" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['BrMispredicts']) sibling = None def compute(self, EV): try: self.val = Spec_Clears_Ratio(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Spec_Clears_Ratio zero division") desc = """ Speculative to Retired ratio of all clears (covering Mispredicts and nukes)""" class Metric_Cond_NT: name = "Cond_NT" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_NT(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_NT zero division") desc = """ Fraction of branches that are non-taken conditionals""" class Metric_Cond_TK: name = "Cond_TK" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches', 'CodeGen', 'PGO']) sibling = None def compute(self, EV): try: self.val = Cond_TK(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Cond_TK zero division") desc = """ Fraction of branches that are taken conditionals""" class Metric_CallRet: name = "CallRet" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = CallRet(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CallRet zero division") desc = """ Fraction of branches that are CALL or RET""" class Metric_Jump: name = "Jump" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Jump(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Jump zero division") desc = """ Fraction of branches that are unconditional (direct or indirect) jumps""" class Metric_Other_Branches: name = "Other_Branches" domain = "Fraction" maxval = 1.0 errcount = 0 area = "Info.Branches" metricgroup = frozenset(['Bad', 'Branches']) sibling = None def compute(self, EV): try: self.val = Other_Branches(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Other_Branches zero division") desc = """ Fraction of branches of other types (not individually covered by other metrics in Info.Branches group)""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L1MPKI_Load: name = "L1MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI_Load zero division") desc = """ L1 cache true misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_All: name = "L2MPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_All zero division") desc = """ L2 cache true misses per kilo instruction for all request types (including speculative)""" class Metric_L2MPKI_Load: name = "L2MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Load zero division") desc = """ L2 cache true misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L2HPKI_All: name = "L2HPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_All zero division") desc = """ L2 cache hits per kilo instruction for all request types (including speculative)""" class Metric_L2HPKI_Load: name = "L2HPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_Load zero division") desc = """ L2 cache hits per kilo instruction for all demand loads (including speculative)""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_FB_HPKI: name = "FB_HPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = FB_HPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FB_HPKI zero division") desc = """ Fill Buffer (FB) hits per kilo instructions for retired demand loads (L1D misses that merge into ongoing miss- handling entries)""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Access_BW: name = "L3_Cache_Access_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_Code_STLB_MPKI: name = "Code_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Fed', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Code_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Code_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) code speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Load_STLB_MPKI: name = "Load_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Load_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data load speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_Store_STLB_MPKI: name = "Store_STLB_MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Store_STLB_MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Store_STLB_MPKI zero division") desc = """ STLB (2nd level TLB) data store speculative misses per kilo instruction (misses of any page-size that complete the page walk)""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_L3_Cache_Access_BW_2T: name = "L3_Cache_Access_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Access_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Access_BW_2T zero division") desc = """ Average per-core data access bandwidth to the L3 cache [GB / sec]""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L3_Miss_Latency: name = "Load_L3_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L3_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L3_Miss_Latency zero division") desc = """ Average Latency for L3 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_UC_Load_PKI: name = "UC_Load_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = UC_Load_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "UC_Load_PKI zero division") desc = """ Un-cacheable retired load per kilo instruction""" class Metric_Bus_Lock_PKI: name = "Bus_Lock_PKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory.Mix" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = Bus_Lock_PKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Bus_Lock_PKI zero division") desc = """ \"Bus lock\" per kilo instruction""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_C0_Wait: name = "C0_Wait" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['C0Wait']) sibling = None def compute(self, EV): try: self.val = C0_Wait(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "C0_Wait zero division") desc = """ Fraction of cycles the processor is waiting yet unhalted; covering legacy PAUSE instruction, as well as C0.1 / C0.2 power-performance optimized states. Sample code of TPAUSE: h ttps://github.com/torvalds/linux/blob/master/arch/x86/lib/de lay.c""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_MEM_Parallel_Reads: name = "MEM_Parallel_Reads" domain = "SystemMetric" maxval = 100 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Parallel_Reads(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Parallel_Reads zero division") desc = """ Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches""" class Metric_Power: name = "Power" domain = "SystemMetric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power', 'SoC']) sibling = None def compute(self, EV): try: self.val = Power(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Power zero division") desc = """ Total package Power in Watts""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = Mispredicts_Resteers() ; r.run(n) ; o["Mispredicts_Resteers"] = n n = Clears_Resteers() ; r.run(n) ; o["Clears_Resteers"] = n n = Unknown_Branches() ; r.run(n) ; o["Unknown_Branches"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = Decoder0_Alone() ; r.run(n) ; o["Decoder0_Alone"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = LSD() ; r.run(n) ; o["LSD"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Other_Mispredicts() ; r.run(n) ; o["Other_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Other_Nukes() ; r.run(n) ; o["Other_Nukes"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Load_STLB_Hit() ; r.run(n) ; o["Load_STLB_Hit"] = n n = Load_STLB_Miss() ; r.run(n) ; o["Load_STLB_Miss"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = L1_Hit_Latency() ; r.run(n) ; o["L1_Hit_Latency"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = Streaming_Stores() ; r.run(n) ; o["Streaming_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Store_STLB_Hit() ; r.run(n) ; o["Store_STLB_Hit"] = n n = Store_STLB_Miss() ; r.run(n) ; o["Store_STLB_Miss"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Serializing_Operation() ; r.run(n) ; o["Serializing_Operation"] = n n = Slow_Pause() ; r.run(n) ; o["Slow_Pause"] = n n = C01_WAIT() ; r.run(n) ; o["C01_WAIT"] = n n = C02_WAIT() ; r.run(n) ; o["C02_WAIT"] = n n = Memory_Fence() ; r.run(n) ; o["Memory_Fence"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Mixing_Vectors() ; r.run(n) ; o["Mixing_Vectors"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_6() ; r.run(n) ; o["Port_6"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = Int_Operations() ; r.run(n) ; o["Int_Operations"] = n n = Int_Vector_128b() ; r.run(n) ; o["Int_Vector_128b"] = n n = Int_Vector_256b() ; r.run(n) ; o["Int_Vector_256b"] = n n = Memory_Operations() ; r.run(n) ; o["Memory_Operations"] = n n = Fused_Instructions() ; r.run(n) ; o["Fused_Instructions"] = n n = Non_Fused_Branches() ; r.run(n) ; o["Non_Fused_Branches"] = n n = Other_Light_Ops() ; r.run(n) ; o["Other_Light_Ops"] = n n = Nop_Instructions() ; r.run(n) ; o["Nop_Instructions"] = n n = Shuffles_256b() ; r.run(n) ; o["Shuffles_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Few_Uops_Instructions() ; r.run(n) ; o["Few_Uops_Instructions"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = Page_Faults() ; r.run(n) ; o["Page_Faults"] = n n = FP_Assists() ; r.run(n) ; o["FP_Assists"] = n n = AVX_Assists() ; r.run(n) ; o["AVX_Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["Mispredicts_Resteers"].parent = o["Branch_Resteers"] o["Clears_Resteers"].parent = o["Branch_Resteers"] o["Unknown_Branches"].parent = o["Branch_Resteers"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["Decoder0_Alone"].parent = o["MITE"] o["DSB"].parent = o["Fetch_Bandwidth"] o["LSD"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Other_Mispredicts"].parent = o["Branch_Mispredicts"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Other_Nukes"].parent = o["Machine_Clears"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Load_STLB_Hit"].parent = o["DTLB_Load"] o["Load_STLB_Miss"].parent = o["DTLB_Load"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["L1_Hit_Latency"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["Streaming_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Store_STLB_Hit"].parent = o["DTLB_Store"] o["Store_STLB_Miss"].parent = o["DTLB_Store"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Serializing_Operation"].parent = o["Core_Bound"] o["Slow_Pause"].parent = o["Serializing_Operation"] o["C01_WAIT"].parent = o["Serializing_Operation"] o["C02_WAIT"].parent = o["Serializing_Operation"] o["Memory_Fence"].parent = o["Serializing_Operation"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Mixing_Vectors"].parent = o["Ports_Utilized_0"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_6"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["Int_Operations"].parent = o["Light_Operations"] o["Int_Vector_128b"].parent = o["Int_Operations"] o["Int_Vector_256b"].parent = o["Int_Operations"] o["Memory_Operations"].parent = o["Light_Operations"] o["Fused_Instructions"].parent = o["Light_Operations"] o["Non_Fused_Branches"].parent = o["Light_Operations"] o["Other_Light_Ops"].parent = o["Light_Operations"] o["Nop_Instructions"].parent = o["Other_Light_Ops"] o["Shuffles_256b"].parent = o["Other_Light_Ops"] o["Heavy_Operations"].parent = o["Retiring"] o["Few_Uops_Instructions"].parent = o["Heavy_Operations"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["Page_Faults"].parent = o["Assists"] o["FP_Assists"].parent = o["Assists"] o["AVX_Assists"].parent = o["Assists"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_Mispredictions() ; r.metric(n) ; o["Mispredictions"] = n n = Metric_Big_Code() ; r.metric(n) ; o["Big_Code"] = n n = Metric_Instruction_Fetch_BW() ; r.metric(n) ; o["Instruction_Fetch_BW"] = n n = Metric_Cache_Memory_Bandwidth() ; r.metric(n) ; o["Cache_Memory_Bandwidth"] = n n = Metric_Cache_Memory_Latency() ; r.metric(n) ; o["Cache_Memory_Latency"] = n n = Metric_Memory_Data_TLBs() ; r.metric(n) ; o["Memory_Data_TLBs"] = n n = Metric_Memory_Synchronization() ; r.metric(n) ; o["Memory_Synchronization"] = n n = Metric_Compute_Bound_Est() ; r.metric(n) ; o["Compute_Bound_Est"] = n n = Metric_Irregular_Overhead() ; r.metric(n) ; o["Irregular_Overhead"] = n n = Metric_Other_Bottlenecks() ; r.metric(n) ; o["Other_Bottlenecks"] = n n = Metric_Branching_Overhead() ; r.metric(n) ; o["Branching_Overhead"] = n n = Metric_Useful_Work() ; r.metric(n) ; o["Useful_Work"] = n n = Metric_Core_Bound_Likely() ; r.metric(n) ; o["Core_Bound_Likely"] = n n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Slots_Utilization() ; r.metric(n) ; o["Slots_Utilization"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_FP_Arith_Utilization() ; r.metric(n) ; o["FP_Arith_Utilization"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_EPC() ; r.metric(n) ; o["EPC"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpFLOP() ; r.metric(n) ; o["IpFLOP"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_IpArith_Scalar_SP() ; r.metric(n) ; o["IpArith_Scalar_SP"] = n n = Metric_IpArith_Scalar_DP() ; r.metric(n) ; o["IpArith_Scalar_DP"] = n n = Metric_IpArith_AVX128() ; r.metric(n) ; o["IpArith_AVX128"] = n n = Metric_IpArith_AVX256() ; r.metric(n) ; o["IpArith_AVX256"] = n n = Metric_IpPause() ; r.metric(n) ; o["IpPause"] = n n = Metric_IpSWPF() ; r.metric(n) ; o["IpSWPF"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_Strings_Cycles() ; r.metric(n) ; o["Strings_Cycles"] = n n = Metric_IpAssist() ; r.metric(n) ; o["IpAssist"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_Fetch_LSD() ; r.metric(n) ; o["Fetch_LSD"] = n n = Metric_Fetch_DSB() ; r.metric(n) ; o["Fetch_DSB"] = n n = Metric_Fetch_MITE() ; r.metric(n) ; o["Fetch_MITE"] = n n = Metric_Fetch_UpC() ; r.metric(n) ; o["Fetch_UpC"] = n n = Metric_LSD_Coverage() ; r.metric(n) ; o["LSD_Coverage"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_Unknown_Branch_Cost() ; r.metric(n) ; o["Unknown_Branch_Cost"] = n n = Metric_DSB_Switch_Cost() ; r.metric(n) ; o["DSB_Switch_Cost"] = n n = Metric_DSB_Misses() ; r.metric(n) ; o["DSB_Misses"] = n n = Metric_DSB_Bandwidth() ; r.metric(n) ; o["DSB_Bandwidth"] = n n = Metric_ICache_Miss_Latency() ; r.metric(n) ; o["ICache_Miss_Latency"] = n n = Metric_IC_Misses() ; r.metric(n) ; o["IC_Misses"] = n n = Metric_IpDSB_Miss_Ret() ; r.metric(n) ; o["IpDSB_Miss_Ret"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_L2MPKI_Code() ; r.metric(n) ; o["L2MPKI_Code"] = n n = Metric_L2MPKI_Code_All() ; r.metric(n) ; o["L2MPKI_Code_All"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Cond_Ntaken() ; r.metric(n) ; o["IpMisp_Cond_Ntaken"] = n n = Metric_IpMisp_Cond_Taken() ; r.metric(n) ; o["IpMisp_Cond_Taken"] = n n = Metric_IpMisp_Ret() ; r.metric(n) ; o["IpMisp_Ret"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Branch_Misprediction_Cost() ; r.metric(n) ; o["Branch_Misprediction_Cost"] = n n = Metric_Spec_Clears_Ratio() ; r.metric(n) ; o["Spec_Clears_Ratio"] = n n = Metric_Cond_NT() ; r.metric(n) ; o["Cond_NT"] = n n = Metric_Cond_TK() ; r.metric(n) ; o["Cond_TK"] = n n = Metric_CallRet() ; r.metric(n) ; o["CallRet"] = n n = Metric_Jump() ; r.metric(n) ; o["Jump"] = n n = Metric_Other_Branches() ; r.metric(n) ; o["Other_Branches"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L1MPKI_Load() ; r.metric(n) ; o["L1MPKI_Load"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_All() ; r.metric(n) ; o["L2MPKI_All"] = n n = Metric_L2MPKI_Load() ; r.metric(n) ; o["L2MPKI_Load"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L2HPKI_All() ; r.metric(n) ; o["L2HPKI_All"] = n n = Metric_L2HPKI_Load() ; r.metric(n) ; o["L2HPKI_Load"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_FB_HPKI() ; r.metric(n) ; o["FB_HPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_L3_Cache_Access_BW() ; r.metric(n) ; o["L3_Cache_Access_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_Code_STLB_MPKI() ; r.metric(n) ; o["Code_STLB_MPKI"] = n n = Metric_Load_STLB_MPKI() ; r.metric(n) ; o["Load_STLB_MPKI"] = n n = Metric_Store_STLB_MPKI() ; r.metric(n) ; o["Store_STLB_MPKI"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Access_BW_2T() ; r.metric(n) ; o["L3_Cache_Access_BW_2T"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L3_Miss_Latency() ; r.metric(n) ; o["Load_L3_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_UC_Load_PKI() ; r.metric(n) ; o["UC_Load_PKI"] = n n = Metric_Bus_Lock_PKI() ; r.metric(n) ; o["Bus_Lock_PKI"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_C0_Wait() ; r.metric(n) ; o["C0_Wait"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_MEM_Parallel_Reads() ; r.metric(n) ; o["MEM_Parallel_Reads"] = n n = Metric_Power() ; r.metric(n) ; o["Power"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Branch_Resteers"].Unknown_Branches = o["Unknown_Branches"] o["Mispredicts_Resteers"].Retiring = o["Retiring"] o["Mispredicts_Resteers"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Mispredicts_Resteers"].Bad_Speculation = o["Bad_Speculation"] o["Mispredicts_Resteers"].Frontend_Bound = o["Frontend_Bound"] o["Mispredicts_Resteers"].Backend_Bound = o["Backend_Bound"] o["Clears_Resteers"].Retiring = o["Retiring"] o["Clears_Resteers"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Clears_Resteers"].Bad_Speculation = o["Bad_Speculation"] o["Clears_Resteers"].Frontend_Bound = o["Frontend_Bound"] o["Clears_Resteers"].Backend_Bound = o["Backend_Bound"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Bad_Speculation"].Retiring = o["Retiring"] o["Bad_Speculation"].Frontend_Bound = o["Frontend_Bound"] o["Bad_Speculation"].Backend_Bound = o["Backend_Bound"] o["Other_Mispredicts"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Retiring = o["Retiring"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Frontend_Bound = o["Frontend_Bound"] o["Machine_Clears"].Backend_Bound = o["Backend_Bound"] o["Other_Nukes"].Machine_Clears = o["Machine_Clears"] o["Other_Nukes"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Nukes"].Retiring = o["Retiring"] o["Other_Nukes"].Backend_Bound = o["Backend_Bound"] o["Other_Nukes"].Bad_Speculation = o["Bad_Speculation"] o["Other_Nukes"].Frontend_Bound = o["Frontend_Bound"] o["DTLB_Load"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Load_STLB_Hit"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Load_STLB_Hit"].DTLB_Load = o["DTLB_Load"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["DTLB_Store"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Store_STLB_Hit"].DTLB_Store = o["DTLB_Store"] o["Store_STLB_Hit"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Serializing_Operation"].C02_WAIT = o["C02_WAIT"] o["Ports_Utilization"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Ports_Utilization"].Retiring = o["Retiring"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["FP_Arith"].Retiring = o["Retiring"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["X87_Use"].Retiring = o["Retiring"] o["FP_Scalar"].Retiring = o["Retiring"] o["FP_Vector"].Retiring = o["Retiring"] o["FP_Vector_128b"].Retiring = o["Retiring"] o["FP_Vector_256b"].Retiring = o["Retiring"] o["Int_Operations"].Retiring = o["Retiring"] o["Int_Operations"].Int_Vector_256b = o["Int_Vector_256b"] o["Int_Operations"].Int_Vector_128b = o["Int_Vector_128b"] o["Int_Vector_128b"].Retiring = o["Retiring"] o["Int_Vector_256b"].Retiring = o["Retiring"] o["Memory_Operations"].Retiring = o["Retiring"] o["Memory_Operations"].Light_Operations = o["Light_Operations"] o["Memory_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Fused_Instructions"].Retiring = o["Retiring"] o["Fused_Instructions"].Light_Operations = o["Light_Operations"] o["Fused_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Non_Fused_Branches"].Retiring = o["Retiring"] o["Non_Fused_Branches"].Light_Operations = o["Light_Operations"] o["Non_Fused_Branches"].Heavy_Operations = o["Heavy_Operations"] o["Other_Light_Ops"].Light_Operations = o["Light_Operations"] o["Other_Light_Ops"].Retiring = o["Retiring"] o["Other_Light_Ops"].Heavy_Operations = o["Heavy_Operations"] o["Other_Light_Ops"].Int_Operations = o["Int_Operations"] o["Other_Light_Ops"].Non_Fused_Branches = o["Non_Fused_Branches"] o["Other_Light_Ops"].FP_Arith = o["FP_Arith"] o["Other_Light_Ops"].Fused_Instructions = o["Fused_Instructions"] o["Other_Light_Ops"].Int_Vector_128b = o["Int_Vector_128b"] o["Other_Light_Ops"].FP_Vector = o["FP_Vector"] o["Other_Light_Ops"].FP_Scalar = o["FP_Scalar"] o["Other_Light_Ops"].X87_Use = o["X87_Use"] o["Other_Light_Ops"].Int_Vector_256b = o["Int_Vector_256b"] o["Other_Light_Ops"].Memory_Operations = o["Memory_Operations"] o["Nop_Instructions"].Retiring = o["Retiring"] o["Nop_Instructions"].Light_Operations = o["Light_Operations"] o["Nop_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Shuffles_256b"].Retiring = o["Retiring"] o["Shuffles_256b"].Light_Operations = o["Light_Operations"] o["Shuffles_256b"].Heavy_Operations = o["Heavy_Operations"] o["Few_Uops_Instructions"].Heavy_Operations = o["Heavy_Operations"] o["Few_Uops_Instructions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] o["Mispredictions"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Mispredictions"].LCP = o["LCP"] o["Mispredictions"].Retiring = o["Retiring"] o["Mispredictions"].Other_Mispredicts = o["Other_Mispredicts"] o["Mispredictions"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Mispredictions"].Frontend_Bound = o["Frontend_Bound"] o["Mispredictions"].DSB_Switches = o["DSB_Switches"] o["Mispredictions"].Backend_Bound = o["Backend_Bound"] o["Mispredictions"].Branch_Resteers = o["Branch_Resteers"] o["Mispredictions"].ICache_Misses = o["ICache_Misses"] o["Mispredictions"].MS_Switches = o["MS_Switches"] o["Mispredictions"].Bad_Speculation = o["Bad_Speculation"] o["Mispredictions"].ITLB_Misses = o["ITLB_Misses"] o["Mispredictions"].Unknown_Branches = o["Unknown_Branches"] o["Mispredictions"].Fetch_Latency = o["Fetch_Latency"] o["Mispredictions"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Big_Code"].LCP = o["LCP"] o["Big_Code"].ICache_Misses = o["ICache_Misses"] o["Big_Code"].DSB_Switches = o["DSB_Switches"] o["Big_Code"].Branch_Resteers = o["Branch_Resteers"] o["Big_Code"].MS_Switches = o["MS_Switches"] o["Big_Code"].ITLB_Misses = o["ITLB_Misses"] o["Big_Code"].Unknown_Branches = o["Unknown_Branches"] o["Big_Code"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].Retiring = o["Retiring"] o["Instruction_Fetch_BW"].Other_Mispredicts = o["Other_Mispredicts"] o["Instruction_Fetch_BW"].DSB_Switches = o["DSB_Switches"] o["Instruction_Fetch_BW"].Backend_Bound = o["Backend_Bound"] o["Instruction_Fetch_BW"].Branch_Resteers = o["Branch_Resteers"] o["Instruction_Fetch_BW"].Fetch_Latency = o["Fetch_Latency"] o["Instruction_Fetch_BW"].ICache_Misses = o["ICache_Misses"] o["Instruction_Fetch_BW"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Instruction_Fetch_BW"].Frontend_Bound = o["Frontend_Bound"] o["Instruction_Fetch_BW"].Bad_Speculation = o["Bad_Speculation"] o["Instruction_Fetch_BW"].ITLB_Misses = o["ITLB_Misses"] o["Instruction_Fetch_BW"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Instruction_Fetch_BW"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Instruction_Fetch_BW"].LCP = o["LCP"] o["Instruction_Fetch_BW"].Clears_Resteers = o["Clears_Resteers"] o["Instruction_Fetch_BW"].MS_Switches = o["MS_Switches"] o["Instruction_Fetch_BW"].Unknown_Branches = o["Unknown_Branches"] o["Cache_Memory_Bandwidth"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Bandwidth"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Bandwidth"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Bandwidth"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Bandwidth"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Bandwidth"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Cache_Memory_Bandwidth"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Bandwidth"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Bandwidth"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Bandwidth"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Bandwidth"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Bandwidth"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Bandwidth"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Bandwidth"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Bandwidth"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Bandwidth"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Bandwidth"].FB_Full = o["FB_Full"] o["Cache_Memory_Bandwidth"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Bandwidth"].DRAM_Bound = o["DRAM_Bound"] o["Cache_Memory_Latency"].L1_Bound = o["L1_Bound"] o["Cache_Memory_Latency"].DTLB_Load = o["DTLB_Load"] o["Cache_Memory_Latency"].False_Sharing = o["False_Sharing"] o["Cache_Memory_Latency"].Data_Sharing = o["Data_Sharing"] o["Cache_Memory_Latency"].L2_Bound = o["L2_Bound"] o["Cache_Memory_Latency"].Memory_Bound = o["Memory_Bound"] o["Cache_Memory_Latency"].DTLB_Store = o["DTLB_Store"] o["Cache_Memory_Latency"].SQ_Full = o["SQ_Full"] o["Cache_Memory_Latency"].Store_Bound = o["Store_Bound"] o["Cache_Memory_Latency"].Split_Loads = o["Split_Loads"] o["Cache_Memory_Latency"].L3_Bound = o["L3_Bound"] o["Cache_Memory_Latency"].FB_Full = o["FB_Full"] o["Cache_Memory_Latency"].Streaming_Stores = o["Streaming_Stores"] o["Cache_Memory_Latency"].Contested_Accesses = o["Contested_Accesses"] o["Cache_Memory_Latency"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Cache_Memory_Latency"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Cache_Memory_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Cache_Memory_Latency"].Store_Latency = o["Store_Latency"] o["Cache_Memory_Latency"].Split_Stores = o["Split_Stores"] o["Cache_Memory_Latency"].Lock_Latency = o["Lock_Latency"] o["Cache_Memory_Latency"].MEM_Latency = o["MEM_Latency"] o["Cache_Memory_Latency"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Cache_Memory_Latency"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Cache_Memory_Latency"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Cache_Memory_Latency"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Data_TLBs"].L1_Bound = o["L1_Bound"] o["Memory_Data_TLBs"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Memory_Data_TLBs"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Memory_Data_TLBs"].DTLB_Load = o["DTLB_Load"] o["Memory_Data_TLBs"].Store_Latency = o["Store_Latency"] o["Memory_Data_TLBs"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Memory_Data_TLBs"].False_Sharing = o["False_Sharing"] o["Memory_Data_TLBs"].DTLB_Store = o["DTLB_Store"] o["Memory_Data_TLBs"].L2_Bound = o["L2_Bound"] o["Memory_Data_TLBs"].Memory_Bound = o["Memory_Bound"] o["Memory_Data_TLBs"].Lock_Latency = o["Lock_Latency"] o["Memory_Data_TLBs"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Memory_Data_TLBs"].Store_Bound = o["Store_Bound"] o["Memory_Data_TLBs"].Split_Stores = o["Split_Stores"] o["Memory_Data_TLBs"].Split_Loads = o["Split_Loads"] o["Memory_Data_TLBs"].L3_Bound = o["L3_Bound"] o["Memory_Data_TLBs"].FB_Full = o["FB_Full"] o["Memory_Data_TLBs"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Data_TLBs"].DRAM_Bound = o["DRAM_Bound"] o["Memory_Synchronization"].L1_Bound = o["L1_Bound"] o["Memory_Synchronization"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Synchronization"].False_Sharing = o["False_Sharing"] o["Memory_Synchronization"].Retiring = o["Retiring"] o["Memory_Synchronization"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Synchronization"].Machine_Clears = o["Machine_Clears"] o["Memory_Synchronization"].Data_Sharing = o["Data_Sharing"] o["Memory_Synchronization"].Memory_Bound = o["Memory_Bound"] o["Memory_Synchronization"].SQ_Full = o["SQ_Full"] o["Memory_Synchronization"].Store_Bound = o["Store_Bound"] o["Memory_Synchronization"].L3_Bound = o["L3_Bound"] o["Memory_Synchronization"].L2_Bound = o["L2_Bound"] o["Memory_Synchronization"].Streaming_Stores = o["Streaming_Stores"] o["Memory_Synchronization"].Contested_Accesses = o["Contested_Accesses"] o["Memory_Synchronization"].DTLB_Store = o["DTLB_Store"] o["Memory_Synchronization"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Memory_Synchronization"].Store_Latency = o["Store_Latency"] o["Memory_Synchronization"].Split_Stores = o["Split_Stores"] o["Memory_Synchronization"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Memory_Synchronization"].Backend_Bound = o["Backend_Bound"] o["Memory_Synchronization"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Memory_Synchronization"].Other_Nukes = o["Other_Nukes"] o["Memory_Synchronization"].DRAM_Bound = o["DRAM_Bound"] o["Compute_Bound_Est"].Serializing_Operation = o["Serializing_Operation"] o["Compute_Bound_Est"].Ports_Utilization = o["Ports_Utilization"] o["Compute_Bound_Est"].C02_WAIT = o["C02_WAIT"] o["Compute_Bound_Est"].Retiring = o["Retiring"] o["Compute_Bound_Est"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Compute_Bound_Est"].Memory_Bound = o["Memory_Bound"] o["Compute_Bound_Est"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Compute_Bound_Est"].Core_Bound = o["Core_Bound"] o["Compute_Bound_Est"].Backend_Bound = o["Backend_Bound"] o["Compute_Bound_Est"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Compute_Bound_Est"].Divider = o["Divider"] o["Compute_Bound_Est"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Heavy_Operations = o["Heavy_Operations"] o["Irregular_Overhead"].Ports_Utilization = o["Ports_Utilization"] o["Irregular_Overhead"].C02_WAIT = o["C02_WAIT"] o["Irregular_Overhead"].Retiring = o["Retiring"] o["Irregular_Overhead"].ICache_Misses = o["ICache_Misses"] o["Irregular_Overhead"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Irregular_Overhead"].Frontend_Bound = o["Frontend_Bound"] o["Irregular_Overhead"].Serializing_Operation = o["Serializing_Operation"] o["Irregular_Overhead"].Core_Bound = o["Core_Bound"] o["Irregular_Overhead"].Bad_Speculation = o["Bad_Speculation"] o["Irregular_Overhead"].ITLB_Misses = o["ITLB_Misses"] o["Irregular_Overhead"].Divider = o["Divider"] o["Irregular_Overhead"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Irregular_Overhead"].Memory_Bound = o["Memory_Bound"] o["Irregular_Overhead"].Machine_Clears = o["Machine_Clears"] o["Irregular_Overhead"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Irregular_Overhead"].LCP = o["LCP"] o["Irregular_Overhead"].Other_Mispredicts = o["Other_Mispredicts"] o["Irregular_Overhead"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Irregular_Overhead"].DSB_Switches = o["DSB_Switches"] o["Irregular_Overhead"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Irregular_Overhead"].Assists = o["Assists"] o["Irregular_Overhead"].Backend_Bound = o["Backend_Bound"] o["Irregular_Overhead"].Branch_Resteers = o["Branch_Resteers"] o["Irregular_Overhead"].Clears_Resteers = o["Clears_Resteers"] o["Irregular_Overhead"].MS_Switches = o["MS_Switches"] o["Irregular_Overhead"].Other_Nukes = o["Other_Nukes"] o["Irregular_Overhead"].Unknown_Branches = o["Unknown_Branches"] o["Irregular_Overhead"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].L1_Bound = o["L1_Bound"] o["Other_Bottlenecks"].C02_WAIT = o["C02_WAIT"] o["Other_Bottlenecks"].Retiring = o["Retiring"] o["Other_Bottlenecks"].Data_Sharing = o["Data_Sharing"] o["Other_Bottlenecks"].L2_Bound = o["L2_Bound"] o["Other_Bottlenecks"].Core_Bound = o["Core_Bound"] o["Other_Bottlenecks"].Ports_Utilization = o["Ports_Utilization"] o["Other_Bottlenecks"].Contested_Accesses = o["Contested_Accesses"] o["Other_Bottlenecks"].Divider = o["Divider"] o["Other_Bottlenecks"].L3_Bound = o["L3_Bound"] o["Other_Bottlenecks"].Ports_Utilized_3m = o["Ports_Utilized_3m"] o["Other_Bottlenecks"].L1_Hit_Latency = o["L1_Hit_Latency"] o["Other_Bottlenecks"].FB_Full = o["FB_Full"] o["Other_Bottlenecks"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Other_Bottlenecks"].Store_Latency = o["Store_Latency"] o["Other_Bottlenecks"].Other_Mispredicts = o["Other_Mispredicts"] o["Other_Bottlenecks"].DSB_Switches = o["DSB_Switches"] o["Other_Bottlenecks"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Other_Bottlenecks"].Ports_Utilized_1 = o["Ports_Utilized_1"] o["Other_Bottlenecks"].Ports_Utilized_2 = o["Ports_Utilized_2"] o["Other_Bottlenecks"].Assists = o["Assists"] o["Other_Bottlenecks"].Backend_Bound = o["Backend_Bound"] o["Other_Bottlenecks"].Branch_Resteers = o["Branch_Resteers"] o["Other_Bottlenecks"].Store_STLB_Miss = o["Store_STLB_Miss"] o["Other_Bottlenecks"].L3_Hit_Latency = o["L3_Hit_Latency"] o["Other_Bottlenecks"].Split_Stores = o["Split_Stores"] o["Other_Bottlenecks"].Heavy_Operations = o["Heavy_Operations"] o["Other_Bottlenecks"].Fetch_Latency = o["Fetch_Latency"] o["Other_Bottlenecks"].DTLB_Load = o["DTLB_Load"] o["Other_Bottlenecks"].False_Sharing = o["False_Sharing"] o["Other_Bottlenecks"].ICache_Misses = o["ICache_Misses"] o["Other_Bottlenecks"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Other_Bottlenecks"].Frontend_Bound = o["Frontend_Bound"] o["Other_Bottlenecks"].Machine_Clears = o["Machine_Clears"] o["Other_Bottlenecks"].Streaming_Stores = o["Streaming_Stores"] o["Other_Bottlenecks"].Memory_Bound = o["Memory_Bound"] o["Other_Bottlenecks"].SQ_Full = o["SQ_Full"] o["Other_Bottlenecks"].Store_Bound = o["Store_Bound"] o["Other_Bottlenecks"].Split_Loads = o["Split_Loads"] o["Other_Bottlenecks"].Bad_Speculation = o["Bad_Speculation"] o["Other_Bottlenecks"].ITLB_Misses = o["ITLB_Misses"] o["Other_Bottlenecks"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Other_Bottlenecks"].Store_Fwd_Blk = o["Store_Fwd_Blk"] o["Other_Bottlenecks"].Serializing_Operation = o["Serializing_Operation"] o["Other_Bottlenecks"].DTLB_Store = o["DTLB_Store"] o["Other_Bottlenecks"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Other_Bottlenecks"].LCP = o["LCP"] o["Other_Bottlenecks"].Load_STLB_Miss = o["Load_STLB_Miss"] o["Other_Bottlenecks"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Other_Bottlenecks"].Lock_Latency = o["Lock_Latency"] o["Other_Bottlenecks"].MEM_Latency = o["MEM_Latency"] o["Other_Bottlenecks"].Clears_Resteers = o["Clears_Resteers"] o["Other_Bottlenecks"].MS_Switches = o["MS_Switches"] o["Other_Bottlenecks"].Other_Nukes = o["Other_Nukes"] o["Other_Bottlenecks"].Unknown_Branches = o["Unknown_Branches"] o["Other_Bottlenecks"].DRAM_Bound = o["DRAM_Bound"] o["Useful_Work"].Retiring = o["Retiring"] o["Useful_Work"].Heavy_Operations = o["Heavy_Operations"] o["Useful_Work"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Useful_Work"].Few_Uops_Instructions = o["Few_Uops_Instructions"] o["Useful_Work"].Assists = o["Assists"] o["Core_Bound_Likely"].Memory_Bound = o["Memory_Bound"] o["Core_Bound_Likely"].Ports_Utilized_0 = o["Ports_Utilized_0"] o["Core_Bound_Likely"].Core_Bound = o["Core_Bound"] o["Core_Bound_Likely"].Ports_Utilization = o["Ports_Utilization"] o["Core_Bound_Likely"].Retiring = o["Retiring"] o["Core_Bound_Likely"].Backend_Bound = o["Backend_Bound"] o["UopPI"].Retiring = o["Retiring"] o["UpTB"].Retiring = o["Retiring"] o["Retire"].Retiring = o["Retiring"] o["DSB_Misses"].LSD = o["LSD"] o["DSB_Misses"].MITE = o["MITE"] o["DSB_Misses"].LCP = o["LCP"] o["DSB_Misses"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Misses"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Misses"].DSB_Switches = o["DSB_Switches"] o["DSB_Misses"].Branch_Resteers = o["Branch_Resteers"] o["DSB_Misses"].ICache_Misses = o["ICache_Misses"] o["DSB_Misses"].MS_Switches = o["MS_Switches"] o["DSB_Misses"].ITLB_Misses = o["ITLB_Misses"] o["DSB_Misses"].DSB = o["DSB"] o["DSB_Misses"].Unknown_Branches = o["Unknown_Branches"] o["DSB_Misses"].Fetch_Latency = o["Fetch_Latency"] o["DSB_Bandwidth"].LSD = o["LSD"] o["DSB_Bandwidth"].Fetch_Bandwidth = o["Fetch_Bandwidth"] o["DSB_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["DSB_Bandwidth"].MITE = o["MITE"] o["DSB_Bandwidth"].DSB = o["DSB"] o["DSB_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].Fetch_Latency = o["Fetch_Latency"] o["IC_Misses"].LCP = o["LCP"] o["IC_Misses"].MS_Switches = o["MS_Switches"] o["IC_Misses"].ICache_Misses = o["ICache_Misses"] o["IC_Misses"].ITLB_Misses = o["ITLB_Misses"] o["IC_Misses"].Unknown_Branches = o["Unknown_Branches"] o["IC_Misses"].DSB_Switches = o["DSB_Switches"] o["IC_Misses"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].Retiring = o["Retiring"] o["Branch_Misprediction_Cost"].ICache_Misses = o["ICache_Misses"] o["Branch_Misprediction_Cost"].Microcode_Sequencer = o["Microcode_Sequencer"] o["Branch_Misprediction_Cost"].Frontend_Bound = o["Frontend_Bound"] o["Branch_Misprediction_Cost"].Bad_Speculation = o["Bad_Speculation"] o["Branch_Misprediction_Cost"].ITLB_Misses = o["ITLB_Misses"] o["Branch_Misprediction_Cost"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Branch_Misprediction_Cost"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Branch_Misprediction_Cost"].LCP = o["LCP"] o["Branch_Misprediction_Cost"].Other_Mispredicts = o["Other_Mispredicts"] o["Branch_Misprediction_Cost"].DSB_Switches = o["DSB_Switches"] o["Branch_Misprediction_Cost"].Backend_Bound = o["Backend_Bound"] o["Branch_Misprediction_Cost"].Branch_Resteers = o["Branch_Resteers"] o["Branch_Misprediction_Cost"].MS_Switches = o["MS_Switches"] o["Branch_Misprediction_Cost"].Unknown_Branches = o["Unknown_Branches"] o["Branch_Misprediction_Cost"].Fetch_Latency = o["Fetch_Latency"] # siblings cross-tree o["Mispredicts_Resteers"].sibling = (o["Branch_Mispredicts"],) o["Clears_Resteers"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["MS_Switches"].sibling = (o["Clears_Resteers"], o["Machine_Clears"], o["L1_Bound"], o["Serializing_Operation"], o["Mixing_Vectors"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Decoder0_Alone"].sibling = (o["Few_Uops_Instructions"],) o["Branch_Mispredicts"].sibling = (o["Mispredicts_Resteers"],) o["Machine_Clears"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"], o["Streaming_Stores"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"],) o["Streaming_Stores"].sibling = (o["FB_Full"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Serializing_Operation"].sibling = (o["MS_Switches"],) o["Mixing_Vectors"].sibling = (o["MS_Switches"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Port_6"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["Int_Vector_128b"], o["Int_Vector_256b"],) o["Int_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_256b"],) o["Int_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"], o["Int_Vector_128b"],) o["Few_Uops_Instructions"].sibling = (o["Decoder0_Alone"],) o["Microcode_Sequencer"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["Mispredictions"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["Cache_Memory_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],) o["Cache_Memory_Latency"].sibling = (o["L3_Hit_Latency"], o["MEM_Latency"],) o["Memory_Data_TLBs"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Memory_Synchronization"].sibling = (o["DTLB_Load"], o["DTLB_Store"],) o["Irregular_Overhead"].sibling = (o["MS_Switches"], o["Microcode_Sequencer"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Misses"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["Branch_Misprediction_Cost"].sibling = (o["Mispredicts_Resteers"], o["Branch_Mispredicts"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
244,538
Python
.py
5,655
37.270203
1,765
0.655489
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,891
event_download.py
andikleen_pmu-tools/event_download.py
#!/usr/bin/env python3 # Copyright (c) 2014-2020, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. # # Automatic event list downloader # # event_download.py download for current cpu # event_download.py -a download all # event_download.py cpustr... Download for specific CPU # # env: # CPUINFO=... override /proc/cpuinfo file # MAPFILE=... override mapfile.csv # PERFMONDIR=... override download prefix for perfmon data, can be a local clone (file:///tmp/perfmon) from __future__ import print_function import sys import re try: from urllib.request import urlopen from urllib.error import URLError except ImportError: from urllib2 import urlopen, URLError import os import string from fnmatch import fnmatch urlpath = os.environ.get('PERFMONDIR', 'https://raw.githubusercontent.com/intel/perfmon/main') mapfile = 'mapfile.csv' modelpath = urlpath + "/" + mapfile def get_cpustr(): cpuinfo = os.getenv("CPUINFO") if cpuinfo is None: cpuinfo = '/proc/cpuinfo' f = open(cpuinfo, 'r') cpu = [None, None, None, None] for j in f: n = j.split() if n[0] == 'vendor_id': cpu[0] = n[2] elif n[0] == 'model' and n[1] == ':': cpu[2] = int(n[2]) elif n[0] == 'cpu' and n[1] == 'family': cpu[1] = int(n[3]) elif n[0] == 'stepping' and n[1] == ':': cpu[3] = int(n[2]) if all(v is not None for v in cpu): break # stepping for SKX only stepping = cpu[0] == "GenuineIntel" and cpu[1] == 6 and cpu[2] == 0x55 if stepping: return "%s-%d-%X-%X" % tuple(cpu) return "%s-%d-%X" % tuple(cpu)[:3] def sanitize(s, a): o = "" for j in s: if j in a: o += j return o def getdir(): try: d = os.getenv("XDG_CACHE_HOME") xd = d if not d: home = os.getenv("HOME") d = "%s/.cache" % (home) d += "/pmu-events" if not os.path.isdir(d): # try to handle the sudo case if not xd: user = os.getenv("SUDO_USER") if user: nd = os.path.expanduser("~" + user) + "/.cache/pmu-events" if os.path.isdir(nd): return nd os.makedirs(d) return d except OSError: raise Exception('Cannot access ' + d) NUM_TRIES = 3 def getfile(url, dirfn, fn): tries = 0 print("Downloading", url, "to", fn) while True: try: f = urlopen(url) data = f.read() except IOError: tries += 1 if tries >= NUM_TRIES: raise print("retrying download") continue break o = open(os.path.join(dirfn, fn), "wb") o.write(data) o.close() f.close() printed = set() def warn_once(msg): if msg not in printed: print(msg, file=sys.stderr) printed.add(msg) def cpu_without_step(match): if match.count("-") < 3: return match n = match.split("-") return "%s-%s-%s" % tuple(n[:3]) allowed_chars = string.ascii_letters + '_-.' + string.digits def parse_map_file(match, key=None, link=True, onlyprint=False, acceptfile=False, hybridkey=None): match2 = cpu_without_step(match) files = [] dirfn = getdir() try: mfn = os.getenv("MAPFILE") if mfn: mapfn = mfn acceptfile = True else: mapfn = os.path.join(dirfn, mapfile) if onlyprint and not os.path.exists(mapfn) and not mfn: print("Download", mapfn, "first for --print") return [] if acceptfile and os.path.exists(mapfn): pass elif not onlyprint and not mfn: getfile(modelpath, dirfn, mapfile) models = open(mapfn) for j in models: if j.startswith("Family-model"): continue n = j.rstrip().split(",") if len(n) < 4: if len(n) > 0: print("Cannot parse", n) continue cpu, version, name, typ = n[:4] if not (fnmatch(cpu, match) or fnmatch(cpu, match2) or fnmatch(match2, cpu) or fnmatch(match, cpu)): continue if key is not None and typ not in key: continue if hybridkey and len(n) >= 7 and n[6] != hybridkey: continue cpu = sanitize(cpu, allowed_chars) url = urlpath + name matchfn = match if matchfn == "*": matchfn = cpu if ".json" not in matchfn: if hybridkey: fn = "%s-%s-%s.json" % (matchfn, sanitize(typ, allowed_chars), hybridkey) else: fn = "%s-%s.json" % (matchfn, sanitize(typ, allowed_chars)) path = os.path.join(dirfn, fn) if acceptfile and os.path.exists(path): if onlyprint: print(path) continue else: try: os.remove(path) except OSError: pass if onlyprint: print(path) continue if mfn: print("error accessing", path) continue try: fn = fn.replace("01234", "4") fn = fn.replace("56789ABCDEF", "5") # XXX getfile(url, dirfn, fn) except URLError as e: print("error accessing %s: %s" % (url, e), file=sys.stderr) if match == '*': continue raise if link: lname = re.sub(r'.*/', '', name) lname = sanitize(lname, allowed_chars) try: os.remove(os.path.join(dirfn, lname)) except OSError: pass try: os.symlink(fn, os.path.join(dirfn, lname)) except OSError as e: print("Cannot link %s to %s:" % (name, lname), e, file=sys.stderr) files.append(fn) models.close() for file_name in ["README.md", "LICENSE"]: if not onlyprint and not os.path.exists(os.path.join(dirfn, file_name)) and not mfn: getfile(urlpath + "/" + file_name, dirfn, file_name) except URLError as e: print("Cannot access event server:", e, file=sys.stderr) warn_once(""" If you need a proxy to access the internet please set it with: \texport https_proxy=http://proxyname... If you are not connected to the internet please run this on a connected system: \tevent_download.py '%s' and then copy ~/.cache/pmu-events to the system under test To get events for all possible CPUs use: \tevent_download.py -a""" % match) except OSError as e: print("Cannot write events file:", e, file=sys.stderr) return files def download(match, key=None, link=True, onlyprint=False, acceptfile=False): return len(parse_map_file(match, key, link, onlyprint, acceptfile)) def download_current(link=False, onlyprint=False): """Download JSON event list for current cpu. Returns >0 when a event list is found""" return download(get_cpustr(), link=link, onlyprint=onlyprint, ) def eventlist_name(name=None, key="core", hybridkey=None): if not name: name = get_cpustr() cache = getdir() fn = name if os.path.exists(fn): return fn if ".json" not in name: if hybridkey: fn = "%s-%s-%s.json" % (name, key, hybridkey) else: fn = "%s-%s.json" % (name, key) if "/" not in fn: fn = "%s/%s" % (cache, fn) if not os.path.exists(fn): files = parse_map_file(name, key, acceptfile=True, hybridkey=hybridkey) if files: return files[0] name = cpu_without_step(name) if "*" in fn: fn = "%s/%s" % (cache, name) elif hybridkey: fn = "%s/%s-%s-%s.json" % (cache, name, key, hybridkey) else: fn = "%s/%s-%s.json" % (cache, name, key) files = parse_map_file(name, key, acceptfile=True, hybridkey=hybridkey) if files: fn = files[0] return fn if __name__ == '__main__': # only import argparse when actually called from command line # this makes ocperf work on older python versions without it. import argparse p = argparse.ArgumentParser(usage='download Intel event files') p.add_argument('--all', '-a', help='Download all available event files', action='store_true') p.add_argument('--verbose', '-v', help='Be verbose', action='store_true') p.add_argument('--mine', help='Print name of current CPU', action='store_true') p.add_argument('--link', help='Create links with the original event file name', action='store_true', default=True) p.add_argument('--print', help='Print file names of all event files instead of downloading. Requires existing mapfile.csv.', dest='print_', action='store_true') p.add_argument('cpus', help='CPU identifiers to download', nargs='*') args = p.parse_args() if args.verbose or args.mine: print(get_cpustr()) if args.mine: sys.exit(0) d = getdir() if args.all: found = download('*', link=args.link, onlyprint=args.print_) elif len(args.cpus) == 0: found = download_current(link=args.link, onlyprint=args.print_) else: found = 0 for j in args.cpus: found += download(j, link=args.link, onlyprint=args.print_) if found == 0 and not args.print_: print("Nothing found", file=sys.stderr) el = eventlist_name() if os.path.exists(el) and not args.print_: print("my event list", el)
