File size: 7,018 Bytes
26d5b81
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
246
247
248
249
250
251
252
253
254
255
256
257
/**
 * NeuroFlow Core Tests - Tensor Operations
 */

#include <iostream>
#include <cassert>
#include <cmath>
#include <chrono>
#include "../include/neuroflow/tensor.hpp"

using namespace neuroflow;

void test_tensor_creation() {
    std::cout << "Testing tensor creation..." << std::endl;
    
    Tensor t1;
    assert(t1.data_size_ == 0);
    
    Tensor t2({32, 512});
    assert(t2.shape_.size() == 2);
    assert(t2.shape_[0] == 32);
    assert(t2.shape_[1] == 512);
    assert(t2.numel() == 32 * 512);
    assert(t2.data_size_ == 32 * 512 * 4);
    
    std::cout << "  PASSED: tensor creation" << std::endl;
}

void test_tensor_reshape() {
    std::cout << "Testing tensor reshape..." << std::endl;
    
    Tensor t({32, 512});
    Tensor r = t.reshape({16, 1024});
    
    assert(r.shape_[0] == 16);
    assert(r.shape_[1] == 1024);
    assert(r.numel() == t.numel());
    assert(!r.owns_data_);  // 零拷贝
    
    std::cout << "  PASSED: tensor reshape (zero-copy)" << std::endl;
}

void test_tensor_clone() {
    std::cout << "Testing tensor clone..." << std::endl;
    
    Tensor t({10, 20});
    float* data = t.as_fp32();
    for (size_t i = 0; i < t.numel(); ++i) {
        data[i] = static_cast<float>(i);
    }
    
    Tensor c = t.clone();
    assert(c.owns_data_);
    assert(c.numel() == t.numel());
    
    float* cdata = c.as_fp32();
    for (size_t i = 0; i < c.numel(); ++i) {
        assert(std::abs(cdata[i] - static_cast<float>(i)) < 1e-6);
    }
    
    std::cout << "  PASSED: tensor clone" << std::endl;
}

void test_gemm() {
    std::cout << "Testing GEMM (matrix multiplication)..." << std::endl;
    
    // 简单测试: A(2,3) @ B(3,2) = C(2,2)
    Tensor A({2, 3});
    Tensor B({3, 2});
    Tensor C({2, 2});
    
    float* a = A.as_fp32();
    float* b = B.as_fp32();
    
    // A = [[1,2,3], [4,5,6]]
    a[0] = 1; a[1] = 2; a[2] = 3;
    a[3] = 4; a[4] = 5; a[5] = 6;
    
    // B = [[7,8], [9,10], [11,12]]
    b[0] = 7;  b[1] = 8;
    b[2] = 9;  b[3] = 10;
    b[4] = 11; b[5] = 12;
    
    TensorOps::gemm(A, B, C);
    
    float* c = C.as_fp32();
    
    // C = [[58,64], [139,154]]
    assert(std::abs(c[0] - 58) < 1e-4);
    assert(std::abs(c[1] - 64) < 1e-4);
    assert(std::abs(c[2] - 139) < 1e-4);
    assert(std::abs(c[3] - 154) < 1e-4);
    
    std::cout << "  PASSED: GEMM basic" << std::endl;
}

void test_gemm_performance() {
    std::cout << "Testing GEMM performance..." << std::endl;
    
    size_t M = 256, K = 512, N = 256;
    
    Tensor A({M, K});
    Tensor B({K, N});
    Tensor C({M, N});
    
    // 填充随机数据
    float* a = A.as_fp32();
    float* b = B.as_fp32();
    for (size_t i = 0; i < A.numel(); ++i) a[i] = static_cast<float>(std::rand()) / RAND_MAX - 0.5f;
    for (size_t i = 0; i < B.numel(); ++i) b[i] = static_cast<float>(std::rand()) / RAND_MAX - 0.5f;
    
    // 预热
    TensorOps::gemm(A, B, C);
    
    // 性能测试
    int iterations = 10;
    auto start = std::chrono::high_resolution_clock::now();
    
    for (int i = 0; i < iterations; ++i) {
        TensorOps::gemm(A, B, C);
    }
    
    auto end = std::chrono::high_resolution_clock::now();
    auto duration = std::chrono::duration_cast<std::chrono::microseconds>(end - start);
    
    double ms = duration.count() / 1000.0 / iterations;
    double gflops = 2.0 * M * K * N / (ms / 1000.0) / 1e9;
    
    std::cout << "  GEMM (256x512x256): " << ms << " ms per iteration, " << gflops << " GFLOPS" << std::endl;
    std::cout << "  PASSED: GEMM performance" << std::endl;
}

void test_layer_norm() {
    std::cout << "Testing LayerNorm..." << std::endl;
    
