{
  "domain": "ai.onnx",
  "name": "MeanVarianceNormalization",
  "sinceVersion": 13,
  "inputs": { "x": { "onnx": "X", "dtype": "T" } },
  "outputs": { "y": { "onnx": "Y", "dtype": "T", "rank": "ranks.x", "shape": "shapes.x" } },
  "attributes": { "axes": { "default": [0, 2, 3] } },
  "typeConstraints": { "T": ["float32", "float16"] },
  "tunables": {
    "WORKGROUP_SIZE": { "default": 256 },
    "SERIAL_WORKGROUP_SIZE": { "default": 256 },
    "SERIAL_TINY_WORKGROUP_SIZE": { "default": 64 },
    "SERIAL_MAX_REDUCTION": { "default": 128 },
    "SERIAL_MIN_ROWS": { "default": 256 },
    "TREE_MEDIUM_WORKGROUP_SIZE": { "default": 64 },
    "VEC4_MIN_REDUCTION": { "default": 8 },
    "FLAT_SPLIT_MIN_ELEMENTS": { "default": 65536 },
    "FLAT_SPLIT_TARGET_ELEMENTS": { "default": 4096 },
    "MAX_FLAT_SPLITS": { "default": 256 }
  },
  "derive": {
    "deviceWorkgroupCap": "min(device.limits.maxComputeInvocationsPerWorkgroup, device.limits.maxComputeWorkgroupSizeX)",
    "foldedDispatchCapacity": "min(device.limits.maxComputeWorkgroupsPerDimension, 65535) * min(device.limits.maxComputeWorkgroupsPerDimension, 65535)",
    "shapeContract": "ranks.x >= 3 and ranks.x <= 8 and ranks.y == ranks.x and sameShape(shapes.y, shapes.x) and f16Ok(dtypes.T)",
    "reduceCount": "(dim(shapes.x, 0) if hasAxis(attrs.axes, 0, ranks.x) else 1) * (dim(shapes.x, 1) if hasAxis(attrs.axes, 1, ranks.x) else 1) * (dim(shapes.x, 2) if hasAxis(attrs.axes, 2, ranks.x) else 1) * (dim(shapes.x, 3) if ranks.x >= 4 and hasAxis(attrs.axes, 3, ranks.x) else 1) * (dim(shapes.x, 4) if ranks.x >= 5 and hasAxis(attrs.axes, 4, ranks.x) else 1) * (dim(shapes.x, 5) if ranks.x >= 6 and hasAxis(attrs.axes, 5, ranks.x) else 1) * (dim(shapes.x, 6) if ranks.x >= 7 and hasAxis(attrs.axes, 6, ranks.x) else 1) * (dim(shapes.x, 7) if ranks.x >= 8 and hasAxis(attrs.axes, 7, ranks.x) else 1)",
    "rowCount": "numel(shapes.x) / max(1, reduceCount)",
    "allAxesReduced": "hasAxis(attrs.axes, 0, ranks.x) and hasAxis(attrs.axes, 1, ranks.x) and hasAxis(attrs.axes, 2, ranks.x) and (ranks.x < 4 or hasAxis(attrs.axes, 3, ranks.x)) and (ranks.x < 5 or hasAxis(attrs.axes, 4, ranks.x)) and (ranks.x < 6 or hasAxis(attrs.axes, 5, ranks.x)) and (ranks.x < 7 or hasAxis(attrs.axes, 6, ranks.x)) and (ranks.x < 8 or hasAxis(attrs.axes, 7, ranks.x))",
    "vec4Eligible": "((ranks.x == 3 and hasAxis(attrs.axes, 2, 3) and (dim(shapes.x, 2) % 4 == 0 or (hasAxis(attrs.axes, 1, 3) and dim(shapes.x, 1) * dim(shapes.x, 2) % 4 == 0) or (hasAxis(attrs.axes, 0, 3) and hasAxis(attrs.axes, 1, 3) and numel(shapes.x) % 4 == 0))) or (ranks.x == 4 and hasAxis(attrs.axes, 3, 4) and (dim(shapes.x, 3) % 4 == 0 or (hasAxis(attrs.axes, 2, 4) and dim(shapes.x, 2) * dim(shapes.x, 3) % 4 == 0) or (hasAxis(attrs.axes, 1, 4) and hasAxis(attrs.axes, 2, 4) and dim(shapes.x, 1) * dim(shapes.x, 2) * dim(shapes.x, 3) % 4 == 0) or (hasAxis(attrs.axes, 0, 4) and hasAxis(attrs.axes, 1, 4) and hasAxis(attrs.axes, 2, 4) and numel(shapes.x) % 4 == 0))) or (ranks.x == 5 and hasAxis(attrs.axes, 4, 5) and (dim(shapes.x, 4) % 4 == 0 or (hasAxis(attrs.axes, 3, 5) and dim(shapes.x, 3) * dim(shapes.x, 4) % 4 == 0) or (hasAxis(attrs.axes, 2, 5) and hasAxis(attrs.axes, 3, 5) and dim(shapes.x, 2) * dim(shapes.x, 3) * dim(shapes.x, 4) % 4 == 0) or (hasAxis(attrs.axes, 