10,437
Python
.py
282
27.79078
128
0.558591
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,892
power_metrics.py
andikleen_pmu-tools/power_metrics.py
# # perf power metrics for toplev # import os class EnergyPackage: name = "Package Energy" desc = """ Package Energy over measurement period in Joules""" unit = "Joules" nogroup = True subplot = "Power" domain = "Package" def compute(self, EV): self.val = EV("power/energy-pkg/", 1) self.thresh = self.val > 0 class EnergyCores: name = "Cores Energy" desc = """ Cores Energy over measurement period in Joules""" unit = "Joules" nogroup = True subplot = "Power" domain = "Package" def compute(self, EV): self.val = EV("power/energy-cores/", 1) self.thresh = self.val > 0 class EnergyRAM: name = "RAM Energy" desc = """ RAM Energy over measurement period in Joules""" unit = "Joules" nogroup = True subplot = "Power" domain = "Package" def compute(self, EV): self.val = EV("power/energy-ram/", 1) self.thresh = self.val > 0 class EnergyGPU: name = "GPU Energy" desc = """ GPU Energy over measurement period in Joules""" unit = "Joules" nogroup = True subplot = "Power" domain = "Package" def compute(self, EV): self.val = EV("power/energy-gpu/", 1) self.thresh = self.val > 1 class Setup: def __init__(self, r): if os.path.exists("/sys/devices/power/events/energy-cores"): r.force_metric(EnergyCores()) r.force_metric(EnergyPackage()) if os.path.exists("/sys/devices/power/events/energy-ram"): r.force_metric(EnergyRAM()) if os.path.exists("/sys/devices/power/events/energy-gpu"): r.force_metric(EnergyGPU())
1,661
Python
.py
57
23.684211
68
0.62015
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,893
bdw_client_ratios.py
andikleen_pmu-tools/bdw_client_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel 5th gen Core / Core M (code named Broadwell) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 8 Mem_L2_Store_Cost = 9 Mem_L3_Weight = 7 Mem_STLB_Hit_Cost = 8 BAClear_Cost = 12 MS_Switches_Cost = 2 Avg_Assist_Cost = 66 Pipeline_Width = 4 OneMillion = 1000000 OneBillion = 1000000000 Energy_Unit = 61 Errata_Whitelist = "BDE69;BDE70" EBS_Mode = 0 DS = 0 # Aux. formulas def Backend_Bound_Cycles(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_TOTAL", level) + EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - Few_Uops_Executed_Threshold(self, EV, level) - Frontend_RS_Empty_Cycles(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) def Cycles_0_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:i1:c1", level)) / 2 if smt_enabled else(EV("CYCLE_ACTIVITY.STALLS_TOTAL", level) - Frontend_RS_Empty_Cycles(self, EV, level)) def Cycles_1_Port_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) - EV("UOPS_EXECUTED.CORE:c2", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level)) def Cycles_2_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c2", level) - EV("UOPS_EXECUTED.CORE:c3", level)) / 2 if smt_enabled else(EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) - EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level)) def Cycles_3m_Ports_Utilized(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c3", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_EXECUTED.CORE:c1", level) / 2) if smt_enabled else EV("UOPS_EXECUTED.CYCLES_GE_1_UOP_EXEC", level) def Fetched_Uops(self, EV, level): return (EV("IDQ.DSB_UOPS", level) + EV("LSD.UOPS", level) + EV("IDQ.MITE_UOPS", level) + EV("IDQ.MS_UOPS", level)) def Few_Uops_Executed_Threshold(self, EV, level): EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) return EV("UOPS_EXECUTED.CYCLES_GE_3_UOPS_EXEC", level) if (IPC(self, EV, level)> 1.8) else EV("UOPS_EXECUTED.CYCLES_GE_2_UOPS_EXEC", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) + 2 * EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + 4 * EV("FP_ARITH_INST_RETIRED.4_FLOPS", level) + 8 * EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_ARITH_INST_RETIRED.SCALAR", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_ARITH_INST_RETIRED.VECTOR", level) def Frontend_RS_Empty_Cycles(self, EV, level): EV("RS_EVENTS.EMPTY_CYCLES", level) return EV("RS_EVENTS.EMPTY_CYCLES", level) if (self.Fetch_Latency.compute(EV)> 0.1) else 0 def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def ITLB_Miss_Cycles(self, EV, level): return (14 * EV("ITLB_MISSES.STLB_HIT", level) + EV("ITLB_MISSES.WALK_DURATION:c1", level) + 7 * EV("ITLB_MISSES.WALK_COMPLETED", level)) def LOAD_L1_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L2_HIT", level) + EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM", level) + EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS", level) def LOAD_L1_MISS_NET(self, EV, level): return LOAD_L1_MISS(self, EV, level) + EV("MEM_LOAD_UOPS_RETIRED.L3_MISS", level) def LOAD_L3_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HIT(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_HITM(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def LOAD_XSNP_MISS(self, EV, level): return EV("MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS", level) * (1 + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level) / LOAD_L1_MISS_NET(self, EV, level)) def Mem_L3_Hit_Fraction(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) / (EV("MEM_LOAD_UOPS_RETIRED.L3_HIT", level) + Mem_L3_Weight * EV("MEM_LOAD_UOPS_RETIRED.L3_MISS", level)) def Mem_Lock_St_Fraction(self, EV, level): return EV("MEM_UOPS_RETIRED.LOCK_LOADS", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) def Memory_Bound_Fraction(self, EV, level): return (EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", level) + EV("RESOURCE_STALLS.SB", level)) / Backend_Bound_Cycles(self, EV, level) def Mispred_Clears_Fraction(self, EV, level): return EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) def ORO_Demand_RFO_C1(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_RFO", level)) , level ) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c4", level)) , level ) def SQ_Full_Cycles(self, EV, level): return (EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) / 2) if smt_enabled else EV("OFFCORE_REQUESTS_BUFFER.SQ_FULL", level) def Store_L2_Hit_Cycles(self, EV, level): return EV("L2_RQSTS.RFO_HIT", level) * Mem_L2_Store_Cost *(1 - Mem_Lock_St_Fraction(self, EV, level)) def Mem_XSNP_HitM_Cost(self, EV, level): return 60 def Mem_XSNP_Hit_Cost(self, EV, level): return 43 def Mem_XSNP_None_Cost(self, EV, level): return 29 def Recovery_Cycles(self, EV, level): return (EV("INT_MISC.RECOVERY_CYCLES_ANY", level) / 2) if smt_enabled else EV("INT_MISC.RECOVERY_CYCLES", level) def Retire_Fraction(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return EV("UOPS_RETIRED.RETIRE_SLOTS", level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return 8 if smt_enabled else 4 # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Uops per taken branch def UpTB(self, EV, level): val = Retired_Slots(self, EV, level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 1.5 return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return Pipeline_Width * CORE_CLKS(self, EV, level) # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common. def FP_Arith_Utilization(self, EV, level): return (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) / (2 * CORE_CLKS(self, EV, level)) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / EV("UOPS_EXECUTED.THREAD:c1", level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return ((EV("CPU_CLK_UNHALTED.THREAD", level) / 2) * (1 + EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_XCLK", level))) if ebs_mode else(EV("CPU_CLK_UNHALTED.THREAD_ANY", level) / 2) if smt_enabled else CLKS(self, EV, level) # Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpLoad(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_LOADS", level) self.thresh = (val < 3) return val # Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses. #Link Opt Guide section: Minimize Register Spills def IpStore(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("MEM_UOPS_RETIRED.ALL_STORES", level) self.thresh = (val < 8) return val # Instructions per Branch (lower number means higher occurrence rate) def IpBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 8) return val # Instructions per (near) call (lower number means higher occurrence rate) def IpCall(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_CALL", level) self.thresh = (val < 200) return val # Instructions per taken branch def IpTB(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) self.thresh = val < Pipeline_Width * 2 + 1 return val # Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes. def BpTkBranch(self, EV, level): return EV("BR_INST_RETIRED.ALL_BRANCHES", level) / EV("BR_INST_RETIRED.NEAR_TAKEN", level) # Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408 def IpFLOP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / FLOP_Count(self, EV, level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW. def IpArith(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (FP_Arith_Scalar(self, EV, level) + FP_Arith_Vector(self, EV, level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_SP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_SINGLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_Scalar_DP(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("FP_ARITH_INST_RETIRED.SCALAR_DOUBLE", level) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX128(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. def IpArith_AVX256(self, EV, level): val = EV("INST_RETIRED.ANY", level) / (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", level) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", level)) self.thresh = (val < 10) return val # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.RETIRE_SLOTS:c1", level) def Execute(self, EV, level): return EV("UOPS_EXECUTED.THREAD", level) / Execute_Cycles(self, EV, level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate) def IpUnknown_Branch(self, EV, level): return Instructions(self, EV, level) / EV("BACLEARS.ANY", level) # Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate) def IpMispredict(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_MISP_RETIRED.ALL_BRANCHES", level) self.thresh = (val < 200) return val # Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate). def IpMisp_Indirect(self, EV, level): val = Instructions(self, EV, level) / (Retire_Fraction(self, EV, level) * EV("BR_MISP_EXEC.INDIRECT", level)) self.thresh = (val < 1000) return val # Actual Average Latency for L1 data-cache miss demand load operations (in core cycles) def Load_Miss_Real_Latency(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / (EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) + EV("MEM_LOAD_UOPS_RETIRED.HIT_LFB", level)) # Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor) def MLP(self, EV, level): return EV("L1D_PEND_MISS.PENDING", level) / EV("L1D_PEND_MISS.PENDING_CYCLES", level) # L1 cache true misses per kilo instruction for retired demand loads def L1MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L1_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache true misses per kilo instruction for retired demand loads def L2MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L2_MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache misses per kilo instruction for all request types (including speculative) def L2MPKI_All(self, EV, level): return 1000 * EV("L2_RQSTS.MISS", level) / EV("INST_RETIRED.ANY", level) # L2 cache misses per kilo instruction for all demand loads (including speculative) def L2MPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_MISS", level) / EV("INST_RETIRED.ANY", level) # Offcore requests (L2 cache miss) per kilo instruction for demand RFOs def L2MPKI_RFO(self, EV, level): return 1000 * EV("OFFCORE_REQUESTS.DEMAND_RFO", level) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all request types (including speculative) def L2HPKI_All(self, EV, level): return 1000 *(EV("L2_RQSTS.REFERENCES", level) - EV("L2_RQSTS.MISS", level)) / EV("INST_RETIRED.ANY", level) # L2 cache hits per kilo instruction for all demand loads (including speculative) def L2HPKI_Load(self, EV, level): return 1000 * EV("L2_RQSTS.DEMAND_DATA_RD_HIT", level) / EV("INST_RETIRED.ANY", level) # L3 cache true misses per kilo instruction for retired demand loads def L3MPKI(self, EV, level): return 1000 * EV("MEM_LOAD_UOPS_RETIRED.L3_MISS", level) / EV("INST_RETIRED.ANY", level) def L1D_Cache_Fill_BW(self, EV, level): return 64 * EV("L1D.REPLACEMENT", level) / OneBillion / Time(self, EV, level) def L2_Cache_Fill_BW(self, EV, level): return 64 * EV("L2_LINES_IN.ALL", level) / OneBillion / Time(self, EV, level) def L3_Cache_Fill_BW(self, EV, level): return 64 * EV("LONGEST_LAT_CACHE.MISS", level) / OneBillion / Time(self, EV, level) # Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses def Page_Walks_Utilization(self, EV, level): val = (EV("ITLB_MISSES.WALK_DURATION:c1", level) + EV("DTLB_LOAD_MISSES.WALK_DURATION:c1", level) + EV("DTLB_STORE_MISSES.WALK_DURATION:c1", level) + 7 *(EV("DTLB_STORE_MISSES.WALK_COMPLETED", level) + EV("DTLB_LOAD_MISSES.WALK_COMPLETED", level) + EV("ITLB_MISSES.WALK_COMPLETED", level))) / CORE_CLKS(self, EV, level) self.thresh = (val > 0.5) return val # Average per-core data fill bandwidth to the L1 data cache [GB / sec] def L1D_Cache_Fill_BW_2T(self, EV, level): return L1D_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L2 cache [GB / sec] def L2_Cache_Fill_BW_2T(self, EV, level): return L2_Cache_Fill_BW(self, EV, level) # Average per-core data fill bandwidth to the L3 cache [GB / sec] def L3_Cache_Fill_BW_2T(self, EV, level): return L3_Cache_Fill_BW(self, EV, level) # Average Latency for L2 cache miss demand Loads def Load_L2_Miss_Latency(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS.DEMAND_DATA_RD", level) # Average Parallel L2 cache miss demand Loads def Load_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.DEMAND_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DEMAND_DATA_RD", level) # Average Parallel L2 cache miss data reads def Data_L2_MLP(self, EV, level): return EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD", level) / EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level) # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / (EV("CPU_CLK_UNHALTED.REF_XCLK_ANY", level) / 2) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return 64 *(EV("UNC_ARB_TRK_REQUESTS.ALL", level) + EV("UNC_ARB_COH_TRK_REQUESTS.ALL", level)) / OneMillion / Time(self, EV, level) / 1000 # Total package Power in Watts def Power(self, EV, level): return EV("UNC_PKG_ENERGY_STATUS", level) * Energy_Unit /(Time(self, EV, level) * OneMillion ) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_CLOCK.SOCKET", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['RS_EVENTS.EMPTY_END'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Pipeline_Width * EV("IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE", 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ICache_Misses: name = "ICache_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'IcMiss']) maxval = None def compute(self, EV): try: self.val = EV("ICACHE.IFDATA_STALL", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ICache_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to instruction cache misses.. Using compiler's Profile-Guided Optimization (PGO) can reduce i-cache misses through improved hot code layout.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['ITLB_MISSES.WALK_COMPLETED'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = ITLB_Miss_Cycles(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = BAClear_Cost *(EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) + EV("MACHINE_CLEARS.COUNT", 3) + EV("BACLEARS.ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class Mispredicts_Resteers: name = "Mispredicts_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP']) maxval = None def compute(self, EV): try: self.val = EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) * self.Branch_Resteers.compute(EV) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) + EV("MACHINE_CLEARS.COUNT", 4) + EV("BACLEARS.ANY", 4)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Mispredicts_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Branch Misprediction at execution stage.""" class Clears_Resteers: name = "Clears_Resteers" domain = "Clocks" area = "FE" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'MachineClears']) maxval = None def compute(self, EV): try: self.val = EV("MACHINE_CLEARS.COUNT", 4) * self.Branch_Resteers.compute(EV) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", 4) + EV("MACHINE_CLEARS.COUNT", 4) + EV("BACLEARS.ANY", 4)) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Clears_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers as a result of Machine Clears.""" class Unknown_Branches: name = "Unknown_Branches" domain = "Clocks" area = "FE" level = 4 htoff = False sample = ['BACLEARS.ANY'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = self.Branch_Resteers.compute(EV) - self.Mispredicts_Resteers.compute(EV) - self.Clears_Resteers.compute(EV) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Unknown_Branches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to new branch address clears. These are fetched branches the Branch Prediction Unit was unable to recognize (e.g. first time the branch is fetched or hitting BTB capacity limit) hence called Unknown Branches""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['IDQ.MS_SWITCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("IDQ.MS_SWITCHES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("ILD_STALL.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class MITE: name = "MITE" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_MITE_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_MITE_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MITE zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to the MITE pipeline (the legacy decode pipeline). This pipeline is used for code that was not pre-cached in the DSB or LSD. For example; inefficiencies due to asymmetric decoders; use of long immediate or LCP can manifest as MITE fetch bandwidth bottleneck.. Consider tuning codegen of 'small hotspots' that can fit in DSB. Read about 'Decoded ICache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class DSB: name = "DSB" domain = "Slots_Estimated" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSB', 'FetchBW']) maxval = None def compute(self, EV): try: self.val = (EV("IDQ.ALL_DSB_CYCLES_ANY_UOPS", 3) - EV("IDQ.ALL_DSB_CYCLES_4_UOPS", 3)) / CORE_CLKS(self, EV, 3) / 2 self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB zero division") return self.val desc = """ This metric represents Core fraction of cycles in which CPU was likely limited due to DSB (decoded uop cache) fetch pipeline. For example; inefficient utilization of the DSB cache structure or bank conflict when reading from it; are categorized here.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_ISSUED.ANY", 1) - Retired_Slots(self, EV, 1) + Pipeline_Width * Recovery_Cycles(self, EV, 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 2) * self.Bad_Speculation.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = 1 -(self.Frontend_Bound.compute(EV) + self.Bad_Speculation.compute(EV) + self.Retiring.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Memory_Bound_Fraction(self, EV, 2) * self.Backend_Bound.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class L1_Bound: name = "L1_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L1_HIT:pp', 'MEM_LOAD_UOPS_RETIRED.HIT_LFB:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = max((EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3) - EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", 3)) / CLKS(self, EV, 3) , 0 ) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L1_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled without loads missing the L1 data cache. The L1 data cache typically has the shortest latency. However; in certain cases like loads blocked on older stores; a load might suffer due to high latency even though it is being satisfied by the L1. Another example is loads who miss in the TLB. These cases are characterized by execution unit stalls; while some non-completed demand load lives in the machine without having that demand load missing the L1 cache.""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT", 4) + EV("DTLB_LOAD_MISSES.WALK_DURATION:c1", 4) + 7 * EV("DTLB_LOAD_MISSES.WALK_COMPLETED", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class Store_Fwd_Blk: name = "Store_Fwd_Blk" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = 13 * EV("LD_BLOCKS.STORE_FORWARD", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Fwd_Blk zero division") return self.val desc = """ This metric roughly estimates fraction of cycles when the memory subsystem had loads blocked since they could not forward data from earlier (in program order) overlapping stores. To streamline memory operations in the pipeline; a load can avoid waiting for memory if a prior in-flight store is writing the data that the load wants to read (store forwarding process). However; in some cases the load may be blocked for a significant time pending the store forward. For example; when the prior store is writing a smaller region than the load is reading.""" class Lock_Latency: name = "Lock_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.LOCK_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_Lock_St_Fraction(self, EV, 4) * ORO_Demand_RFO_C1(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Lock_Latency zero division") return self.val desc = """ This metric represents fraction of cycles the CPU spent handling cache misses due to lock operations. Due to the microarchitecture handling of locks; they are classified as L1_Bound regardless of what memory source satisfied them.""" class Split_Loads: name = "Split_Loads" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = 1.0 def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("LD_BLOCKS.NO_SR", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Loads zero division") return self.val desc = """ This metric estimates fraction of cycles handling memory load split accesses - load that cross 64-byte cache line boundary. . Consider aligning data or hot structure fields. See the Optimization Manual for more details""" class G4K_Aliasing: name = "4K_Aliasing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("LD_BLOCKS_PARTIAL.ADDRESS_ALIAS", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "G4K_Aliasing zero division") return self.val desc = """ This metric estimates how often memory load accesses were aliased by preceding stores (in program order) with a 4K address offset. False match is possible; which incur a few cycles load re-issue. However; the short re-issue duration is often hidden by the out-of-order core and HW optimizations; hence a user may safely ignore a high value of this metric unless it manages to propagate up into parent nodes of the hierarchy (e.g. to L1_Bound).. Consider reducing independent loads/stores accesses with 4K offsets. See the Optimization Manual for more details""" class FB_Full: name = "FB_Full" domain = "Clocks_Calculated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW']) maxval = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 4) * EV("L1D_PEND_MISS.FB_FULL:c1", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.3) except ZeroDivisionError: handle_error(self, "FB_Full zero division") return self.val desc = """ This metric does a *rough estimation* of how often L1D Fill Buffer unavailability limited additional L1D miss memory access requests to proceed. The higher the metric value; the deeper the memory hierarchy level the misses are satisfied from (metric values >1 are valid). Often it hints on approaching bandwidth limits (to L2 cache; L3 cache or external memory).. See $issueBW and $issueSL hints. Avoid software prefetches if indeed memory BW limited.""" class L2_Bound: name = "L2_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L2_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = (EV("CYCLE_ACTIVITY.STALLS_L1D_MISS", 3) - EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L2_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to L2 cache accesses by loads. Avoiding cache misses (i.e. L1 misses/L2 hits) can improve the latency and increase performance.""