sglang_v0.5.2/pytorch_2.8.0/third_party/fbgemm/bench/EmbeddingQuantizeFloatToFlo...

99 lines
2.7 KiB
C++

/*
* Copyright (c) Meta Platforms, Inc. and affiliates.
* All rights reserved.
*
* This source code is licensed under the BSD-style license found in the
* LICENSE file in the root directory of this source tree.
*/
#include <chrono>
#include <initializer_list>
#include <iomanip>
#include <iostream>
#include <vector>
#ifdef _OPENMP
#include <omp.h>
#endif
#include "./BenchUtils.h"
#include "fbgemm/QuantUtils.h"
#include "fbgemm/Types.h"
using namespace std;
using namespace fbgemm;
// T is the type of scale and bias
template <typename T>
void performance_test() {
constexpr int NWARMUP = 4;
constexpr int NITER = 256;
if (is_same<T, float16>::value) {
cout << "With result as float16" << endl;
} else {
cout << "With result as float" << endl;
}
cout << setw(6) << "rows" << "," << setw(6) << "cols" << "," << setw(16)
<< "elems_per_usec" << "," << setw(10) << "GB/Sec" << endl;
for (int rowSize : {100, 120, 1000}) {
for (int colSize : {16, 64, 128, 256, 512, 1024, 2048}) {
aligned_vector<uint8_t> inpVec(rowSize * colSize);
randFill<uint8_t>(inpVec, 0, 20);
int out_emb_cols = colSize;
if (is_same<T, float16>::value) {
out_emb_cols /= 2;
}
int outVecSize = rowSize * (out_emb_cols + 2 * sizeof(T));
aligned_vector<T> outVec(outVecSize);
double duration = 0.0f;
int constexpr kNumRepeats = is_same<T, float16>::value ? 16 : 32;
duration = measureWithWarmup(
[&]() {
for (int i = 0; i < kNumRepeats; ++i) {
Fused8BitRowwiseQuantizedSBFloatToFloatOrHalf(
inpVec.data(), rowSize, colSize, outVec.data());
}
},
NWARMUP,
NITER,
[&]() {
cache_evict(inpVec);
cache_evict(outVec);
});
float elements_per_usec =
rowSize * colSize * kNumRepeats / (duration * 1e6);
duration *= 1e9; // convert to ns
long bytes_read = rowSize * colSize * sizeof(float) * kNumRepeats;
float gigabyes_per_sec = bytes_read / duration;
cout << setw(6) << rowSize << ", " << setw(6) << colSize << ",";
cout << setw(16) << std::fixed << std::setprecision(2)
<< elements_per_usec << ", ";
cout << setw(10) << std::fixed << std::setprecision(2) << gigabyes_per_sec
<< endl;
} // for each cols
} // for each rows
} // performance_test
int main() {
#ifdef _OPENMP
// Use 1 thread unless OMP_NUM_THREADS is explicit set.
const char* val = getenv("OMP_NUM_THREADS");
if (val == nullptr || !*val) {
omp_set_num_threads(1);
}
#endif
performance_test<float16>();
performance_test<float>();
return 0;
}