1, 5) and hasAxis(attrs.axes, 2, 5) and hasAxis(attrs.axes, 3, 5) and dim(shapes.x, 1) * dim(shapes.x, 2) * dim(shapes.x, 3) * dim(shapes.x, 4) % 4 == 0) or (allAxesReduced and numel(shapes.x) % 4 == 0))) or (ranks.x == 6 and hasAxis(attrs.axes, 5, 6) and (dim(shapes.x, 5) % 4 == 0 or (hasAxis(attrs.axes, 4, 6) and dim(shapes.x, 4) * dim(shapes.x, 5) % 4 == 0) or (hasAxis(attrs.axes, 3, 6) and hasAxis(attrs.axes, 4, 6) and dim(shapes.x, 3) * dim(shapes.x, 4) * dim(shapes.x, 5) % 4 == 0) or (hasAxis(attrs.axes, 2, 6) and hasAxis(attrs.axes, 3, 6) and hasAxis(attrs.axes, 4, 6) and dim(shapes.x, 2) * dim(shapes.x, 3) * dim(shapes.x, 4) * dim(shapes.x, 5) % 4 == 0) or (hasAxis(attrs.axes, 1, 6) and hasAxis(attrs.axes, 2, 6) and hasAxis(attrs.axes, 3, 6) and hasAxis(attrs.axes, 4, 6) and dim(shapes.x, 1) * dim(shapes.x, 2) * dim(shapes.x, 3) * dim(shapes.x, 4) * dim(shapes.x, 5) % 4 == 0) or (allAxesReduced and numel(shapes.x) % 4 == 0))))",
    "maxWorkgroupSize": "min(tunables.WORKGROUP_SIZE, deviceWorkgroupCap)",
    "minSubgroupSize": "device.adapterInfo.subgroupMinSize if has(device.adapterInfo, \"subgroupMinSize\") else 1",
    "useSubgroups": "device.features.has(\"subgroups\") and has(device.adapterInfo, \"subgroupMinSize\") and minSubgroupSize > 0",
    "scalarWorkgroupSize": "min(maxWorkgroupSize, tunables.TREE_MEDIUM_WORKGROUP_SIZE) if not useSubgroups and reduceCount > tunables.TREE_MEDIUM_WORKGROUP_SIZE and reduceCount <= 2 * tunables.TREE_MEDIUM_WORKGROUP_SIZE else min(maxWorkgroupSize, max(1, pow2ceil(reduceCount)))",
    "vectorWorkgroupSize": "min(maxWorkgroupSize, max(1, pow2ceil(ceilDiv(reduceCount, 4))))",
    "serialWorkgroupSize": "min(device.limits.maxComputeInvocationsPerWorkgroup, device.limits.maxComputeWorkgroupSizeX, (tunables.SERIAL_TINY_WORKGROUP_SIZE if reduceCount <= 4 else tunables.SERIAL_WORKGROUP_SIZE))",
    "rowDispatchFits": "rowCount <= foldedDispatchCapacity",
    "serialDispatchFits": "ceilDiv(rowCount, serialWorkgroupSize) <= foldedDispatchCapacity",
    "applyDispatchFits": "ceilDiv(numel(shapes.y), maxWorkgroupSize) <= foldedDispatchCapacity",
    "scalarStorageFits": "scalarWorkgroupSize * 8 <= device.limits.maxComputeWorkgroupStorageSize",
    "vectorStorageFits": "vectorWorkgroupSize * 8 <= device.limits.maxComputeWorkgroupStorageSize",
    "flatSplit": "min(tunables.MAX_FLAT_SPLITS, pow2ceil(ceilDiv(numel(shapes.x), tunables.FLAT_SPLIT_TARGET_ELEMENTS)))",
    "flatScratchBytes": "flatSplit * 8",
    "flatPathFits": "flatSplit <= min(device.limits.maxComputeWorkgroupsPerDimension, 65535) and flatScratchBytes <= device.limits.maxStorageBufferBindingSize and flatScratchBytes <= device.limits.maxBufferSize and maxWorkgroupSize * 8 <= device.limits.maxComputeWorkgroupStorageSize and applyDispatchFits"
  },
  "when": ["shapeContract"],
  "bindings": {
    "x": { "buffer": "read-only-storage", "elementType": "$scalar" },
    "params": { "buffer": "uniform", "struct": [{ "name": "count", "type": "u32", "value": "numel(shapes.x)" }] },
    "x_2": { "name": "x", "buffer": "read-only-storage", "elementType": "$ioElement" },
    "y_2": { "name": "y", "buffer": "storage", "elementType": "$ioElement" },
    "params_4": {
      "name": "params",
      "buffer": "uniform",