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = Mem_L3_Hit_Fraction(self, EV, 3) * EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class Contested_Accesses: name = "Contested_Accesses" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM:pp', 'MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_XSNP_HitM_Cost(self, EV, 4) * LOAD_XSNP_HITM(self, EV, 4) + Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_MISS(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Contested_Accesses zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to contested accesses. Contested accesses occur when data written by one Logical Processor are read by another Logical Processor on a different Physical Core. Examples of contested accesses include synchronizations such as locks; true data sharing such as modified locked variables; and false sharing.""" class Data_Sharing: name = "Data_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_Hit_Cost(self, EV, 4) * LOAD_XSNP_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Data_Sharing zero division") return self.val desc = """ This metric estimates fraction of cycles while the memory subsystem was handling synchronizations due to data-sharing accesses. Data shared by multiple Logical Processors (even just read shared) may cause increased access latency due to cache coherency. Excessive data sharing can drastically harm multithreaded performance.""" class L3_Hit_Latency: name = "L3_Hit_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_None_Cost(self, EV, 4) * LOAD_L3_HIT(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Hit_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles with demand load accesses that hit the L3 cache under unloaded scenarios (possibly L3 latency limited). Avoiding private cache misses (i.e. L2 misses/L3 hits) will improve the latency; reduce contention with sibling physical cores and increase performance. Note the value of this node may overlap with its siblings.""" class SQ_Full: name = "SQ_Full" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = SQ_Full_Cycles(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.3) and self.parent.thresh except ZeroDivisionError: handle_error(self, "SQ_Full zero division") return self.val desc = """ This metric measures fraction of cycles where the Super Queue (SQ) was full taking into account all request-types and both hardware SMT threads (Logical Processors).""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.L3_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = (1 - Mem_L3_Hit_Fraction(self, EV, 3)) * EV("CYCLE_ACTIVITY.STALLS_L2_MISS", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_UOPS_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SB", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Store_Latency: name = "Store_Latency" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = 1.0 def compute(self, EV): try: self.val = (Store_L2_Hit_Cycles(self, EV, 4) + (1 - Mem_Lock_St_Fraction(self, EV, 4)) * ORO_Demand_RFO_C1(self, EV, 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU spent handling L1D store misses. Store accesses usually less impact out-of-order core performance; however; holding resources for longer time can lead into undesired implications (e.g. contention on L1D fill-buffer entries - see FB_Full). Consider to avoid/reduce unnecessary (or easily load-able/computable) memory store.""" class False_Sharing: name = "False_Sharing" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_LOAD_UOPS_L3_HIT_RETIRED.XSNP_HITM:pp', 'OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM'] errcount = 0 sibling = None metricgroup = frozenset(['BvMS', 'DataSharing', 'Offcore', 'Snoop']) maxval = 1.0 def compute(self, EV): try: self.val = Mem_XSNP_HitM_Cost(self, EV, 4) * EV("OFFCORE_RESPONSE.DEMAND_RFO.L3_HIT.SNOOP_HITM", 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "False_Sharing zero division") return self.val desc = """ This metric roughly estimates how often CPU was handling synchronizations due to False Sharing. False Sharing is a multithreading hiccup; where multiple Logical Processors contend on different data-elements mapped into the same cache line. . False Sharing can be easily avoided by padding to make Logical Processors access different lines.""" class Split_Stores: name = "Split_Stores" domain = "Core_Utilization" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.SPLIT_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = 2 * EV("MEM_UOPS_RETIRED.SPLIT_STORES", 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Split_Stores zero division") return self.val desc = """ This metric represents rate of split store accesses. Consider aligning your data to the 64-byte cache line granularity.""" class DTLB_Store: name = "DTLB_Store" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_STORE_MISSES.STLB_HIT", 4) + EV("DTLB_STORE_MISSES.WALK_DURATION:c1", 4) + 7 * EV("DTLB_STORE_MISSES.WALK_COMPLETED", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Store zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles spent handling first-level data TLB store misses. As with ordinary data caching; focus on improving data locality and reducing working-set size to reduce DTLB overhead. Additionally; consider using profile-guided optimization (PGO) to collocate frequently-used data on the same page. Try using larger page sizes for large amounts of frequently- used data.""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.FPU_DIV_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.FPU_DIV_ACTIVE", 3) / CORE_CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = (Backend_Bound_Cycles(self, EV, 3) - EV("RESOURCE_STALLS.SB", 3) - EV("CYCLE_ACTIVITY.STALLS_MEM_ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_0: name = "Ports_Utilized_0" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_0_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_0 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed no uops on any execution port (Logical Processor cycles since ICL, Physical Core cycles otherwise). Long-latency instructions like divides may contribute to this metric.. Check assembly view and Appendix C in Optimization Manual to find out instructions with say 5 or more cycles latency.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Ports_Utilized_1: name = "Ports_Utilized_1" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_1_Port_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_1 zero division") return self.val desc = """ This metric represents fraction of cycles where the CPU executed total of 1 uop per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). This can be due to heavy data-dependency among software instructions; or over oversubscribing a particular hardware resource. In some other cases with high 1_Port_Utilized and L1_Bound; this metric can point to L1 data-cache latency bottleneck that may not necessarily manifest with complete execution starvation (due to the short L1 latency e.g. walking a linked list) - looking at the assembly can be helpful.""" class Ports_Utilized_2: name = "Ports_Utilized_2" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_2_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_2 zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 2 uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise). Loop Vectorization -most compilers feature auto- Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Ports_Utilized_3m: name = "Ports_Utilized_3m" domain = "Clocks" area = "BE/Core" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvCB', 'PortsUtil']) maxval = None def compute(self, EV): try: self.val = Cycles_3m_Ports_Utilized(self, EV, 4) / CORE_CLKS(self, EV, 4) self.thresh = (self.val > 0.4) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilized_3m zero division") return self.val desc = """ This metric represents fraction of cycles CPU executed total of 3 or more uops per cycle on all execution ports (Logical Processor cycles since ICL, Physical Core cycles otherwise).""" class ALU_Op_Utilization: name = "ALU_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_0", 5) + EV("UOPS_DISPATCHED_PORT.PORT_1", 5) + EV("UOPS_DISPATCHED_PORT.PORT_5", 5) + EV("UOPS_DISPATCHED_PORT.PORT_6", 5)) / (4 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.4) except ZeroDivisionError: handle_error(self, "ALU_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution ports for ALU operations.""" class Port_0: name = "Port_0" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_0'] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_0", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_0 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 0 ALU and 2nd branch""" class Port_1: name = "Port_1" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_1'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_1", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_1 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 1 (ALU)""" class Port_5: name = "Port_5" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_5'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_5", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_5 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 5 ALU. See section 'Handling Port 5 Pressure' in Optimization Manual:. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Port_6: name = "Port_6" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_6'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_6", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_6 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 6 Primary Branch and simple ALU""" class Load_Op_Utilization: name = "Load_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_DISPATCHED_PORT.PORT_2", 5) + EV("UOPS_DISPATCHED_PORT.PORT_3", 5) + EV("UOPS_DISPATCHED_PORT.PORT_7", 5) - EV("UOPS_DISPATCHED_PORT.PORT_4", 5)) / (2 * CORE_CLKS(self, EV, 5)) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Load_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Load operations""" class Port_2: name = "Port_2" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_2'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_2", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_2 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 2 Loads and Store-address""" class Port_3: name = "Port_3" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_3'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_3", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_3 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 3 Loads and Store-address""" class Store_Op_Utilization: name = "Store_Op_Utilization" domain = "Core_Execution" area = "BE/Core" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 5) / CORE_CLKS(self, EV, 5) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Store_Op_Utilization zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port for Store operations""" class Port_4: name = "Port_4" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_4'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_4", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_4 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 4 (Store-data)""" class Port_7: name = "Port_7" domain = "Core_Clocks" area = "BE/Core" level = 6 htoff = False sample = ['UOPS_DISPATCHED_PORT.PORT_7'] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = EV("UOPS_DISPATCHED_PORT.PORT_7", 6) / CORE_CLKS(self, EV, 6) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Port_7 zero division") return self.val desc = """ This metric represents Core fraction of cycles CPU dispatched uops on execution port 7 simple Store-address""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.RETIRE_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. . Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = EV("INST_RETIRED.X87", 4) * UopPI(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / Retired_Slots(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.128B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.128B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_ARITH_INST_RETIRED.256B_PACKED_DOUBLE", 5) + EV("FP_ARITH_INST_RETIRED.256B_PACKED_SINGLE", 5)) / Retired_Slots(self, EV, 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Microcode_Sequencer.compute(EV) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['IDQ.MS_UOPS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = Retire_Fraction(self, EV, 3) * EV("IDQ.MS_UOPS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Assists: name = "Assists" domain = "Slots_Estimated" area = "RET" level = 4 htoff = False sample = ['OTHER_ASSISTS.ANY_WB_ASSIST'] errcount = 0 sibling = None metricgroup = frozenset(['BvIO']) maxval = 1.0 def compute(self, EV): try: self.val = Avg_Assist_Cost * EV("OTHER_ASSISTS.ANY_WB_ASSIST", 4) / SLOTS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Assists zero division") return self.val desc = """ This metric estimates fraction of slots the CPU retired uops delivered by the Microcode_Sequencer as a result of Assists. Assists are long sequences of uops that are required in certain corner-cases for operations that cannot be handled natively by the execution pipeline. For example; when working with very small floating point values (so-called Denormals); the FP units are not set up to perform these operations natively. Instead; a sequence of instructions to perform the computation on the Denormals is injected into the pipeline. Since these microcode sequences might be dozens of uops long; Assists can be extremely deleterious to performance and they can be avoided in many cases.""" class CISC: name = "CISC" domain = "Slots" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset([]) maxval = None def compute(self, EV): try: self.val = max(0 , self.Microcode_Sequencer.compute(EV) - self.Assists.compute(EV)) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "CISC zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU retired uops originated from CISC (complex instruction set computer) instruction. A CISC instruction has multiple uops that are required to perform the instruction's functionality as in the case of read-modify-write as an example. Since these instructions require multiple uops they may or may not imply sub-optimal use of machine resources.""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_UpTB: name = "UpTB" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = UpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 1.5 except ZeroDivisionError: handle_error_metric(self, "UpTB zero division") desc = """ Uops per taken branch""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_FP_Arith_Utilization: name = "FP_Arith_Utilization" domain = "Core_Metric" maxval = 2.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = FP_Arith_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FP_Arith_Utilization zero division") desc = """ Actual per-core usage of the Floating Point non-X87 execution units (regardless of precision or vector-width). Values > 1 are possible due to Fused-Multiply Add FMA counting - common.""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_IpLoad: name = "IpLoad" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpLoad(self, EV, 0) self.thresh = (self.val < 3) except ZeroDivisionError: handle_error_metric(self, "IpLoad zero division") desc = """ Instructions per Load (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpStore: name = "IpStore" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['InsType']) sibling = None def compute(self, EV): try: self.val = IpStore(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpStore zero division") desc = """ Instructions per Store (lower number means higher occurrence rate). Tip: reduce memory accesses.""" class Metric_IpBranch: name = "IpBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpBranch(self, EV, 0) self.thresh = (self.val < 8) except ZeroDivisionError: handle_error_metric(self, "IpBranch zero division") desc = """ Instructions per Branch (lower number means higher occurrence rate)""" class Metric_IpCall: name = "IpCall" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpCall(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpCall zero division") desc = """ Instructions per (near) call (lower number means higher occurrence rate)""" class Metric_IpTB: name = "IpTB" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'FetchBW', 'Frontend', 'PGO']) sibling = None def compute(self, EV): try: self.val = IpTB(self, EV, 0) self.thresh = self.val < Pipeline_Width * 2 + 1 except ZeroDivisionError: handle_error_metric(self, "IpTB zero division") desc = """ Instructions per taken branch""" class Metric_BpTkBranch: name = "BpTkBranch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Branches', 'Fed', 'PGO']) sibling = None def compute(self, EV): try: self.val = BpTkBranch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "BpTkBranch zero division") desc = """ Branch instructions per taken branch. . Can be used to approximate PGO-likelihood for non-loopy codes.""" class Metric_IpFLOP: name = "IpFLOP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpFLOP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpFLOP zero division") desc = """ Instructions per Floating Point (FP) Operation (lower number means higher occurrence rate). Reference: Tuning Performance via Metrics with Expectations. https://doi.org/10.1109/LCA.2019.2916408""" class Metric_IpArith: name = "IpArith" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith zero division") desc = """ Instructions per FP Arithmetic instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting. Approximated prior to BDW.""" class Metric_IpArith_Scalar_SP: name = "IpArith_Scalar_SP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_SP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_SP zero division") desc = """ Instructions per FP Arithmetic Scalar Single-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_Scalar_DP: name = "IpArith_Scalar_DP" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpScalar', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_Scalar_DP(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_Scalar_DP zero division") desc = """ Instructions per FP Arithmetic Scalar Double-Precision instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX128: name = "IpArith_AVX128" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX128(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX128 zero division") desc = """ Instructions per FP Arithmetic AVX/SSE 128-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_IpArith_AVX256: name = "IpArith_AVX256" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Flops', 'FpVector', 'InsType']) sibling = None def compute(self, EV): try: self.val = IpArith_AVX256(self, EV, 0) self.thresh = (self.val < 10) except ZeroDivisionError: handle_error_metric(self, "IpArith_AVX256 zero division") desc = """ Instructions per FP Arithmetic AVX* 256-bit instruction (lower number means higher occurrence rate). Values < 1 are possible due to intentional FMA double counting.""