    Tensor x({2, 4});
    Tensor weight({4});
    Tensor bias({4});
    
    float* data = x.as_fp32();
    data[0] = 1; data[1] = 2; data[2] = 3; data[3] = 4;
    data[4] = 5; data[5] = 6; data[6] = 7; data[7] = 8;
    
    float* w = weight.as_fp32();
    float* b = bias.as_fp32();
    for (size_t i = 0; i < 4; ++i) {
        w[i] = 1.0f;
        b[i] = 0.0f;
    }
    
    TensorOps::layer_norm(x, weight, bias);
    
    // 验证均值接近0,方差接近1
    float* out = x.as_fp32();
    
    // 第一行均值
    float mean1 = 0;
    for (size_t i = 0; i < 4; ++i) mean1 += out[i];
    mean1 /= 4;
    assert(std::abs(mean1) < 1e-4);
    
    std::cout << "  PASSED: LayerNorm" << std::endl;
}

void test_gelu() {
    std::cout << "Testing GELU..." << std::endl;
    
    Tensor x({5});
    float* data = x.as_fp32();
    data[0] = -1; data[1] = 0; data[2] = 1; data[3] = 2; data[4] = 3;
    
    TensorOps::gelu(x);
    
    // GELU(0) ≈ 0
    assert(std::abs(data[1]) < 1e-4);
    
    // GELU(1) ≈ 0.841
    assert(std::abs(data[2] - 0.841f) < 0.01);
    
    std::cout << "  PASSED: GELU" << std::endl;
}

void test_softmax() {
    std::cout << "Testing Softmax..." << std::endl;
    
    Tensor x({2, 4});
    float* data = x.as_fp32();
    data[0] = 1; data[1] = 2; data[2] = 3; data[3] = 4;
    data[4] = 0; data[5] = 0; data[6] = 0; data[7] = 0;
    
    TensorOps::softmax(x);
    
    // 验证每行和为1
    float sum1 = 0;
    for (size_t i = 0; i < 4; ++i) sum1 += data[i];
    assert(std::abs(sum1 - 1.0f) < 1e-4);
    
    float sum2 = 0;
    for (size_t i = 4; i < 8; ++i) sum2 += data[i];
    assert(std::abs(sum2 - 1.0f) < 1e-4);
    
    std::cout << "  PASSED: Softmax" << std::endl;
}

void test_quantization() {
    std::cout << "Testing INT8 quantization..." << std::endl;
    
    Tensor fp32({4, 8});
    float* data = fp32.as_fp32();
    for (size_t i = 0; i < fp32.numel(); ++i) {
        data[i] = (static_cast<float>(std::rand()) / RAND_MAX - 0.5f) * 10;
    }
    
    Tensor int8({4, 8}, QuantType::INT8);
    Tensor scale({4});
    
    TensorOps::quantize_int8(fp32, int8, scale);
    
    // 反量化
    Tensor dequant({4, 8});
    TensorOps::dequantize_int8(int8, dequant, scale);
    
    // 验证误差小于量化精度
    float* original = fp32.as_fp32();
    float* restored = dequant.as_fp32();
    
    float max_error = 0;
    for (size_t i = 0; i < fp32.numel(); ++i) {
        float err = std::abs(original[i] - restored[i]);
        max_error = std::max(max_error, err);
    }
    
    std::cout << "  Max quantization error: " << max_error << std::endl;
    std::cout << "  PASSED: INT8 quantization" << std::endl;
}

int main(int argc, char** argv) {
    std::cout << "========================================" << std::endl;
    std::cout << "NeuroFlow Core - Tensor Tests" << std::endl;
    std::cout << "========================================" << std::endl;
    
    test_tensor_creation();
    test_tensor_reshape();
    test_tensor_clone();
    test_gemm();
    test_gemm_performance();
    test_layer_norm();
    test_gelu();
    test_softmax();
    test_quantization();
    
    std::cout << "========================================" << std::endl;
    std::cout << "All tests PASSED!" << std::endl;
    std::cout << "========================================" << std::endl;
    
    return 0;
}