      "struct": [{ "name": "rows", "type": "u32", "value": "rowCount" }]
    }
  },
  "variants": [
    {
      "id": "empty_noop",
      "priority": 200,
      "when": ["numel(shapes.x) == 0"],
      "passes": [
        {
          "id": "noop",
          "name": "MeanVarianceNormalization.Empty",
          "shader": "noop.wgsl.jinja",
          "bindings": [{ "name": "params", "struct": [{ "name": "dummy", "type": "u32", "value": 0 }] }],
          "dispatch": { "x": 0 }
        }
      ]
    },
    {
      "id": "all_axes_flat_split",
      "priority": 120,
      "when": ["numel(shapes.x) > 0", "allAxesReduced", "numel(shapes.x) >= tunables.FLAT_SPLIT_MIN_ELEMENTS", "flatPathFits"],
      "derive": { "scalar": "dtypes.T" },
      "intermediates": [
        { "id": "partials", "dtype": "float32", "shape": "[flatSplit, 2]" },
        { "id": "stats", "dtype": "float32", "shape": "[2]" }
      ],
      "passes": [
        {
          "id": "partials",
          "name": "MeanVarianceNormalization.FlatPartials",
          "shader": "norm-flat-splitk-partials.wgsl.jinja",
          "derive": {
            "workgroupSizeSpec": "maxWorkgroupSize",
            "splitSpec": "flatSplit",
            "usesF16Spec": "dtypes.T == \"f16\""
          },
          "bindings": ["x", { "name": "partials", "buffer": "storage", "elementType": "vec2<f32>" }, "params"],
          "dispatch": { "x": "min(flatSplit, 65535)", "y": "ceilDiv(flatSplit, 65535)", "z": 1 }
        },
        {
          "id": "combine",
          "name": "MeanVarianceNormalization.FlatCombine",
          "shader": "norm-flat-splitk-combine.wgsl.jinja",
          "derive": { "splitSpec": "flatSplit", "usesF16Spec": "dtypes.T == \"f16\"" },
          "bindings": [
            "x",
            { "name": "partials", "buffer": "read-only-storage", "elementType": "vec2<f32>" },
            { "name": "stats", "buffer": "storage", "elementType": "f32", "length": 2 },
            "params"
          ],
          "dispatch": { "x": "min(1, 65535)", "y": "ceilDiv(1, 65535)", "z": 1 }
        },
        {
          "id": "apply",
          "name": "MeanVarianceNormalization.FlatApply",
          "shader": "norm-flat-apply.wgsl.jinja",
          "derive": {
            "workgroupSizeSpec": "maxWorkgroupSize",
            "scalar": "dtypes.T",
            "usesF16Spec": "dtypes.T == \"f16\""
          },
          "bindings": [
            "x",
            { "name": "stats", "buffer": "read-only-storage", "elementType": "f32", "length": 2 },
            { "arg": "y", "elementType": "$scalar" },
            { "name": "params", "struct": [{ "name": "count", "type": "u32", "value": "numel(shapes.y)" }] }
          ],
          "dispatch": {
            "x": "min(ceilDiv((numel(shapes.y)), (maxWorkgroupSize)), 65535)",
            "y": "ceilDiv(ceilDiv((numel(shapes.y)), (maxWorkgroupSize)), 65535)",
            "z": 1
          }
        }
      ]
    },
    {
      "id": "serial_rows",
      "priority": 115,
      "when": ["numel(shapes.x) > 0", "reduceCount <= tunables.SERIAL_MAX_REDUCTION", "rowCount >= tunables.SERIAL_MIN_ROWS", "serialDispatchFits"],
      "derive": { "scalar": "dtypes.T", "ioElement": "dtypes.T" },
      "passes": [
        {
          "id": "main",
          "name": "MeanVarianceNormalization.SerialRows",