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_Execute: name = "Execute" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Cor', 'Pipeline', 'PortsUtil', 'SMT']) sibling = None def compute(self, EV): try: self.val = Execute(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute zero division") desc = """ """ class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_IpUnknown_Branch: name = "IpUnknown_Branch" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['Fed']) sibling = None def compute(self, EV): try: self.val = IpUnknown_Branch(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IpUnknown_Branch zero division") desc = """ Instructions per speculative Unknown Branch Misprediction (BAClear) (lower number means higher occurrence rate)""" class Metric_IpMispredict: name = "IpMispredict" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BadSpec', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMispredict(self, EV, 0) self.thresh = (self.val < 200) except ZeroDivisionError: handle_error_metric(self, "IpMispredict zero division") desc = """ Number of Instructions per non-speculative Branch Misprediction (JEClear) (lower number means higher occurrence rate)""" class Metric_IpMisp_Indirect: name = "IpMisp_Indirect" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.Bad_Spec" metricgroup = frozenset(['Bad', 'BrMispredicts']) sibling = None def compute(self, EV): try: self.val = IpMisp_Indirect(self, EV, 0) self.thresh = (self.val < 1000) except ZeroDivisionError: handle_error_metric(self, "IpMisp_Indirect zero division") desc = """ Instructions per retired Mispredicts for indirect CALL or JMP branches (lower number means higher occurrence rate).""" class Metric_Load_Miss_Real_Latency: name = "Load_Miss_Real_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryLat']) sibling = None def compute(self, EV): try: self.val = Load_Miss_Real_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_Miss_Real_Latency zero division") desc = """ Actual Average Latency for L1 data-cache miss demand load operations (in core cycles)""" class Metric_MLP: name = "MLP" domain = "Metric" maxval = 10.0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBound', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MLP zero division") desc = """ Memory-Level-Parallelism (average number of L1 miss demand load when there is at least one such miss. Per-Logical Processor)""" class Metric_L1MPKI: name = "L1MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L1MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1MPKI zero division") desc = """ L1 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI: name = "L2MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'Backend', 'CacheHits']) sibling = None def compute(self, EV): try: self.val = L2MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI zero division") desc = """ L2 cache true misses per kilo instruction for retired demand loads""" class Metric_L2MPKI_All: name = "L2MPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_All zero division") desc = """ L2 cache misses per kilo instruction for all request types (including speculative)""" class Metric_L2MPKI_Load: name = "L2MPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2MPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_Load zero division") desc = """ L2 cache misses per kilo instruction for all demand loads (including speculative)""" class Metric_L2MPKI_RFO: name = "L2MPKI_RFO" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheMisses', 'Offcore']) sibling = None def compute(self, EV): try: self.val = L2MPKI_RFO(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2MPKI_RFO zero division") desc = """ Offcore requests (L2 cache miss) per kilo instruction for demand RFOs""" class Metric_L2HPKI_All: name = "L2HPKI_All" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_All(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_All zero division") desc = """ L2 cache hits per kilo instruction for all request types (including speculative)""" class Metric_L2HPKI_Load: name = "L2HPKI_Load" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['CacheHits', 'Mem']) sibling = None def compute(self, EV): try: self.val = L2HPKI_Load(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2HPKI_Load zero division") desc = """ L2 cache hits per kilo instruction for all demand loads (including speculative)""" class Metric_L3MPKI: name = "L3MPKI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem']) sibling = None def compute(self, EV): try: self.val = L3MPKI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3MPKI zero division") desc = """ L3 cache true misses per kilo instruction for retired demand loads""" class Metric_L1D_Cache_Fill_BW: name = "L1D_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW zero division") desc = """ """ class Metric_L2_Cache_Fill_BW: name = "L2_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW zero division") desc = """ """ class Metric_L3_Cache_Fill_BW: name = "L3_Cache_Fill_BW" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Memory" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW zero division") desc = """ """ class Metric_Page_Walks_Utilization: name = "Page_Walks_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.Memory.TLB" metricgroup = frozenset(['Mem', 'MemoryTLB']) sibling = None def compute(self, EV): try: self.val = Page_Walks_Utilization(self, EV, 0) self.thresh = (self.val > 0.5) except ZeroDivisionError: handle_error_metric(self, "Page_Walks_Utilization zero division") desc = """ Utilization of the core's Page Walker(s) serving STLB misses triggered by instruction/Load/Store accesses""" class Metric_L1D_Cache_Fill_BW_2T: name = "L1D_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L1D_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L1D_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L1 data cache [GB / sec]""" class Metric_L2_Cache_Fill_BW_2T: name = "L2_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L2_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L2_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L2 cache [GB / sec]""" class Metric_L3_Cache_Fill_BW_2T: name = "L3_Cache_Fill_BW_2T" domain = "Core_Metric" maxval = 0 errcount = 0 area = "Info.Memory.Core" metricgroup = frozenset(['Mem', 'MemoryBW']) sibling = None def compute(self, EV): try: self.val = L3_Cache_Fill_BW_2T(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "L3_Cache_Fill_BW_2T zero division") desc = """ Average per-core data fill bandwidth to the L3 cache [GB / sec]""" class Metric_Load_L2_Miss_Latency: name = "Load_L2_Miss_Latency" domain = "Clocks_Latency" maxval = 1000 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_Lat', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_Miss_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_Miss_Latency zero division") desc = """ Average Latency for L2 cache miss demand Loads""" class Metric_Load_L2_MLP: name = "Load_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Load_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Load_L2_MLP zero division") desc = """ Average Parallel L2 cache miss demand Loads""" class Metric_Data_L2_MLP: name = "Data_L2_MLP" domain = "Metric" maxval = 100 errcount = 0 area = "Info.Memory.Latency" metricgroup = frozenset(['Memory_BW', 'Offcore']) sibling = None def compute(self, EV): try: self.val = Data_L2_MLP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Data_L2_MLP zero division") desc = """ Average Parallel L2 cache miss data reads""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_Power: name = "Power" domain = "SystemMetric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power', 'SoC']) sibling = None def compute(self, EV): try: self.val = Power(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Power zero division") desc = """ Total package Power in Watts""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ICache_Misses() ; r.run(n) ; o["ICache_Misses"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = Mispredicts_Resteers() ; r.run(n) ; o["Mispredicts_Resteers"] = n n = Clears_Resteers() ; r.run(n) ; o["Clears_Resteers"] = n n = Unknown_Branches() ; r.run(n) ; o["Unknown_Branches"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = MITE() ; r.run(n) ; o["MITE"] = n n = DSB() ; r.run(n) ; o["DSB"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = L1_Bound() ; r.run(n) ; o["L1_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = Store_Fwd_Blk() ; r.run(n) ; o["Store_Fwd_Blk"] = n n = Lock_Latency() ; r.run(n) ; o["Lock_Latency"] = n n = Split_Loads() ; r.run(n) ; o["Split_Loads"] = n n = G4K_Aliasing() ; r.run(n) ; o["G4K_Aliasing"] = n n = FB_Full() ; r.run(n) ; o["FB_Full"] = n n = L2_Bound() ; r.run(n) ; o["L2_Bound"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = Contested_Accesses() ; r.run(n) ; o["Contested_Accesses"] = n n = Data_Sharing() ; r.run(n) ; o["Data_Sharing"] = n n = L3_Hit_Latency() ; r.run(n) ; o["L3_Hit_Latency"] = n n = SQ_Full() ; r.run(n) ; o["SQ_Full"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Store_Latency() ; r.run(n) ; o["Store_Latency"] = n n = False_Sharing() ; r.run(n) ; o["False_Sharing"] = n n = Split_Stores() ; r.run(n) ; o["Split_Stores"] = n n = DTLB_Store() ; r.run(n) ; o["DTLB_Store"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Ports_Utilized_0() ; r.run(n) ; o["Ports_Utilized_0"] = n n = Ports_Utilized_1() ; r.run(n) ; o["Ports_Utilized_1"] = n n = Ports_Utilized_2() ; r.run(n) ; o["Ports_Utilized_2"] = n n = Ports_Utilized_3m() ; r.run(n) ; o["Ports_Utilized_3m"] = n n = ALU_Op_Utilization() ; r.run(n) ; o["ALU_Op_Utilization"] = n n = Port_0() ; r.run(n) ; o["Port_0"] = n n = Port_1() ; r.run(n) ; o["Port_1"] = n n = Port_5() ; r.run(n) ; o["Port_5"] = n n = Port_6() ; r.run(n) ; o["Port_6"] = n n = Load_Op_Utilization() ; r.run(n) ; o["Load_Op_Utilization"] = n n = Port_2() ; r.run(n) ; o["Port_2"] = n n = Port_3() ; r.run(n) ; o["Port_3"] = n n = Store_Op_Utilization() ; r.run(n) ; o["Store_Op_Utilization"] = n n = Port_4() ; r.run(n) ; o["Port_4"] = n n = Port_7() ; r.run(n) ; o["Port_7"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n n = Assists() ; r.run(n) ; o["Assists"] = n n = CISC() ; r.run(n) ; o["CISC"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ICache_Misses"].parent = o["Fetch_Latency"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["Mispredicts_Resteers"].parent = o["Branch_Resteers"] o["Clears_Resteers"].parent = o["Branch_Resteers"] o["Unknown_Branches"].parent = o["Branch_Resteers"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["MITE"].parent = o["Fetch_Bandwidth"] o["DSB"].parent = o["Fetch_Bandwidth"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Memory_Bound"].parent = o["Backend_Bound"] o["L1_Bound"].parent = o["Memory_Bound"] o["DTLB_Load"].parent = o["L1_Bound"] o["Store_Fwd_Blk"].parent = o["L1_Bound"] o["Lock_Latency"].parent = o["L1_Bound"] o["Split_Loads"].parent = o["L1_Bound"] o["G4K_Aliasing"].parent = o["L1_Bound"] o["FB_Full"].parent = o["L1_Bound"] o["L2_Bound"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["Contested_Accesses"].parent = o["L3_Bound"] o["Data_Sharing"].parent = o["L3_Bound"] o["L3_Hit_Latency"].parent = o["L3_Bound"] o["SQ_Full"].parent = o["L3_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Store_Bound"].parent = o["Memory_Bound"] o["Store_Latency"].parent = o["Store_Bound"] o["False_Sharing"].parent = o["Store_Bound"] o["Split_Stores"].parent = o["Store_Bound"] o["DTLB_Store"].parent = o["Store_Bound"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Ports_Utilized_0"].parent = o["Ports_Utilization"] o["Ports_Utilized_1"].parent = o["Ports_Utilization"] o["Ports_Utilized_2"].parent = o["Ports_Utilization"] o["Ports_Utilized_3m"].parent = o["Ports_Utilization"] o["ALU_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_0"].parent = o["ALU_Op_Utilization"] o["Port_1"].parent = o["ALU_Op_Utilization"] o["Port_5"].parent = o["ALU_Op_Utilization"] o["Port_6"].parent = o["ALU_Op_Utilization"] o["Load_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_2"].parent = o["Load_Op_Utilization"] o["Port_3"].parent = o["Load_Op_Utilization"] o["Store_Op_Utilization"].parent = o["Ports_Utilized_3m"] o["Port_4"].parent = o["Store_Op_Utilization"] o["Port_7"].parent = o["Store_Op_Utilization"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["Heavy_Operations"].parent = o["Retiring"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] o["Assists"].parent = o["Microcode_Sequencer"] o["CISC"].parent = o["Microcode_Sequencer"] # user visible metrics n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_UpTB() ; r.metric(n) ; o["UpTB"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_FP_Arith_Utilization() ; r.metric(n) ; o["FP_Arith_Utilization"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_IpLoad() ; r.metric(n) ; o["IpLoad"] = n n = Metric_IpStore() ; r.metric(n) ; o["IpStore"] = n n = Metric_IpBranch() ; r.metric(n) ; o["IpBranch"] = n n = Metric_IpCall() ; r.metric(n) ; o["IpCall"] = n n = Metric_IpTB() ; r.metric(n) ; o["IpTB"] = n n = Metric_BpTkBranch() ; r.metric(n) ; o["BpTkBranch"] = n n = Metric_IpFLOP() ; r.metric(n) ; o["IpFLOP"] = n n = Metric_IpArith() ; r.metric(n) ; o["IpArith"] = n n = Metric_IpArith_Scalar_SP() ; r.metric(n) ; o["IpArith_Scalar_SP"] = n n = Metric_IpArith_Scalar_DP() ; r.metric(n) ; o["IpArith_Scalar_DP"] = n n = Metric_IpArith_AVX128() ; r.metric(n) ; o["IpArith_AVX128"] = n n = Metric_IpArith_AVX256() ; r.metric(n) ; o["IpArith_AVX256"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_Execute() ; r.metric(n) ; o["Execute"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_IpUnknown_Branch() ; r.metric(n) ; o["IpUnknown_Branch"] = n n = Metric_IpMispredict() ; r.metric(n) ; o["IpMispredict"] = n n = Metric_IpMisp_Indirect() ; r.metric(n) ; o["IpMisp_Indirect"] = n n = Metric_Load_Miss_Real_Latency() ; r.metric(n) ; o["Load_Miss_Real_Latency"] = n n = Metric_MLP() ; r.metric(n) ; o["MLP"] = n n = Metric_L1MPKI() ; r.metric(n) ; o["L1MPKI"] = n n = Metric_L2MPKI() ; r.metric(n) ; o["L2MPKI"] = n n = Metric_L2MPKI_All() ; r.metric(n) ; o["L2MPKI_All"] = n n = Metric_L2MPKI_Load() ; r.metric(n) ; o["L2MPKI_Load"] = n n = Metric_L2MPKI_RFO() ; r.metric(n) ; o["L2MPKI_RFO"] = n n = Metric_L2HPKI_All() ; r.metric(n) ; o["L2HPKI_All"] = n n = Metric_L2HPKI_Load() ; r.metric(n) ; o["L2HPKI_Load"] = n n = Metric_L3MPKI() ; r.metric(n) ; o["L3MPKI"] = n n = Metric_L1D_Cache_Fill_BW() ; r.metric(n) ; o["L1D_Cache_Fill_BW"] = n n = Metric_L2_Cache_Fill_BW() ; r.metric(n) ; o["L2_Cache_Fill_BW"] = n n = Metric_L3_Cache_Fill_BW() ; r.metric(n) ; o["L3_Cache_Fill_BW"] = n n = Metric_Page_Walks_Utilization() ; r.metric(n) ; o["Page_Walks_Utilization"] = n n = Metric_L1D_Cache_Fill_BW_2T() ; r.metric(n) ; o["L1D_Cache_Fill_BW_2T"] = n n = Metric_L2_Cache_Fill_BW_2T() ; r.metric(n) ; o["L2_Cache_Fill_BW_2T"] = n n = Metric_L3_Cache_Fill_BW_2T() ; r.metric(n) ; o["L3_Cache_Fill_BW_2T"] = n n = Metric_Load_L2_Miss_Latency() ; r.metric(n) ; o["Load_L2_Miss_Latency"] = n n = Metric_Load_L2_MLP() ; r.metric(n) ; o["Load_L2_MLP"] = n n = Metric_Data_L2_MLP() ; r.metric(n) ; o["Data_L2_MLP"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_Power() ; r.metric(n) ; o["Power"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Mispredicts_Resteers"].Branch_Resteers = o["Branch_Resteers"] o["Clears_Resteers"].Branch_Resteers = o["Branch_Resteers"] o["Unknown_Branches"].Clears_Resteers = o["Clears_Resteers"] o["Unknown_Branches"].Branch_Resteers = o["Branch_Resteers"] o["Unknown_Branches"].Mispredicts_Resteers = o["Mispredicts_Resteers"] o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Branch_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Backend_Bound"].Retiring = o["Retiring"] o["Backend_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Backend_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Retiring = o["Retiring"] o["Memory_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Backend_Bound = o["Backend_Bound"] o["Memory_Bound"].Fetch_Latency = o["Fetch_Latency"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Core_Bound"].Retiring = o["Retiring"] o["Core_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Core_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Core_Bound"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilization"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilized_0"].Fetch_Latency = o["Fetch_Latency"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["Heavy_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Microcode_Sequencer = o["Microcode_Sequencer"] o["CISC"].Assists = o["Assists"] # siblings cross-tree o["Mispredicts_Resteers"].sibling = (o["Branch_Mispredicts"],) o["Clears_Resteers"].sibling = (o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["MS_Switches"].sibling = (o["Clears_Resteers"], o["Machine_Clears"], o["L1_Bound"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Branch_Mispredicts"].sibling = (o["Mispredicts_Resteers"],) o["Machine_Clears"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["L1_Bound"], o["Contested_Accesses"], o["Data_Sharing"], o["False_Sharing"], o["Microcode_Sequencer"],) o["L1_Bound"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["Ports_Utilized_1"], o["Microcode_Sequencer"],) o["DTLB_Load"].sibling = (o["DTLB_Store"],) o["Lock_Latency"].sibling = (o["Store_Latency"],) o["FB_Full"].sibling = (o["SQ_Full"], o["MEM_Bandwidth"], o["Store_Latency"],) o["Contested_Accesses"].sibling = (o["Machine_Clears"], o["Data_Sharing"], o["False_Sharing"],) o["Data_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["False_Sharing"],) o["L3_Hit_Latency"].sibling = (o["MEM_Latency"],) o["L3_Hit_Latency"].overlap = True o["SQ_Full"].sibling = (o["FB_Full"], o["MEM_Bandwidth"],) o["MEM_Bandwidth"].sibling = (o["FB_Full"], o["SQ_Full"],) o["MEM_Latency"].sibling = (o["L3_Hit_Latency"],) o["Store_Latency"].sibling = (o["Lock_Latency"], o["FB_Full"],) o["Store_Latency"].overlap = True o["False_Sharing"].sibling = (o["Machine_Clears"], o["Contested_Accesses"], o["Data_Sharing"],) o["Split_Stores"].sibling = (o["Port_4"],) o["DTLB_Store"].sibling = (o["DTLB_Load"],) o["Ports_Utilized_1"].sibling = (o["L1_Bound"],) o["Ports_Utilized_2"].sibling = (o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_0"].sibling = (o["Ports_Utilized_2"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_1"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_5"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_6"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["Port_4"].sibling = (o["Split_Stores"],) o["FP_Scalar"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["Ports_Utilized_2"], o["Port_0"], o["Port_1"], o["Port_5"], o["Port_6"], o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"],) o["Microcode_Sequencer"].sibling = (o["Clears_Resteers"], o["MS_Switches"], o["Machine_Clears"], o["L1_Bound"],) o["IpTB"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DRAM_BW_Use"].sibling = (o["FB_Full"], o["SQ_Full"], o["MEM_Bandwidth"],)
142,042
Python
.py
3,523
34.632132
323
0.656478
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,894
tl-barplot.py
andikleen_pmu-tools/tl-barplot.py
#!/usr/bin/env python3 # plot toplev -I... -x, -o ...csv output as bar plot # from __future__ import print_function import os import re import argparse from math import isnan, trunc from collections import defaultdict import matplotlib if os.getenv('DISPLAY') is None: matplotlib.use('Agg') import matplotlib.pyplot as plt import gen_level import tldata def parse_args(): p = argparse.ArgumentParser(usage='plot toplev -I... -x, output as bar plot') p.add_argument('file', help='CSV file to plot') p.add_argument('--output', '-o', help='Save figure to file (.pdf/.png/etc). Otherwise show.', nargs='?') p.add_argument('--verbose', '-v', help='Plot all data values even if below threshold', action='store_true') p.add_argument('--xkcd', help='Enable XKCD mode (with new matplotlib). Please install Humor Sans.', action='store_true') p.add_argument('--title', help='Set title of plot', nargs='?') p.add_argument('--quiet', help='Be quiet', action='store_true') p.add_argument('--cpu', help='CPU to plot (by default first)') # XXX return p.parse_args() args = parse_args() try: import brewer2mpl except ImportError: if not args.quiet: print("pip install brewer2mpl for better colors") if args.xkcd: plt.xkcd() data = tldata.TLData(args.file, args.verbose) data.update() levels = data.levels timestamps = data.times ratios = defaultdict(list) # XXX plot multiple cpus instead cpu = None if args.cpu: cpu = args.cpu elif 'CLKS' in data.headers and len(data.vals) > 0: # pick CPU with highest utilization. XXX look at all time series util = sorted([(data.vals[0][x], x[1]) for x in data.vals[0].keys() if x[0] == 'CLKS'], reverse=True) cpu = util[0][1] elif len(data.cpus) > 0: cpu = sorted(sorted(data.cpus), key=len, reverse=True)[0] def cpumatch(x, cpu, base): return x.startswith(cpu) or x == base if cpu: base = None m = re.match(r'C\d+', cpu) if m: base = m.group(0) aliases = [x for x in data.cpus if cpumatch(x, cpu, base)] print("plotting cpus:", " ".join(sorted(aliases))) else: aliases = [] if len(aliases) == 0: aliases = [None] for h in data.headers: def findval(d): for c in aliases: if (h, c) in d: return d[(h, c)] return float('nan') ratios[h] = list(map(findval, data.vals)) def valid_row(r): s = sum(r) #if sum([0 if isnan(x) else 1 for x in r]) < len(r)/80.: # return False return s != 0.0 and not isnan(s) def get_colors(non_null): if 'brewer2mpl' in globals(): num_color = max(min(len(non_null), 11), 3) all_colors = brewer2mpl.get_map('Spectral', 'Diverging', num_color).hex_colors else: all_colors = None return all_colors def set_title(ax, t): try: ax.set_title(t, {'fontsize': 6}, loc='right') except AttributeError: ax.set_title(t) def suffix(x): dot = x.rfind('.') if dot >= 0: return x[dot + 1:] return x n = 0 numplots = len(levels.keys()) ax = None yset = False max_legend = 0 xaxis = None legend_bbox = (0., 0., -0.07, -0.03) legend_loc = 2 for l in tldata.level_order(data): non_null = [x for x in levels[l] if valid_row(ratios[x])] if not non_null: n += 1 continue all_colors = get_colors(non_null) ax = plt.subplot2grid((numplots, 1), (n, 0), sharex=xaxis) plt.tight_layout() set_title(ax, l) r = [[y if not isnan(y) else 0.0 for y in ratios[x]] for x in non_null] if gen_level.is_metric(non_null[0]): for j, name in zip(r, non_null): stack = ax.plot(timestamps, j, label=name) leg = plt.legend(ncol=6, loc=legend_loc, bbox_to_anchor=legend_bbox, prop={'size':6}) low = min([min(ratios[x]) for x in non_null]) high = max([max(ratios[x]) for x in non_null]) if not isnan(low) and not isnan(high): ax.yaxis.set_ticks([low, trunc(((high - low) / 2.0) / 100.) * 100., high]) else: stack = ax.stackplot(timestamps, *r, colors=all_colors) ax.set_ylim(0, 100) ax.yaxis.set_ticks([0., 50., 100.]) p = [plt.Rectangle((0, 0), 1, 1, fc=pc.get_facecolor()[0]) for pc in stack] leg = plt.legend(p, list(map(suffix, non_null)), ncol=6, bbox_to_anchor=legend_bbox, loc=legend_loc, prop={'size':6}) leg.get_frame().set_alpha(0.5) for j in ax.get_xticklabels() + ax.get_yticklabels(): j.set_fontsize(6) if not xaxis: xaxis = ax #if n >= 2 and not yset and l != -1: # ax.set_ylabel('(% of execution time)') # yset = True if n != numplots: max_legend = max(len(non_null), max_legend) #ax.margins(0, 0) n += 1 if len(timestamps) == 1: plt.gca().axes.get_xaxis().set_visible(False) plt.subplots_adjust(hspace=1.5 if max_legend > 6 else 0.9, bottom=0.20, top=0.95) if args.title: #plt.subplot(numplots, 1, 1) plt.title(args.title) if args.output: plt.savefig(args.output) else: plt.show()
5,231
Python
.py
154
28.266234
124
0.604826
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,895
jkt_server_ratios.py
andikleen_pmu-tools/jkt_server_ratios.py
# -*- coding: latin-1 -*- # # auto generated TopDown/TMA 4.8-full-perf description for Intel Xeon E5 (code named SandyBridge EP) # Please see http://ark.intel.com for more details on these CPUs. # # References: # http://bit.ly/tma-ispass14 # http://halobates.de/blog/p/262 # https://sites.google.com/site/analysismethods/yasin-pubs # https://download.01.org/perfmon/ # https://github.com/andikleen/pmu-tools/wiki/toplev-manual # # Helpers print_error = lambda msg: False smt_enabled = False ebs_mode = False version = "4.8-full-perf" base_frequency = -1.0 Memory = 0 Average_Frequency = 0.0 num_cores = 1 num_threads = 1 num_sockets = 1 def handle_error(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 obj.thresh = False def handle_error_metric(obj, msg): print_error(msg) obj.errcount += 1 obj.val = 0 # Constants Exe_Ports = 6 Mem_L3_Weight = 7 Mem_STLB_Hit_Cost = 7 BAClear_Cost = 12 MS_Switches_Cost = 3 Pipeline_Width = 4 OneMillion = 1000000 OneBillion = 1000000000 EBS_Mode = 0 DS = 1 # Aux. formulas def Backend_Bound_Cycles(self, EV, level): return (STALLS_TOTAL(self, EV, level) + EV("UOPS_DISPATCHED.THREAD:c1", level) - Few_Uops_Executed_Threshold(self, EV, level) - Frontend_RS_Empty_Cycles(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) def DurationTimeInSeconds(self, EV, level): return EV("interval-ms", 0) / 1000 def Execute_Cycles(self, EV, level): return (EV("UOPS_DISPATCHED.CORE:c1", level) / 2) if smt_enabled else EV("UOPS_DISPATCHED.CORE:c1", level) def Fetched_Uops(self, EV, level): return (EV("IDQ.DSB_UOPS", level) + EV("LSD.UOPS", level) + EV("IDQ.MITE_UOPS", level) + EV("IDQ.MS_UOPS", level)) def Few_Uops_Executed_Threshold(self, EV, level): EV("UOPS_DISPATCHED.THREAD:c3", level) EV("UOPS_DISPATCHED.THREAD:c2", level) return EV("UOPS_DISPATCHED.THREAD:c3", level) if (IPC(self, EV, level)> 1.8) else EV("UOPS_DISPATCHED.THREAD:c2", level) # Floating Point computational (arithmetic) Operations Count def FLOP_Count(self, EV, level): return (1 *(EV("FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE", level) + EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", level)) + 2 * EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", level) + 4 *(EV("FP_COMP_OPS_EXE.SSE_PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_DOUBLE", level)) + 8 * EV("SIMD_FP_256.PACKED_SINGLE", level)) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Scalar(self, EV, level): return EV("FP_COMP_OPS_EXE.SSE_SCALAR_SINGLE", level) + EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", level) # Floating Point computational (arithmetic) Operations Count def FP_Arith_Vector(self, EV, level): return EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", level) + EV("FP_COMP_OPS_EXE.SSE_PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_SINGLE", level) + EV("SIMD_FP_256.PACKED_DOUBLE", level) def Frontend_RS_Empty_Cycles(self, EV, level): EV("RS_EVENTS.EMPTY_CYCLES", level) return EV("RS_EVENTS.EMPTY_CYCLES", level) if (self.Fetch_Latency.compute(EV)> 0.1) else 0 def Frontend_Latency_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("IDQ_UOPS_NOT_DELIVERED.CYCLES_0_UOPS_DELIV.CORE", level)) , level ) def HighIPC(self, EV, level): val = IPC(self, EV, level) / Pipeline_Width return val def ITLB_Miss_Cycles(self, EV, level): return (12 * EV("ITLB_MISSES.STLB_HIT", level) + EV("ITLB_MISSES.WALK_DURATION", level)) def Mem_L3_Hit_Fraction(self, EV, level): return EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) / (EV("MEM_LOAD_UOPS_RETIRED.LLC_HIT", level) + Mem_L3_Weight * EV("MEM_LOAD_UOPS_RETIRED.LLC_MISS", level)) def Memory_Bound_Fraction(self, EV, level): return (STALLS_MEM_ANY(self, EV, level) + EV("RESOURCE_STALLS.SB", level)) / Backend_Bound_Cycles(self, EV, level) def Mispred_Clears_Fraction(self, EV, level): return EV("BR_MISP_RETIRED.ALL_BRANCHES", level) / (EV("BR_MISP_RETIRED.ALL_BRANCHES", level) + EV("MACHINE_CLEARS.COUNT", level)) def ORO_DRD_Any_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.CYCLES_WITH_DATA_RD", level)) , level ) def ORO_DRD_BW_Cycles(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("OFFCORE_REQUESTS_OUTSTANDING.ALL_DATA_RD:c6", level)) , level ) def STALLS_MEM_ANY(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("CYCLE_ACTIVITY.STALLS_L1D_PENDING", level)) , level ) def STALLS_TOTAL(self, EV, level): return EV(lambda EV , level : min(EV("CPU_CLK_UNHALTED.THREAD", level) , EV("CYCLE_ACTIVITY.CYCLES_NO_DISPATCH", level)) , level ) def Recovery_Cycles(self, EV, level): return (EV("INT_MISC.RECOVERY_CYCLES_ANY", level) / 2) if smt_enabled else EV("INT_MISC.RECOVERY_CYCLES", level) def Retire_Fraction(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_ISSUED.ANY", level) # Retired slots per Logical Processor def Retired_Slots(self, EV, level): return EV("UOPS_RETIRED.RETIRE_SLOTS", level) # Number of logical processors (enabled or online) on the target system def Num_CPUs(self, EV, level): return 8 if smt_enabled else 4 # Instructions Per Cycle (per Logical Processor) def IPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(self, EV, level) # Uops Per Instruction def UopPI(self, EV, level): val = Retired_Slots(self, EV, level) / EV("INST_RETIRED.ANY", level) self.thresh = (val > 1.05) return val # Cycles Per Instruction (per Logical Processor) def CPI(self, EV, level): return 1 / IPC(self, EV, level) # Per-Logical Processor actual clocks when the Logical Processor is active. def CLKS(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD", level) # Total issue-pipeline slots (per-Physical Core till ICL; per-Logical Processor ICL onward) def SLOTS(self, EV, level): return Pipeline_Width * CORE_CLKS(self, EV, level) # The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of "execute" at rename stage. def Execute_per_Issue(self, EV, level): return EV("UOPS_DISPATCHED.THREAD", level) / EV("UOPS_ISSUED.ANY", level) # Instructions Per Cycle across hyper-threads (per physical core) def CoreIPC(self, EV, level): return EV("INST_RETIRED.ANY", level) / CORE_CLKS(self, EV, level) # Floating Point Operations Per Cycle def FLOPc(self, EV, level): return FLOP_Count(self, EV, level) / CORE_CLKS(self, EV, level) # Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical-processor) def ILP(self, EV, level): return EV("UOPS_DISPATCHED.THREAD", level) / Execute_Cycles(self, EV, level) # Core actual clocks when any Logical Processor is active on the Physical Core def CORE_CLKS(self, EV, level): return ((EV("CPU_CLK_UNHALTED.THREAD", level) / 2) * (1 + EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / EV("CPU_CLK_UNHALTED.REF_XCLK", level))) if ebs_mode else(EV("CPU_CLK_UNHALTED.THREAD_ANY", level) / 2) if smt_enabled else CLKS(self, EV, level) # Total number of retired Instructions def Instructions(self, EV, level): return EV("INST_RETIRED.ANY", level) # Average number of Uops retired in cycles where at least one uop has retired. def Retire(self, EV, level): return Retired_Slots(self, EV, level) / EV("UOPS_RETIRED.RETIRE_SLOTS:c1", level) # Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture-and-technology/64-ia-32-architectures-optimization-manual.html def DSB_Coverage(self, EV, level): val = EV("IDQ.DSB_UOPS", level) / Fetched_Uops(self, EV, level) self.thresh = (val < 0.7) and HighIPC(self, EV, 1) return val # Average CPU Utilization (percentage) def CPU_Utilization(self, EV, level): return CPUs_Utilized(self, EV, level) / Num_CPUs(self, EV, level) # Average number of utilized CPUs def CPUs_Utilized(self, EV, level): return EV("CPU_CLK_UNHALTED.REF_TSC", level) / EV("msr/tsc/", 0) # Measured Average Core Frequency for unhalted processors [GHz] def Core_Frequency(self, EV, level): return Turbo_Utilization(self, EV, level) * EV("msr/tsc/", 0) / OneBillion / Time(self, EV, level) # Measured Average Uncore Frequency for the SoC [GHz] def Uncore_Frequency(self, EV, level): return Socket_CLKS(self, EV, level) / 1e9 / Time(self, EV, level) # Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width def GFLOPs(self, EV, level): return (FLOP_Count(self, EV, level) / OneBillion) / Time(self, EV, level) # Average Frequency Utilization relative nominal frequency def Turbo_Utilization(self, EV, level): return CLKS(self, EV, level) / EV("CPU_CLK_UNHALTED.REF_TSC", level) # Fraction of cycles where both hardware Logical Processors were active def SMT_2T_Utilization(self, EV, level): return 1 - EV("CPU_CLK_UNHALTED.ONE_THREAD_ACTIVE", level) / (EV("CPU_CLK_UNHALTED.REF_XCLK_ANY", level) / 2) if smt_enabled else 0 # Fraction of cycles spent in the Operating System (OS) Kernel mode def Kernel_Utilization(self, EV, level): val = EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("CPU_CLK_UNHALTED.THREAD", level) self.thresh = (val > 0.05) return val # Cycles Per Instruction for the Operating System (OS) Kernel mode def Kernel_CPI(self, EV, level): return EV("CPU_CLK_UNHALTED.THREAD_P:SUP", level) / EV("INST_RETIRED.ANY_P:SUP", level) # Average external Memory Bandwidth Use for reads and writes [GB / sec] def DRAM_BW_Use(self, EV, level): return (64 *(EV("UNC_M_CAS_COUNT.RD", level) + EV("UNC_M_CAS_COUNT.WR", level)) / OneBillion) / Time(self, EV, level) # Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches. def MEM_Read_Latency(self, EV, level): return OneBillion *(EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182", level) / EV("UNC_C_TOR_INSERTS.MISS_OPCODE:Match=0x182", level)) / (Socket_CLKS(self, EV, level) / Time(self, EV, level)) # Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches def MEM_Parallel_Reads(self, EV, level): return EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182", level) / EV("UNC_C_TOR_OCCUPANCY.MISS_OPCODE:Match=0x182:c1", level) # Run duration time in seconds def Time(self, EV, level): val = EV("interval-s", 0) self.thresh = (val < 1) return val # Socket actual clocks when any core is active on that socket def Socket_CLKS(self, EV, level): return EV("UNC_C_CLOCKTICKS:one_unit", level) # Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate] def IpFarBranch(self, EV, level): val = EV("INST_RETIRED.ANY", level) / EV("BR_INST_RETIRED.FAR_BRANCH:USER", level) self.thresh = (val < 1000000) return val # Event groups class Frontend_Bound: name = "Frontend_Bound" domain = "Slots" area = "FE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvFB', 'BvIO', 'TmaL1', 'PGO']) maxval = None def compute(self, EV): try: self.val = EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Frontend_Bound zero division") return self.val desc = """ This category represents fraction of slots where the processor's Frontend undersupplies its Backend. Frontend denotes the first part of the processor core responsible to fetch operations that are executed later on by the Backend part. Within the Frontend; a branch predictor predicts the next address to fetch; cache-lines are fetched from the memory subsystem; parsed into instructions; and lastly decoded into micro-operations (uops). Ideally the Frontend can issue Pipeline_Width uops every cycle to the Backend. Frontend Bound denotes unutilized issue-slots when there is no Backend stall; i.e. bubbles where Frontend delivered no uops while Backend could have accepted them. For example; stalls due to instruction-cache misses would be categorized under Frontend Bound.""" class Fetch_Latency: name = "Fetch_Latency" domain = "Slots" area = "FE" level = 2 htoff = False sample = ['RS_EVENTS.EMPTY_END'] errcount = 0 sibling = None metricgroup = frozenset(['Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Pipeline_Width * Frontend_Latency_Cycles(self, EV, 2) / SLOTS(self, EV, 2) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Fetch_Latency zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend latency issues. For example; instruction- cache misses; iTLB misses or fetch stalls after a branch misprediction are categorized under Frontend Latency. In such cases; the Frontend eventually delivers no uops for some period.""" class ITLB_Misses: name = "ITLB_Misses" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['ITLB_MISSES.WALK_COMPLETED'] errcount = 0 sibling = None metricgroup = frozenset(['BigFootprint', 'BvBC', 'FetchLat', 'MemoryTLB']) maxval = None def compute(self, EV): try: self.val = ITLB_Miss_Cycles(self, EV, 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "ITLB_Misses zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Instruction TLB (ITLB) misses.. Consider large 2M pages for code (selectively prefer hot large-size function, due to limited 2M entries). Linux options: standard binaries use libhugetlbfs; Hfsort.. https://github. com/libhugetlbfs/libhugetlbfs;https://research.fb.com/public ations/optimizing-function-placement-for-large-scale-data- center-applications-2/""" class Branch_Resteers: name = "Branch_Resteers" domain = "Clocks" area = "FE" level = 3 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = BAClear_Cost *(EV("BR_MISP_RETIRED.ALL_BRANCHES", 3) + EV("MACHINE_CLEARS.COUNT", 3) + EV("BACLEARS.ANY", 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Resteers zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to Branch Resteers. Branch Resteers estimates the Frontend delay in fetching operations from corrected path; following all sorts of miss-predicted branches. For example; branchy code with lots of miss-predictions might get categorized under Branch Resteers. Note the value of this node may overlap with its siblings.""" class MS_Switches: name = "MS_Switches" domain = "Clocks_Estimated" area = "FE" level = 3 htoff = False sample = ['IDQ.MS_SWITCHES'] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat', 'MicroSeq']) maxval = 1.0 def compute(self, EV): try: self.val = MS_Switches_Cost * EV("IDQ.MS_SWITCHES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MS_Switches zero division") return self.val desc = """ This metric estimates the fraction of cycles when the CPU was stalled due to switches of uop delivery to the Microcode Sequencer (MS). Commonly used instructions are optimized for delivery by the DSB (decoded i-cache) or MITE (legacy instruction decode) pipelines. Certain operations cannot be handled natively by the execution pipeline; and must be performed by microcode (small programs injected into the execution stream). Switching to the MS too often can negatively impact performance. The MS is designated to deliver long uop flows required by CISC instructions like CPUID; or uncommon conditions like Floating Point Assists when dealing with Denormals.""" class LCP: name = "LCP" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("ILD_STALL.LCP", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "LCP zero division") return self.val desc = """ This metric represents fraction of cycles CPU was stalled due to Length Changing Prefixes (LCPs). Using proper compiler flags or Intel Compiler by default will certainly avoid this.""" class DSB_Switches: name = "DSB_Switches" domain = "Clocks" area = "FE" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['DSBmiss', 'FetchLat']) maxval = None def compute(self, EV): try: self.val = EV("DSB2MITE_SWITCHES.PENALTY_CYCLES", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DSB_Switches zero division") return self.val desc = """ This metric represents fraction of cycles the CPU was stalled due to switches from DSB to MITE pipelines. The DSB (decoded i-cache) is a Uop Cache where the front-end directly delivers Uops (micro operations) avoiding heavy x86 decoding. The DSB pipeline has shorter latency and delivered higher bandwidth than the MITE (legacy instruction decode pipeline). Switching between the two pipelines can cause penalties hence this metric measures the exposed penalty.. See section 'Optimization for Decoded Icache' in Optimization Manual:. http://www.intel.com/content/www/us/en /architecture-and-technology/64-ia-32-architectures- optimization-manual.html""" class Fetch_Bandwidth: name = "Fetch_Bandwidth" domain = "Slots" area = "FE" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['FetchBW', 'Frontend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Frontend_Bound.compute(EV) - self.Fetch_Latency.compute(EV) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Fetch_Bandwidth zero division") return self.val desc = """ This metric represents fraction of slots the CPU was stalled due to Frontend bandwidth issues. For example; inefficiencies at the instruction decoders; or restrictions for caching in the DSB (decoded uops cache) are categorized under Fetch Bandwidth. In such cases; the Frontend typically delivers suboptimal amount of uops to the Backend.""" class Bad_Speculation: name = "Bad_Speculation" domain = "Slots" area = "BAD" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['TmaL1']) maxval = None def compute(self, EV): try: self.val = (EV("UOPS_ISSUED.ANY", 1) - Retired_Slots(self, EV, 1) + Pipeline_Width * Recovery_Cycles(self, EV, 1)) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.15) except ZeroDivisionError: handle_error(self, "Bad_Speculation zero division") return self.val desc = """ This category represents fraction of slots wasted due to incorrect speculations. This include slots used to issue uops that do not eventually get retired and slots for which the issue-pipeline was blocked due to recovery from earlier incorrect speculation. For example; wasted work due to miss- predicted branches are categorized under Bad Speculation category. Incorrect data speculation followed by Memory Ordering Nukes is another example.""" class Branch_Mispredicts: name = "Branch_Mispredicts" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['BR_MISP_RETIRED.ALL_BRANCHES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BrMispredicts', 'BvMP', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Mispred_Clears_Fraction(self, EV, 2) * self.Bad_Speculation.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Branch_Mispredicts zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Branch Misprediction. These slots are either wasted by uops fetched from an incorrectly speculated program path; or stalls when the out-of-order part of the machine needs to recover its state from a speculative path.. Using profile feedback in the compiler may help. Please see the Optimization Manual for general strategies for addressing branch misprediction issues.. http://www.intel.com/content/www/us/en/architecture-and- technology/64-ia-32-architectures-optimization-manual.html""" class Machine_Clears: name = "Machine_Clears" domain = "Slots" area = "BAD" level = 2 htoff = False sample = ['MACHINE_CLEARS.COUNT'] errcount = 0 sibling = None metricgroup = frozenset(['BadSpec', 'BvMS', 'MachineClears', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Bad_Speculation.compute(EV) - self.Branch_Mispredicts.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Machine_Clears zero division") return self.val desc = """ This metric represents fraction of slots the CPU has wasted due to Machine Clears. These slots are either wasted by uops fetched prior to the clear; or stalls the out-of-order portion of the machine needs to recover its state after the clear. For example; this can happen due to memory ordering Nukes (e.g. Memory Disambiguation) or Self-Modifying-Code (SMC) nukes.. See \"Memory Disambiguation\" in Optimization Manual and:. https://software.intel.com/sites/default/files/ m/d/4/1/d/8/sma.pdf""" class Backend_Bound: name = "Backend_Bound" domain = "Slots" area = "BE" level = 1 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvOB', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = 1 -(self.Frontend_Bound.compute(EV) + self.Bad_Speculation.compute(EV) + self.Retiring.compute(EV)) self.thresh = (self.val > 0.2) except ZeroDivisionError: handle_error(self, "Backend_Bound zero division") return self.val desc = """ This category represents fraction of slots where no uops are being delivered due to a lack of required resources for accepting new uops in the Backend. Backend is the portion of the processor core where the out-of-order scheduler dispatches ready uops into their respective execution units; and once completed these uops get retired according to program order. For example; stalls due to data-cache misses or stalls due to the divider unit being overloaded are both categorized under Backend Bound. Backend Bound is further divided into two main categories: Memory Bound and Core Bound.""" class Memory_Bound: name = "Memory_Bound" domain = "Slots" area = "BE/Mem" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = Memory_Bound_Fraction(self, EV, 2) * self.Backend_Bound.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Memory_Bound zero division") return self.val desc = """ This metric represents fraction of slots the Memory subsystem within the Backend was a bottleneck. Memory Bound estimates fraction of slots where pipeline is likely stalled due to demand load or store instructions. This accounts mainly for (1) non-completed in-flight memory demand loads which coincides with execution units starvation; in addition to (2) cases where stores could impose backpressure on the pipeline when many of them get buffered at the same time (less common out of the two).""" class DTLB_Load: name = "DTLB_Load" domain = "Clocks_Estimated" area = "BE/Mem" level = 4 htoff = False sample = ['MEM_UOPS_RETIRED.STLB_MISS_LOADS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['BvMT', 'MemoryTLB']) maxval = 1.0 def compute(self, EV): try: self.val = (Mem_STLB_Hit_Cost * EV("DTLB_LOAD_MISSES.STLB_HIT", 4) + EV("DTLB_LOAD_MISSES.WALK_DURATION", 4)) / CLKS(self, EV, 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DTLB_Load zero division") return self.val desc = """ This metric roughly estimates the fraction of cycles where the Data TLB (DTLB) was missed by load accesses. TLBs (Translation Look-aside Buffers) are processor caches for recently used entries out of the Page Tables that are used to map virtual- to physical-addresses by the operating system. This metric approximates the potential delay of demand loads missing the first-level data TLB (assuming worst case scenario with back to back misses to different pages). This includes hitting in the second-level TLB (STLB) as well as performing a hardware page walk on an STLB miss..""" class L3_Bound: name = "L3_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_HIT:pp'] errcount = 0 sibling = None metricgroup = frozenset(['CacheHits', 'MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = Mem_L3_Hit_Fraction(self, EV, 3) * EV("CYCLE_ACTIVITY.STALLS_L2_PENDING", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "L3_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled due to loads accesses to L3 cache or contended with a sibling Core. Avoiding cache misses (i.e. L2 misses/L3 hits) can improve the latency and increase performance.""" class DRAM_Bound: name = "DRAM_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_LOAD_UOPS_RETIRED.LLC_MISS:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = 1.0 def compute(self, EV): try: self.val = (1 - Mem_L3_Hit_Fraction(self, EV, 3)) * EV("CYCLE_ACTIVITY.STALLS_L2_PENDING", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "DRAM_Bound zero division") return self.val desc = """ This metric estimates how often the CPU was stalled on accesses to external memory (DRAM) by loads. Better caching can improve the latency and increase performance.""" class MEM_Bandwidth: name = "MEM_Bandwidth" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvMB', 'MemoryBW', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_BW_Cycles(self, EV, 4) / CLKS(self, EV, 4) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Bandwidth zero division") return self.val desc = """ This metric estimates fraction of cycles where the core's performance was likely hurt due to approaching bandwidth limits of external memory - DRAM ([SPR-HBM] and/or HBM). The underlying heuristic assumes that a similar off-core traffic is generated by all IA cores. This metric does not aggregate non-data-read requests by this logical processor; requests from other IA Logical Processors/Physical Cores/sockets; or other non-IA devices like GPU; hence the maximum external memory bandwidth limits may or may not be approached when this metric is flagged (see Uncore counters for that).. Improve data accesses to reduce cacheline transfers from/to memory. Examples: 1) Consume all bytes of a each cacheline before it is evicted (e.g. reorder structure elements and split non-hot ones), 2) merge computed-limited with BW-limited loops, 3) NUMA optimizations in multi-socket system. Note: software prefetches will not help BW-limited application..""" class MEM_Latency: name = "MEM_Latency" domain = "Clocks" area = "BE/Mem" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['BvML', 'MemoryLat', 'Offcore']) maxval = None def compute(self, EV): try: self.val = ORO_DRD_Any_Cycles(self, EV, 4) / CLKS(self, EV, 4) - self.MEM_Bandwidth.