          "shader": "mean-variance-normalization-serial-rows.wgsl.jinja",
          "derive": {
            "xShape": "shapes.x",
            "reduce": ["hasAxis(attrs.axes, 0, ranks.x)", "hasAxis(attrs.axes, 1, ranks.x)", "hasAxis(attrs.axes, 2, ranks.x)", "hasAxis(attrs.axes, 3, ranks.x)", "hasAxis(attrs.axes, 4, ranks.x)", "hasAxis(attrs.axes, 5, ranks.x)", "hasAxis(attrs.axes, 6, ranks.x)", "hasAxis(attrs.axes, 7, ranks.x)"],
            "workgroupSizeSpec": "serialWorkgroupSize",
            "scalar": "dtypes.T",
            "usesF16Spec": "dtypes.T == \"f16\""
          },
          "bindings": ["x_2", "y_2", "params_4"],
          "dispatch": {
            "x": "min(ceilDiv((rowCount), (serialWorkgroupSize)), 65535)",
            "y": "ceilDiv(ceilDiv((rowCount), (serialWorkgroupSize)), 65535)",
            "z": 1
          }
        }
      ]
    },
    {
      "id": "cooperative_vec4",
      "priority": 110,
      "when": ["numel(shapes.x) > 0", "reduceCount >= tunables.VEC4_MIN_REDUCTION", "vec4Eligible", "rowDispatchFits", "vectorStorageFits"],
      "derive": { "scalar": "dtypes.T", "ioElement": "\"vec4<\" ~ dtypes.T ~ \">\"" },
      "passes": [
        {
          "id": "main",
          "name": "MeanVarianceNormalization.CooperativeVec4",
          "shader": "mean-variance-normalization-subgroup.wgsl.jinja",
          "subgroupCollectivesWidth": "portable",
          "derive": {
            "xShape": "shapes.x",
            "reduce": ["hasAxis(attrs.axes, 0, ranks.x)", "hasAxis(attrs.axes, 1, ranks.x)", "hasAxis(attrs.axes, 2, ranks.x)", "hasAxis(attrs.axes, 3, ranks.x)", "hasAxis(attrs.axes, 4, ranks.x)", "hasAxis(attrs.axes, 5, ranks.x)", "hasAxis(attrs.axes, 6, ranks.x)", "hasAxis(attrs.axes, 7, ranks.x)"],
            "wg": "vectorWorkgroupSize",
            "scalar": "dtypes.T",
            "vecType": "\"vec4<\" ~ dtypes.T ~ \">\"",
            "usesF16Spec": "dtypes.T == \"f16\"",
            "vectorizedSpec": true
          },
          "bindings": ["x_2", "y_2", "params_4"],
          "dispatch": { "x": "min(rowCount, 65535)", "y": "ceilDiv(rowCount, 65535)", "z": 1 }
        }
      ]
    },
    {
      "id": "cooperative_scalar",
      "priority": 100,
      "when": ["numel(shapes.x) > 0", "rowDispatchFits", "scalarStorageFits"],
      "derive": { "scalar": "dtypes.T", "ioElement": "dtypes.T" },
      "passes": [
        {
          "id": "main",
          "name": "MeanVarianceNormalization.CooperativeScalar",
          "shader": "mean-variance-normalization-subgroup.wgsl.jinja",
          "subgroupCollectivesWidth": "portable",
          "derive": {
            "xShape": "shapes.x",
            "reduce": ["hasAxis(attrs.axes, 0, ranks.x)", "hasAxis(attrs.axes, 1, ranks.x)", "hasAxis(attrs.axes, 2, ranks.x)", "hasAxis(attrs.axes, 3, ranks.x)", "hasAxis(attrs.axes, 4, ranks.x)", "hasAxis(attrs.axes, 5, ranks.x)", "hasAxis(attrs.axes, 6, ranks.x)", "hasAxis(attrs.axes, 7, ranks.x)"],
            "wg": "scalarWorkgroupSize",
            "scalar": "dtypes.T",
            "usesF16Spec": "dtypes.T == \"f16\"",
            "vectorizedSpec": false
          },
          "bindings": ["x_2", "y_2", "params_4"],
          "dispatch": { "x": "min(rowCount, 65535)", "y": "ceilDiv(rowCount, 65535)", "z": 1 }
        }
      ]
    }
  ]
}