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "MEM_Latency zero division") return self.val desc = """ This metric estimates fraction of cycles where the performance was likely hurt due to latency from external memory - DRAM ([SPR-HBM] and/or HBM). This metric does not aggregate requests from other Logical Processors/Physical Cores/sockets (see Uncore counters for that).. Improve data accesses or interleave them with compute. Examples: 1) Data layout re-structuring, 2) Software Prefetches (also through the compiler)..""" class Store_Bound: name = "Store_Bound" domain = "Stalls" area = "BE/Mem" level = 3 htoff = False sample = ['MEM_UOPS_RETIRED.ALL_STORES:pp'] errcount = 0 sibling = None metricgroup = frozenset(['MemoryBound', 'TmaL3mem']) maxval = None def compute(self, EV): try: self.val = EV("RESOURCE_STALLS.SB", 3) / CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Store_Bound zero division") return self.val desc = """ This metric estimates how often CPU was stalled due to RFO store memory accesses; RFO store issue a read-for-ownership request before the write. Even though store accesses do not typically stall out-of-order CPUs; there are few cases where stores can lead to actual stalls. This metric will be flagged should RFO stores be a bottleneck.""" class Core_Bound: name = "Core_Bound" domain = "Slots" area = "BE/Core" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Backend', 'TmaL2', 'Compute']) maxval = None def compute(self, EV): try: self.val = self.Backend_Bound.compute(EV) - self.Memory_Bound.compute(EV) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Core_Bound zero division") return self.val desc = """ This metric represents fraction of slots where Core non- memory issues were of a bottleneck. Shortage in hardware compute resources; or dependencies in software's instructions are both categorized under Core Bound. Hence it may indicate the machine ran out of an out-of-order resource; certain execution units are overloaded or dependencies in program's data- or instruction-flow are limiting the performance (e.g. FP-chained long-latency arithmetic operations).. Tip: consider Port Saturation analysis as next step.""" class Divider: name = "Divider" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = ['ARITH.FPU_DIV_ACTIVE'] errcount = 0 sibling = None metricgroup = frozenset(['BvCB']) maxval = 1.0 def compute(self, EV): try: self.val = EV("ARITH.FPU_DIV_ACTIVE", 3) / CORE_CLKS(self, EV, 3) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Divider zero division") return self.val desc = """ This metric represents fraction of cycles where the Divider unit was active. Divide and square root instructions are performed by the Divider unit and can take considerably longer latency than integer or Floating Point addition; subtraction; or multiplication.""" class Ports_Utilization: name = "Ports_Utilization" domain = "Clocks" area = "BE/Core" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['PortsUtil']) maxval = None def compute(self, EV): try: self.val = (Backend_Bound_Cycles(self, EV, 3) - EV("RESOURCE_STALLS.SB", 3) - STALLS_MEM_ANY(self, EV, 3)) / CLKS(self, EV, 3) self.thresh = (self.val > 0.15) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Ports_Utilization zero division") return self.val desc = """ This metric estimates fraction of cycles the CPU performance was potentially limited due to Core computation issues (non divider-related). Two distinct categories can be attributed into this metric: (1) heavy data-dependency among contiguous instructions would manifest in this metric - such cases are often referred to as low Instruction Level Parallelism (ILP). (2) Contention on some hardware execution unit other than Divider. For example; when there are too many multiply operations.. Loop Vectorization -most compilers feature auto-Vectorization options today- reduces pressure on the execution ports as multiple elements are calculated with same uop.""" class Retiring: name = "Retiring" domain = "Slots" area = "RET" level = 1 htoff = False sample = ['UOPS_RETIRED.RETIRE_SLOTS'] errcount = 0 sibling = None metricgroup = frozenset(['BvUW', 'TmaL1']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 1) / SLOTS(self, EV, 1) self.thresh = (self.val > 0.7) or self.Heavy_Operations.thresh except ZeroDivisionError: handle_error(self, "Retiring zero division") return self.val desc = """ This category represents fraction of slots utilized by useful work i.e. issued uops that eventually get retired. Ideally; all pipeline slots would be attributed to the Retiring category. Retiring of 100% would indicate the maximum Pipeline_Width throughput was achieved. Maximizing Retiring typically increases the Instructions-per-cycle (see IPC metric). Note that a high Retiring value does not necessary mean there is no room for more performance. For example; Heavy-operations or Microcode Assists are categorized under Retiring. They often indicate suboptimal performance and can often be optimized or avoided. . A high Retiring value for non-vectorized code may be a good hint for programmer to consider vectorizing his code. Doing so essentially lets more computations be done without significantly increasing number of instructions thus improving the performance.""" class Light_Operations: name = "Light_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = ['INST_RETIRED.PREC_DIST'] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Retiring.compute(EV) - self.Heavy_Operations.compute(EV) self.thresh = (self.val > 0.6) except ZeroDivisionError: handle_error(self, "Light_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring light-weight operations -- instructions that require no more than one uop (micro-operation). This correlates with total number of instructions used by the program. A uops-per-instruction (see UopPI metric) ratio of 1 or less should be expected for decently optimized code running on Intel Core/Xeon products. While this often indicates efficient X86 instructions were executed; high value does not necessarily mean better performance cannot be achieved. . Focus on techniques that reduce instruction count or result in more efficient instructions generation such as vectorization.""" class FP_Arith: name = "FP_Arith" domain = "Uops" area = "RET" level = 3 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['HPC']) maxval = None def compute(self, EV): try: self.val = self.X87_Use.compute(EV) + self.FP_Scalar.compute(EV) + self.FP_Vector.compute(EV) self.thresh = (self.val > 0.2) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Arith zero division") return self.val desc = """ This metric represents overall arithmetic floating-point (FP) operations fraction the CPU has executed (retired). Note this metric's value may exceed its parent due to use of \"Uops\" CountDomain and FMA double-counting.""" class X87_Use: name = "X87_Use" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute']) maxval = None def compute(self, EV): try: self.val = Retired_Slots(self, EV, 4) * EV("FP_COMP_OPS_EXE.X87", 4) / EV("UOPS_DISPATCHED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "X87_Use zero division") return self.val desc = """ This metric serves as an approximation of legacy x87 usage. It accounts for instructions beyond X87 FP arithmetic operations; hence may be used as a thermometer to avoid X87 high usage and preferably upgrade to modern ISA. See Tip under Tuning Hint.. Tip: consider compiler flags to generate newer AVX (or SSE) instruction sets; which typically perform better and feature vectors.""" class FP_Scalar: name = "FP_Scalar" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = None def compute(self, EV): try: self.val = FP_Arith_Scalar(self, EV, 4) / EV("UOPS_DISPATCHED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Scalar zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) scalar uops fraction the CPU has retired. May overcount due to FMA double counting.. Investigate what limits (compiler) generation of vector code.""" class FP_Vector: name = "FP_Vector" domain = "Uops" area = "RET" level = 4 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = FP_Arith_Vector(self, EV, 4) / EV("UOPS_DISPATCHED.THREAD", 4) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector zero division") return self.val desc = """ This metric approximates arithmetic floating-point (FP) vector uops fraction the CPU has retired aggregated across all vector widths. May overcount due to FMA double counting.. Check if vector width is expected""" class FP_Vector_128b: name = "FP_Vector_128b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("FP_COMP_OPS_EXE.SSE_SCALAR_DOUBLE", 5) + EV("FP_COMP_OPS_EXE.SSE_PACKED_DOUBLE", 5)) / EV("UOPS_DISPATCHED.THREAD", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_128b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 128-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class FP_Vector_256b: name = "FP_Vector_256b" domain = "Uops" area = "RET" level = 5 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Compute', 'Flops']) maxval = 1.0 def compute(self, EV): try: self.val = (EV("SIMD_FP_256.PACKED_DOUBLE", 5) + EV("SIMD_FP_256.PACKED_SINGLE", 5)) / EV("UOPS_DISPATCHED.THREAD", 5) self.thresh = (self.val > 0.1) and self.parent.thresh except ZeroDivisionError: handle_error(self, "FP_Vector_256b zero division") return self.val desc = """ This metric approximates arithmetic FP vector uops fraction the CPU has retired for 256-bit wide vectors. May overcount due to FMA double counting.. Try to exploit wider vector length""" class Heavy_Operations: name = "Heavy_Operations" domain = "Slots" area = "RET" level = 2 htoff = False sample = [] errcount = 0 sibling = None metricgroup = frozenset(['Retire', 'TmaL2']) maxval = None def compute(self, EV): try: self.val = self.Microcode_Sequencer.compute(EV) self.thresh = (self.val > 0.1) except ZeroDivisionError: handle_error(self, "Heavy_Operations zero division") return self.val desc = """ This metric represents fraction of slots where the CPU was retiring heavy-weight operations -- instructions that require two or more uops or micro-coded sequences. This highly-correlates with the uop length of these instructions/sequences.""" class Microcode_Sequencer: name = "Microcode_Sequencer" domain = "Slots" area = "RET" level = 3 htoff = False sample = ['IDQ.MS_UOPS'] errcount = 0 sibling = None metricgroup = frozenset(['MicroSeq']) maxval = None def compute(self, EV): try: self.val = Retire_Fraction(self, EV, 3) * EV("IDQ.MS_UOPS", 3) / SLOTS(self, EV, 3) self.thresh = (self.val > 0.05) and self.parent.thresh except ZeroDivisionError: handle_error(self, "Microcode_Sequencer zero division") return self.val desc = """ This metric represents fraction of slots the CPU was retiring uops fetched by the Microcode Sequencer (MS) unit. The MS is used for CISC instructions not supported by the default decoders (like repeat move strings; or CPUID); or by microcode assists used to address some operation modes (like in Floating Point assists). These cases can often be avoided..""" class Metric_IPC: name = "IPC" domain = "Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Ret', 'Summary']) sibling = None def compute(self, EV): try: self.val = IPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "IPC zero division") desc = """ Instructions Per Cycle (per Logical Processor)""" class Metric_UopPI: name = "UopPI" domain = "Metric" maxval = 2.0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Ret', 'Retire']) sibling = None def compute(self, EV): try: self.val = UopPI(self, EV, 0) self.thresh = (self.val > 1.05) except ZeroDivisionError: handle_error_metric(self, "UopPI zero division") desc = """ Uops Per Instruction""" class Metric_CPI: name = "CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline', 'Mem']) sibling = None def compute(self, EV): try: self.val = CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPI zero division") desc = """ Cycles Per Instruction (per Logical Processor)""" class Metric_CLKS: name = "CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Pipeline']) sibling = None def compute(self, EV): try: self.val = CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CLKS zero division") desc = """ Per-Logical Processor actual clocks when the Logical Processor is active.""" class Metric_SLOTS: name = "SLOTS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['TmaL1']) sibling = None def compute(self, EV): try: self.val = SLOTS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SLOTS zero division") desc = """ Total issue-pipeline slots (per-Physical Core till ICL; per- Logical Processor ICL onward)""" class Metric_Execute_per_Issue: name = "Execute_per_Issue" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Thread" metricgroup = frozenset(['Cor', 'Pipeline']) sibling = None def compute(self, EV): try: self.val = Execute_per_Issue(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Execute_per_Issue zero division") desc = """ The ratio of Executed- by Issued-Uops. Ratio > 1 suggests high rate of uop micro-fusions. Ratio < 1 suggest high rate of \"execute\" at rename stage.""" class Metric_CoreIPC: name = "CoreIPC" domain = "Core_Metric" maxval = Pipeline_Width + 2 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'SMT', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = CoreIPC(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CoreIPC zero division") desc = """ Instructions Per Cycle across hyper-threads (per physical core)""" class Metric_FLOPc: name = "FLOPc" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['Ret', 'Flops']) sibling = None def compute(self, EV): try: self.val = FLOPc(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "FLOPc zero division") desc = """ Floating Point Operations Per Cycle""" class Metric_ILP: name = "ILP" domain = "Metric" maxval = Exe_Ports errcount = 0 area = "Info.Core" metricgroup = frozenset(['Backend', 'Cor', 'Pipeline', 'PortsUtil']) sibling = None def compute(self, EV): try: self.val = ILP(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "ILP zero division") desc = """ Instruction-Level-Parallelism (average number of uops executed when there is execution) per thread (logical- processor)""" class Metric_CORE_CLKS: name = "CORE_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.Core" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = CORE_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CORE_CLKS zero division") desc = """ Core actual clocks when any Logical Processor is active on the Physical Core""" class Metric_Instructions: name = "Instructions" domain = "Count" maxval = 0 errcount = 0 area = "Info.Inst_Mix" metricgroup = frozenset(['Summary', 'TmaL1']) sibling = None def compute(self, EV): try: self.val = Instructions(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Instructions zero division") desc = """ Total number of retired Instructions""" class Metric_Retire: name = "Retire" domain = "Metric" maxval = 0 errcount = 0 area = "Info.Pipeline" metricgroup = frozenset(['Pipeline', 'Ret']) sibling = None def compute(self, EV): try: self.val = Retire(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Retire zero division") desc = """ Average number of Uops retired in cycles where at least one uop has retired.""" class Metric_DSB_Coverage: name = "DSB_Coverage" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.Frontend" metricgroup = frozenset(['DSB', 'Fed', 'FetchBW']) sibling = None def compute(self, EV): try: self.val = DSB_Coverage(self, EV, 0) self.thresh = (self.val < 0.7) and HighIPC(self, EV, 1) except ZeroDivisionError: handle_error_metric(self, "DSB_Coverage zero division") desc = """ Fraction of Uops delivered by the DSB (aka Decoded ICache; or Uop Cache). See section 'Decoded ICache' in Optimization Manual. http://www.intel.com/content/www/us/en/architecture- and-technology/64-ia-32-architectures-optimization- manual.html""" class Metric_CPU_Utilization: name = "CPU_Utilization" domain = "Metric" maxval = 1 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'Summary']) sibling = None def compute(self, EV): try: self.val = CPU_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPU_Utilization zero division") desc = """ Average CPU Utilization (percentage)""" class Metric_CPUs_Utilized: name = "CPUs_Utilized" domain = "Metric" maxval = 300 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = CPUs_Utilized(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "CPUs_Utilized zero division") desc = """ Average number of utilized CPUs""" class Metric_Core_Frequency: name = "Core_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary', 'Power']) sibling = None def compute(self, EV): try: self.val = Core_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Core_Frequency zero division") desc = """ Measured Average Core Frequency for unhalted processors [GHz]""" class Metric_Uncore_Frequency: name = "Uncore_Frequency" domain = "SystemMetric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Uncore_Frequency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Uncore_Frequency zero division") desc = """ Measured Average Uncore Frequency for the SoC [GHz]""" class Metric_GFLOPs: name = "GFLOPs" domain = "Metric" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['Cor', 'Flops', 'HPC']) sibling = None def compute(self, EV): try: self.val = GFLOPs(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "GFLOPs zero division") desc = """ Giga Floating Point Operations Per Second. Aggregate across all supported options of: FP precisions, scalar and vector instructions, vector-width""" class Metric_Turbo_Utilization: name = "Turbo_Utilization" domain = "Core_Metric" maxval = 10.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Power']) sibling = None def compute(self, EV): try: self.val = Turbo_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Turbo_Utilization zero division") desc = """ Average Frequency Utilization relative nominal frequency""" class Metric_SMT_2T_Utilization: name = "SMT_2T_Utilization" domain = "Core_Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SMT']) sibling = None def compute(self, EV): try: self.val = SMT_2T_Utilization(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "SMT_2T_Utilization zero division") desc = """ Fraction of cycles where both hardware Logical Processors were active""" class Metric_Kernel_Utilization: name = "Kernel_Utilization" domain = "Metric" maxval = 1.0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_Utilization(self, EV, 0) self.thresh = (self.val > 0.05) except ZeroDivisionError: handle_error_metric(self, "Kernel_Utilization zero division") desc = """ Fraction of cycles spent in the Operating System (OS) Kernel mode""" class Metric_Kernel_CPI: name = "Kernel_CPI" domain = "Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['OS']) sibling = None def compute(self, EV): try: self.val = Kernel_CPI(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Kernel_CPI zero division") desc = """ Cycles Per Instruction for the Operating System (OS) Kernel mode""" class Metric_DRAM_BW_Use: name = "DRAM_BW_Use" domain = "GB/sec" maxval = 200 errcount = 0 area = "Info.System" metricgroup = frozenset(['HPC', 'MemOffcore', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = DRAM_BW_Use(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "DRAM_BW_Use zero division") desc = """ Average external Memory Bandwidth Use for reads and writes [GB / sec]""" class Metric_MEM_Read_Latency: name = "MEM_Read_Latency" domain = "NanoSeconds" maxval = 1000 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryLat', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Read_Latency(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Read_Latency zero division") desc = """ Average latency of data read request to external memory (in nanoseconds). Accounts for demand loads and L1/L2 prefetches.""" class Metric_MEM_Parallel_Reads: name = "MEM_Parallel_Reads" domain = "SystemMetric" maxval = 100 errcount = 0 area = "Info.System" metricgroup = frozenset(['Mem', 'MemoryBW', 'SoC']) sibling = None def compute(self, EV): try: self.val = MEM_Parallel_Reads(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "MEM_Parallel_Reads zero division") desc = """ Average number of parallel data read requests to external memory. Accounts for demand loads and L1/L2 prefetches""" class Metric_Time: name = "Time" domain = "Seconds" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Summary']) sibling = None def compute(self, EV): try: self.val = Time(self, EV, 0) self.thresh = (self.val < 1) except ZeroDivisionError: handle_error_metric(self, "Time zero division") desc = """ Run duration time in seconds""" class Metric_Socket_CLKS: name = "Socket_CLKS" domain = "Count" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['SoC']) sibling = None def compute(self, EV): try: self.val = Socket_CLKS(self, EV, 0) self.thresh = True except ZeroDivisionError: handle_error_metric(self, "Socket_CLKS zero division") desc = """ Socket actual clocks when any core is active on that socket""" class Metric_IpFarBranch: name = "IpFarBranch" domain = "Inst_Metric" maxval = 0 errcount = 0 area = "Info.System" metricgroup = frozenset(['Branches', 'OS']) sibling = None def compute(self, EV): try: self.val = IpFarBranch(self, EV, 0) self.thresh = (self.val < 1000000) except ZeroDivisionError: handle_error_metric(self, "IpFarBranch zero division") desc = """ Instructions per Far Branch ( Far Branches apply upon transition from application to operating system, handling interrupts, exceptions) [lower number means higher occurrence rate]""" # Schedule class Setup: def __init__(self, r): o = dict() n = Frontend_Bound() ; r.run(n) ; o["Frontend_Bound"] = n n = Fetch_Latency() ; r.run(n) ; o["Fetch_Latency"] = n n = ITLB_Misses() ; r.run(n) ; o["ITLB_Misses"] = n n = Branch_Resteers() ; r.run(n) ; o["Branch_Resteers"] = n n = MS_Switches() ; r.run(n) ; o["MS_Switches"] = n n = LCP() ; r.run(n) ; o["LCP"] = n n = DSB_Switches() ; r.run(n) ; o["DSB_Switches"] = n n = Fetch_Bandwidth() ; r.run(n) ; o["Fetch_Bandwidth"] = n n = Bad_Speculation() ; r.run(n) ; o["Bad_Speculation"] = n n = Branch_Mispredicts() ; r.run(n) ; o["Branch_Mispredicts"] = n n = Machine_Clears() ; r.run(n) ; o["Machine_Clears"] = n n = Backend_Bound() ; r.run(n) ; o["Backend_Bound"] = n n = Memory_Bound() ; r.run(n) ; o["Memory_Bound"] = n n = DTLB_Load() ; r.run(n) ; o["DTLB_Load"] = n n = L3_Bound() ; r.run(n) ; o["L3_Bound"] = n n = DRAM_Bound() ; r.run(n) ; o["DRAM_Bound"] = n n = MEM_Bandwidth() ; r.run(n) ; o["MEM_Bandwidth"] = n n = MEM_Latency() ; r.run(n) ; o["MEM_Latency"] = n n = Store_Bound() ; r.run(n) ; o["Store_Bound"] = n n = Core_Bound() ; r.run(n) ; o["Core_Bound"] = n n = Divider() ; r.run(n) ; o["Divider"] = n n = Ports_Utilization() ; r.run(n) ; o["Ports_Utilization"] = n n = Retiring() ; r.run(n) ; o["Retiring"] = n n = Light_Operations() ; r.run(n) ; o["Light_Operations"] = n n = FP_Arith() ; r.run(n) ; o["FP_Arith"] = n n = X87_Use() ; r.run(n) ; o["X87_Use"] = n n = FP_Scalar() ; r.run(n) ; o["FP_Scalar"] = n n = FP_Vector() ; r.run(n) ; o["FP_Vector"] = n n = FP_Vector_128b() ; r.run(n) ; o["FP_Vector_128b"] = n n = FP_Vector_256b() ; r.run(n) ; o["FP_Vector_256b"] = n n = Heavy_Operations() ; r.run(n) ; o["Heavy_Operations"] = n n = Microcode_Sequencer() ; r.run(n) ; o["Microcode_Sequencer"] = n # parents o["Fetch_Latency"].parent = o["Frontend_Bound"] o["ITLB_Misses"].parent = o["Fetch_Latency"] o["Branch_Resteers"].parent = o["Fetch_Latency"] o["MS_Switches"].parent = o["Fetch_Latency"] o["LCP"].parent = o["Fetch_Latency"] o["DSB_Switches"].parent = o["Fetch_Latency"] o["Fetch_Bandwidth"].parent = o["Frontend_Bound"] o["Branch_Mispredicts"].parent = o["Bad_Speculation"] o["Machine_Clears"].parent = o["Bad_Speculation"] o["Memory_Bound"].parent = o["Backend_Bound"] o["DTLB_Load"].parent = o["Memory_Bound"] o["L3_Bound"].parent = o["Memory_Bound"] o["DRAM_Bound"].parent = o["Memory_Bound"] o["MEM_Bandwidth"].parent = o["DRAM_Bound"] o["MEM_Latency"].parent = o["DRAM_Bound"] o["Store_Bound"].parent = o["Memory_Bound"] o["Core_Bound"].parent = o["Backend_Bound"] o["Divider"].parent = o["Core_Bound"] o["Ports_Utilization"].parent = o["Core_Bound"] o["Light_Operations"].parent = o["Retiring"] o["FP_Arith"].parent = o["Light_Operations"] o["X87_Use"].parent = o["FP_Arith"] o["FP_Scalar"].parent = o["FP_Arith"] o["FP_Vector"].parent = o["FP_Arith"] o["FP_Vector_128b"].parent = o["FP_Vector"] o["FP_Vector_256b"].parent = o["FP_Vector"] o["Heavy_Operations"].parent = o["Retiring"] o["Microcode_Sequencer"].parent = o["Heavy_Operations"] # user visible metrics n = Metric_IPC() ; r.metric(n) ; o["IPC"] = n n = Metric_UopPI() ; r.metric(n) ; o["UopPI"] = n n = Metric_CPI() ; r.metric(n) ; o["CPI"] = n n = Metric_CLKS() ; r.metric(n) ; o["CLKS"] = n n = Metric_SLOTS() ; r.metric(n) ; o["SLOTS"] = n n = Metric_Execute_per_Issue() ; r.metric(n) ; o["Execute_per_Issue"] = n n = Metric_CoreIPC() ; r.metric(n) ; o["CoreIPC"] = n n = Metric_FLOPc() ; r.metric(n) ; o["FLOPc"] = n n = Metric_ILP() ; r.metric(n) ; o["ILP"] = n n = Metric_CORE_CLKS() ; r.metric(n) ; o["CORE_CLKS"] = n n = Metric_Instructions() ; r.metric(n) ; o["Instructions"] = n n = Metric_Retire() ; r.metric(n) ; o["Retire"] = n n = Metric_DSB_Coverage() ; r.metric(n) ; o["DSB_Coverage"] = n n = Metric_CPU_Utilization() ; r.metric(n) ; o["CPU_Utilization"] = n n = Metric_CPUs_Utilized() ; r.metric(n) ; o["CPUs_Utilized"] = n n = Metric_Core_Frequency() ; r.metric(n) ; o["Core_Frequency"] = n n = Metric_Uncore_Frequency() ; r.metric(n) ; o["Uncore_Frequency"] = n n = Metric_GFLOPs() ; r.metric(n) ; o["GFLOPs"] = n n = Metric_Turbo_Utilization() ; r.metric(n) ; o["Turbo_Utilization"] = n n = Metric_SMT_2T_Utilization() ; r.metric(n) ; o["SMT_2T_Utilization"] = n n = Metric_Kernel_Utilization() ; r.metric(n) ; o["Kernel_Utilization"] = n n = Metric_Kernel_CPI() ; r.metric(n) ; o["Kernel_CPI"] = n n = Metric_DRAM_BW_Use() ; r.metric(n) ; o["DRAM_BW_Use"] = n n = Metric_MEM_Read_Latency() ; r.metric(n) ; o["MEM_Read_Latency"] = n n = Metric_MEM_Parallel_Reads() ; r.metric(n) ; o["MEM_Parallel_Reads"] = n n = Metric_Time() ; r.metric(n) ; o["Time"] = n n = Metric_Socket_CLKS() ; r.metric(n) ; o["Socket_CLKS"] = n n = Metric_IpFarBranch() ; r.metric(n) ; o["IpFarBranch"] = n # references between groups o["Fetch_Bandwidth"].Frontend_Bound = o["Frontend_Bound"] o["Fetch_Bandwidth"].Fetch_Latency = o["Fetch_Latency"] o["Branch_Mispredicts"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Bad_Speculation = o["Bad_Speculation"] o["Machine_Clears"].Branch_Mispredicts = o["Branch_Mispredicts"] o["Backend_Bound"].Retiring = o["Retiring"] o["Backend_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Backend_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Retiring = o["Retiring"] o["Memory_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Memory_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Memory_Bound"].Backend_Bound = o["Backend_Bound"] o["Memory_Bound"].Fetch_Latency = o["Fetch_Latency"] o["MEM_Latency"].MEM_Bandwidth = o["MEM_Bandwidth"] o["Core_Bound"].Retiring = o["Retiring"] o["Core_Bound"].Frontend_Bound = o["Frontend_Bound"] o["Core_Bound"].Memory_Bound = o["Memory_Bound"] o["Core_Bound"].Backend_Bound = o["Backend_Bound"] o["Core_Bound"].Bad_Speculation = o["Bad_Speculation"] o["Core_Bound"].Fetch_Latency = o["Fetch_Latency"] o["Ports_Utilization"].Fetch_Latency = o["Fetch_Latency"] o["Retiring"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Retiring = o["Retiring"] o["Light_Operations"].Heavy_Operations = o["Heavy_Operations"] o["Light_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] o["FP_Arith"].FP_Scalar = o["FP_Scalar"] o["FP_Arith"].X87_Use = o["X87_Use"] o["FP_Arith"].FP_Vector = o["FP_Vector"] o["Heavy_Operations"].Microcode_Sequencer = o["Microcode_Sequencer"] # siblings cross-tree o["MS_Switches"].sibling = (o["Machine_Clears"], o["Microcode_Sequencer"],) o["LCP"].sibling = (o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DSB_Switches"].sibling = (o["LCP"], o["Fetch_Bandwidth"],) o["Fetch_Bandwidth"].sibling = (o["LCP"], o["DSB_Switches"],) o["Machine_Clears"].sibling = (o["MS_Switches"], o["Microcode_Sequencer"],) o["FP_Scalar"].sibling = (o["FP_Vector"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector"].sibling = (o["FP_Scalar"], o["FP_Vector_128b"], o["FP_Vector_256b"],) o["FP_Vector_128b"].sibling = (o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_256b"],) o["FP_Vector_256b"].sibling = (o["FP_Scalar"], o["FP_Vector"], o["FP_Vector_128b"],) o["Microcode_Sequencer"].sibling = (o["MS_Switches"], o["Machine_Clears"],) o["DSB_Coverage"].sibling = (o["LCP"], o["DSB_Switches"], o["Fetch_Bandwidth"],) o["DRAM_BW_Use"].sibling = (o["MEM_Bandwidth"],)
67,078
Python
.py
1,671
34.648713
306
0.666247
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,896
interval-normalize.py
andikleen_pmu-tools/interval-normalize.py
#!/usr/bin/env python3 # convert perf stat -Ixxx -x, / toplev -Ixxx -x, output to normalized output # this version buffers all data in memory, so it can use a lot of memory. # t1,ev1,num1 # t1,ev2,num1 # t2,ev1,num3 # -> # timestamp,ev1,ev2 # t1,num1,num2 # t2,num3,, # when the input has CPU generate separate lines for each CPU (may need post filtering) from __future__ import print_function import sys import csv import argparse import collections import csv_formats ap = argparse.ArgumentParser(description= 'Normalize CSV data from perf or toplev. All values are printed on a single line.') ap.add_argument('inputfile', type=argparse.FileType('r'), default=sys.stdin, nargs='?') ap.add_argument('--output', '-o', type=argparse.FileType('w'), default=sys.stdout, nargs='?') ap.add_argument('--cpu', nargs='?', help='Only output for this cpu') ap.add_argument('--na', nargs='?', help='Value to use if data is not available', default="") ap.add_argument('--error-exit', action='store_true', help='Force error exit on parse error') ap.add_argument('--normalize-cpu', action='store_true', help='Normalize CPUs into unique columns too') args = ap.parse_args() printed_header = False timestamp = None events = collections.OrderedDict() out = [] times = [] cpus = [] rc = csv.reader(args.inputfile) res = [] writer = csv.writer(args.output, lineterminator='\n') lastcpu = None cpu = None lineno = 1 for row in rc: if len(row) > 0 and (row[0] == "Timestamp" or row[0].startswith("#")): lineno += 1 continue r = csv_formats.parse_csv_row(row, error_exit=args.error_exit) if r is None: print("at line %d" % lineno, file=sys.stderr) lineno += 1 continue ts, cpu, ev, val = r.ts, r.cpu, r.ev, r.val if ts != timestamp or (cpu != lastcpu and not args.normalize_cpu): if timestamp: if args.cpu and cpu != args.cpu: continue # delay in case we didn't see all headers # only need to do that for toplev, directly output for perf? # could limit buffering to save memory? out.append(res) times.append(timestamp) cpus.append(cpu) res = [] timestamp = ts lastcpu = cpu if cpu is not None and args.normalize_cpu: ev = cpu + " " + ev # use a list for row storage to keep memory requirements down if ev not in events: events[ev] = len(res) ind = events[ev] if ind >= len(res): res += [None] * ((ind + 1) - len(res)) res[ind] = val lineno += 1 if res and not (args.cpu and cpu != args.cpu): out.append(res) times.append(timestamp) cpus.append(cpu) def resolve(row, ind): if ind >= len(row): return args.na v = row[ind] if v is None: return args.na return v def cpulist(): if args.normalize_cpu: return [] if cpu is not None: return ["CPU"] return [] keys = events.keys() writer.writerow(["Timestamp"] + cpulist() + list(keys)) for row, ts, cpunum in zip(out, times, cpus): writer.writerow([ts] + ([cpunum] if (cpu is not None and not args.normalize_cpu) else []) + ([resolve(row, events[x]) for x in keys]))
3,252
Python
.py
94
29.712766
102
0.639568
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,897
tsx_metrics.py
andikleen_pmu-tools/tsx_metrics.py
# # TSX metrics # # XXX force all these into a single group # XXX: force % in caller import os def TXCycles(EV, level): return EV("cpu/cycles-t/", level) / EV("cycles", level) class TransactionalCycles: name = "Transactional cycles" desc = """ Percent cycles spent in a transaction. When low or zero either the program does not use locks (or other transactions), or the locks are not enabled with lock elision.""" subplot = "TSX" unit = "%" sample = ["mem_uops_retired.lock_loads"] server = True def compute(self, EV): try: self.val = TXCycles(EV, 1) * 100. self.thresh = (self.val >= 0.01) except ZeroDivisionError: self.val = 0 self.thresh = False class AbortedCycles: name = "Aborted cycles" desc = """ Percent cycles wasted in transaction aborts. When a significant part of the transactional cycles start sampling for abort causes.""" subplot = "TSX" unit = "%" sample = ["cpu/tx-abort/pp", "cpu/hle-abort/pp"] server = True def compute(self, EV): try: self.val = ((EV("cpu/cycles-t/", 1) - EV("cpu/cycles-ct/", 1)) / EV("cycles", 1)) * 100. self.thresh = (self.val >= 0.01) except ZeroDivisionError: self.val = 0 self.thresh = False class AverageRTM: name = "Average RTM transaction length" desc = """ Average RTM transaction length. Assumes most transactions are RTM. When low consider increasing the size of the critical sections to lower overhead.""" subplot = "TSX Latencies" unit = "cycles" server = True def compute(self, EV): try: self.val = EV("cpu/cycles-t/", 1) / EV("RTM_RETIRED.START", 1) self.thresh = TXCycles(EV, 1) >= 0.01 and self.val > 0 except ZeroDivisionError: self.val = 0 self.thresh = False class AverageHLE: name = "Average HLE transaction length" desc = """ Average HLE transaction length. Assumes most transactions are HLE. When low consider increasing the size of the critical sections to lower overhead.""" subplot = "TSX Latencies" unit = "cycles" def compute(self, EV): try: self.val = EV("cpu/cycles-t/", 1) / EV("HLE_RETIRED.START", 1) self.thresh = TXCycles(EV, 1) >= 0.01 and self.val > 0 except ZeroDivisionError: self.val = 0 self.thresh = False class Setup: def __init__(self, r): # XXX allow override if os.path.exists("/sys/devices/cpu/events/cycles-t"): r.force_metric(TransactionalCycles()) r.force_metric(AbortedCycles()) r.force_metric(AverageRTM()) #r.force_metric(AverageHLE())
2,765
Python
.py
77
29.246753
100
0.622015
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,898
genretlat.py
andikleen_pmu-tools/genretlat.py
#!/usr/bin/env python3 # -*- coding: utf-8 # generate return latencies to tune toplev model # Copyright (c) 2023, Intel Corporation # Author: Andi Kleen # # This program is free software; you can redistribute it and/or modify it # under the terms and conditions of the GNU General Public License, # version 2, as published by the Free Software Foundation. # # This program is distributed in the hope it will be useful, but WITHOUT # ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or # FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for # more details. from __future__ import print_function import subprocess as subp import os import sys import ocperf import json import argparse import random from dummyarith import DummyArith from collections import Counter, defaultdict from copy import copy import csv def cmd(args, l): if args.verbose: print(" ".join(l)) return l class SamplePerfRun(object): """Run perf record to collect Timed PEBS information and generate averages for event weights.""" def __init__(self, args): self.pi = self.pr = self.ps = None self.args = args def execute(self, perf, evsamples, origevents, pargs): self.perf = perf self.rmap = { x: y for x, y in zip(evsamples, origevents) } pr = subp.Popen(cmd(self.args, [perf, "record", "-W", "-B", "-c", "%d" % self.args.interval, "-e", ",".join(evsamples), "-o", "-"] + pargs), stdout=subp.PIPE) pi = subp.Popen(cmd(self.args, [perf, "inject"]), stdin=pr.stdout, stdout=subp.PIPE) ps = subp.Popen(cmd(self.args, [perf, "script", "-i", "-", "--ns", "-F", "time,retire_lat,event"]), stdin=pi.stdout, stdout=subp.PIPE) pi.stdout.close() pr.stdout.close() self.pr, self.pi, self.ps = pr, pi, ps def handle_samples(self): for l in self.ps.stdout: l = l.decode() n = l.split() ts, event, weight = float(n[0].replace(":","")), n[1].replace(":",""), int(n[2]) if self.args.csvplot: self.args.csvplot.writerow([ts, self.rmap[event] if event in self.rmap else event, weight]) if self.args.fake: yield event, random.randint(0, 100) else: yield event, weight self.pr.wait() self.pi.wait() self.ps.wait() NUM_SPARKS = 8 SPARK_SHIFT = 8 def spark_f(l): min_ = min(l) f = int(max(((max(l) - min_) << SPARK_SHIFT) / (NUM_SPARKS - 1), 1)) return f, min_ # this samples unlike normal spark lines def gen_spark_buckets(l): if len(l) == 0: return [], 0, 1 f, min_ = spark_f(l) return [((x - min_) << SPARK_SHIFT) / f for x in random.sample(l, min(NUM_SPARKS, len(l)))], min_, f def lookup(s, i, fb): if 0 <= i <= len(s) - 1: return s[i] return fb def gen_spark(buckets, min_, f): return "".join([lookup("▁▂▃▄▅▆▇█", int((int((x - min_)) << SPARK_SHIFT) / f), " ") for x in buckets]) def gen_stat(samples): # { # "COUNT": 5358917, # "MIN": 0, # "MAX": 65535, # "MEAN": 3.23, # "MEDIAN": 0, # "NZ_MEDIAN": 1, # "MODE": 0, # "MODE_COUNT": 3631698, # "NZ_MODE": 1, # "NZ_MODE_COUNT": 1213029, # "BUCKETS": 2344 # }, nz = [x for x in samples if x != 0.0] buckets = Counter(samples) nz_buckets = copy(buckets) if 0 in nz_buckets: del nz_buckets[0] spark, min_, f = gen_spark_buckets(samples) spark_nz, min_nz, f_nz = gen_spark_buckets(nz) return { "COUNT": len(samples), "MIN": min(samples), "MAX": max(samples), "MEAN": round(float(sum(samples)) / len(samples), 2), "MEDIAN": sorted(samples)[len(samples)//2] if len(samples) > 0 else 0.0, "NZ_MEDIAN": sorted(nz)[len(nz)//2] if len(nz) > 0 else 0.0, "MODE": buckets.most_common(1)[0][0] if len(buckets) > 0 else 0.0, "MODE_COUNT": buckets.most_common(1)[0][1] if len(buckets) > 0 else 0, "NZ_MODE": nz_buckets.most_common(1)[0][0] if len(nz_buckets) > 0 else 0, "NZ_MODE_COUNT": nz_buckets.most_common(1)[0][1] if len(nz_buckets) > 0 else 0, "BUCKETS": len(buckets), "F": f, "F_NZ": f_nz, "MIN_NZ": min_nz, "SPARK_BUCKETS": ",".join(["%d" % x for x in spark]), "SPARK_BUCKETS_NZ": ",".join(["%d" % x for x in spark_nz]), } def human_output(data): d = data["Data"] for ev in sorted(d.keys()): print("%s: " % ev, end="") for m in sorted(d[ev].keys()): if m.startswith("SPARK"): if d[ev][m] == "": continue l = [(int(x) if x.isdecimal() else 0) for x in d[ev][m].split(",")] s = gen_spark(l, d[ev]["MIN"], d[ev]["F_NZ" if m.endswith("_NZ") else "F"]) print("%s %s " % (m.lower(), s), end="") else: print("%s %s " % (m.lower(), d[ev][m]), end="") print() def get_model_number(): with open("/proc/cpuinfo") as f: for l in f: n = l.split() if len(n) >= 3 and n[0] == "model" and n[1] == ":": return int(n[2]) return 0 def find_model(args): if not args.cpu: cpu = open("/sys/devices/cpu/caps/pmu_name").read().strip() if cpu == "meteorlake_hybrid": args.cpu = "mtl" elif cpu == "sapphire_rapids": model = get_model_number() if model == 0xad or model == 0xae: args.cpu = "gnr" else: sys.exit("Unsupported CPU %d" % model) elif cpu == "granite_rapids": args.cpu = "gnr" else: sys.exit("Unsupported CPU %s" % cpu) if args.cpu == "mtl": import mtl_rwc_ratios return mtl_rwc_ratios elif args.cpu == "gnr": import gnr_server_ratios return gnr_server_ratios sys.exit("Unknown cpu %s" % args.cpu) def gen_events(args): model = find_model(args) nodes = [] class Runner(object): def metric(self, n): nodes.append(n) n.thresh = True def run(self, n): nodes.append(n) n.thresh = True model.Setup(Runner()) events = set() def collect(name, level): if level == 999: events.add(name + ":pp") return DummyArith() for n in nodes: n.compute(collect) return sorted(events) def getevent(emap, e): ev = emap.getevent(e) if ev is None: print(e, "not found") return "dummy" return ev.output() def init_args(): ap = argparse.ArgumentParser('Generate JSON of retirement latencies to tune toplev') ap.add_argument('--output', '-o', type=argparse.FileType('w'), default=sys.stdout, help="") ap.add_argument('--verbose', '-v', action='store_true', help='be verbose') ap.add_argument('--fake', action='store_true', help=argparse.SUPPRESS) ap.add_argument('--interval', '-i', type=int, default=103, help="Interval for sampling") ap.add_argument('--pmu', '-p', nargs='*', default=["cpu", "cpu_core"], help="for which PMUs to collect") ap.add_argument('--quiet', '-q', action='store_true') ap.add_argument('--csv', '-c', type=argparse.FileType('w'), help="Generate CSV file with pushout latencies", default=None) ap.add_argument('--cpu', help="Set CPU type (gnr, mtl)") args, rest = ap.parse_known_args() if args.csv: args.csvplot = csv.writer(args.csv) else: args.csvplot = None return args, rest def main(): args, rest = init_args() events = gen_events(args) pmus = ocperf.find_pmus() emap = ocperf.find_emap(pmu=[x for x in pmus if x != "cpu_atom"][0]) if emap is None: sys.exit("Cannot find json event list") pevents = [getevent(emap, e) for e in events] if args.verbose: print(events) s = SamplePerfRun(args) perf = os.getenv("PERF") if perf is None: perf = "perf" s.execute(perf, pevents, events, rest) samples = defaultdict(list) for ev, weight in s.handle_samples(): samples[ev].append(weight) data = { "Data": { ev.upper().replace("CPU_CORE","").replace("/","").replace(":","").replace("RETIRED_", "RETIRED."): gen_stat(s) for ev, s in samples.items() if "/" not in ev or any([x in ev for x in args.pmu]) } } json.dump(data, args.output, indent=2, sort_keys=True) if not args.quiet: human_output(data) if __name__ == '__main__': main()
9,118
Python
.py
239
29.485356
133
0.541629
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)
27,899
simple_ratios.py
andikleen_pmu-tools/simple_ratios.py
# # Simple 5 event top level model # print_error = lambda msg: False version = "1.0" check_event = lambda ev: False # Constants PipelineWidth = 4 uops_retired_slots = "UOPS_RETIRED.RETIRE_SLOTS" def CLKS(EV): return EV("CPU_CLK_UNHALTED.THREAD", 1) def SLOTS(EV): return PipelineWidth * CLKS(EV) # Instructions Per Cycle def IPC(EV, level): return EV("INST_RETIRED.ANY", level) / CLKS(EV) # Uops Per Instruction def UPI(EV, level): return EV(uops_retired_slots, level) / EV("INST_RETIRED.ANY", level) # Average Frequency Utilization relative nominal frequency def TurboUtilization(EV, level): return CLKS(EV) / EV("CPU_CLK_UNHALTED.REF_TSC", level) class FrontendBound: name = "Frontend Bound" domain = "Slots" desc = """ This category reflects slots where the Frontend of the processor undersupplies its Backend.""" level = 1 def compute(self, EV): try: self.val = EV("IDQ_UOPS_NOT_DELIVERED.CORE", 1) / SLOTS(EV) self.thresh = self.val > 0.2 except ZeroDivisionError: self.val = 0 self.thresh = False return self.val class BadSpeculation: name = "Bad Speculation" domain = "Slots" desc = """ This category reflects slots wasted due to incorrect speculations, which include slots used to allocate uops that do not eventually get retired and slots for which allocation was blocked due to recovery from earlier incorrect speculation. For example, wasted work due to miss-predicted branches is categorized under the Bad Speculation category""" level = 1 def compute(self, EV): try: self.val = ( EV("UOPS_ISSUED.ANY", 1) - EV(uops_retired_slots, 1) + PipelineWidth * EV("INT_MISC.RECOVERY_CYCLES", 1) ) / SLOTS(EV) self.thresh = self.val > 0.1 except ZeroDivisionError: self.val = 0 self.thresh = False return self.val class BackendBound: name = "Backend Bound" domain = "Slots" desc = """ This category reflects slots where no uops are being delivered due to a lack of required resources for accepting more uops in the Backend of the pipeline.""" level = 1 def compute(self, EV): try: self.val = 1 - ( self.FrontendBound.compute(EV) + self.BadSpeculation.compute(EV) + self.Retiring.compute(EV) ) self.thresh = self.val > 0.2 except ZeroDivisionError: self.val = 0 self.thresh = False return self.val class Retiring: name = "Retiring" domain = "Slots" desc = """ This category reflects slots utilized by good uops i.e. allocated uops that eventually get retired.""" level = 1 def compute(self, EV): try: self.val = EV(uops_retired_slots, 1) / SLOTS(EV) self.thresh = self.val > 0.7 except ZeroDivisionError: self.val = 0 self.thresh = False return self.val class Metric_IPC: name = "IPC" desc = """ Instructions Per Cycle""" errcount = 0 def compute(self, EV): try: self.val = IPC(EV, 0) except ZeroDivisionError: print_error("IPC zero division") self.errcount += 1 self.val = 0 class Metric_UPI: name = "UPI" desc = """ Uops Per Instruction""" errcount = 0 def compute(self, EV): try: self.val = UPI(EV, 0) except ZeroDivisionError: print_error("UPI zero division") self.errcount += 1 self.val = 0 class Metric_TurboUtilization: name = "TurboUtilization" desc = """ Average Frequency Utilization relative nominal frequency""" errcount = 0 def compute(self, EV): try: self.val = TurboUtilization(EV, 0) except ZeroDivisionError: print_error("TurboUtilization zero division") self.errcount += 1 self.val = 0 class Setup: def __init__(self, r): if check_event("UOPS_RETIRED.SLOTS"): global uops_retired_slots uops_retired_slots = "UOPS_RETIRED.SLOTS" o = dict() n = FrontendBound() ; r.run(n) o["FrontendBound"] = n n = BadSpeculation() ; r.run(n) o["BadSpeculation"] = n n = BackendBound() ; r.run(n) o["BackendBound"] = n n = Retiring() ; r.run(n) o["Retiring"] = n o["FrontendBound"].parent = None o["BadSpeculation"].parent = None o["BackendBound"].parent = None o["Retiring"].parent = None o["BackendBound"].FrontendBound = o["FrontendBound"] o["BackendBound"].BadSpeculation = o["BadSpeculation"] o["BackendBound"].Retiring = o["Retiring"] o["FrontendBound"].sibling = None o["BadSpeculation"].sibling = None o["BackendBound"].sibling = None o["Retiring"].sibling = None o["FrontendBound"].sample = [] o["BadSpeculation"].sample = [] o["BackendBound"].sample = [] o["Retiring"].sample = [] # user visible metrics n = Metric_IPC() ; r.metric(n) n = Metric_UPI() ; r.metric(n) n = Metric_TurboUtilization() ; r.metric(n)
5,235
Python
.py
154
27.058442
143
0.620574
andikleen/pmu-tools
1,984
331
178
GPL-2.0
9/5/2024, 5:13:50 PM (Europe/Amsterdam)