hailin
/
sglang_v0.5.2
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity
sglang_v0.5.2/pytorch_2.8.0/third_party/XNNPACK/bench/gemm.h

793 lines
30 KiB
C++
Raw Permalink Blame History

// Copyright (c) Facebook, Inc. and its affiliates.
// All rights reserved.
//
// Copyright 2019 Google LLC
//
// This source code is licensed under the BSD-style license found in the
// LICENSE file in the root directory of this source tree.
#pragma once
#include <benchmark/benchmark.h>
#include "xnnpack/math.h"
#include <cstdint>
#include <initializer_list>
#include <functional>
#include <string>
#include <vector>
/*
* This is a lightweight wrapper around the underlying benchmark::internal::Benchmark
* class, which handles registration of benchmark case args. It provides default values
* for blocksize and skips blockwise-specific cases for non-blockwise kernels.
*/
class BenchmarkWrapper {
public:
BenchmarkWrapper(benchmark::internal::Benchmark* benchmark, bool blockwise)
: benchmark_(benchmark), blockwise_(blockwise), last_k_(0), last_m_(0), last_n_(0) {}
void ArgNames(const std::initializer_list<std::string>& names) {
std::vector<std::string> vals(names.begin(), names.end());
// Args are standardized as M, N, K, [BL] for gemm kernels.
if (blockwise_ && names.size() < 4) {
vals.push_back("BL");
}
else if (!blockwise_ && vals.size() == 4) {
// Drop blocksize argument for non-blockwise kernels. This declutters the
// benchmark output slightly.
vals.pop_back();
}
benchmark_->ArgNames(vals);
}
void Args(const std::initializer_list<int64_t>& args) {
std::vector<int64_t> vals(args.begin(), args.end());
// Args are standardized as M, N, K, [BL] for gemm kernels.
auto m = vals[0];
auto n = vals[1];
auto k = vals[2];
// For non-blockwise kernels, skip test cases that differ only by blocksize.
if (!blockwise_ && m == last_m_ && n == last_n_ && k == last_k_) {
return;
}
if (blockwise_ && args.size() < 4) {
// Use K as default blocksize for parity with non-blockwise kernels.
// This is equivalent to per-channel quantization.
vals.push_back(round_up_po2(k, 32));
}
else if (!blockwise_ && vals.size() == 4) {
// Drop blocksize argument for non-blockwise kernels. This declutters the
// benchmark output slightly.
vals.pop_back();
}
last_m_ = m;
last_n_ = n;
last_k_ = k;
benchmark_->Args(vals);
}
private:
benchmark::internal::Benchmark* benchmark_;
bool blockwise_;
int64_t last_k_;
int64_t last_m_;
int64_t last_n_;
};
void Configure(benchmark::internal::Benchmark* benchmark, void (setup_method)(BenchmarkWrapper*), bool blockwise) {
BenchmarkWrapper wrapper(benchmark, blockwise);
setup_method(&wrapper);
benchmark->UseRealTime();
}
#define BENCHMARK_CASE(gemm_fn, test_case_name, name, setup_method, blockwise) \
BENCHMARK_CAPTURE(gemm_fn, test_case_name, name) \
->Apply([](benchmark::internal::Benchmark* b) { return Configure(b, &setup_method, blockwise); }) \
->UseRealTime();
#define BENCHMARK_GEMM_BASE(gemm_fn, blockwise) \
BENCHMARK_CASE(gemm_fn, mobilenet_v1, "MobileNet v1", MobileNetV1GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, mobilenet_v2, "MobileNet v2", MobileNetV2GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, mobilenet_v3_small, "MobileNet v3 Small", MobileNetV3SmallGemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, mobilenet_v3_large, "MobileNet v3 Large", MobileNetV3LargeGemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v1_g1, "ShuffleNet v1 (1 group)", ShuffleNetV1G1GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v1_g2, "ShuffleNet v1 (2 groups)", ShuffleNetV1G2GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v1_g3, "ShuffleNet v1 (3 groups)", ShuffleNetV1G3GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v1_g4, "ShuffleNet v1 (4 groups)", ShuffleNetV1G4GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v1_g8, "ShuffleNet v1 (8 groups)", ShuffleNetV1G8GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v2_x05, "ShuffleNet v2 0.5X", ShuffleNetV2X05GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v2_x10, "ShuffleNet v2 1.0X", ShuffleNetV2X10GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v2_x15, "ShuffleNet v2 1.5X", ShuffleNetV2X15GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, shufflenet_v2_x20, "ShuffleNet v2 2.0X", ShuffleNetV2X20GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, inception_v3, "Inception v3", InceptionV3GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, resnet18, "ResNet-18", ResNet18GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, resnet50, "ResNet-50", ResNet50GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, squeezenet_v10, "SqueezeNet 1.0", SqueezeNetV10GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, squeezenet_v11, "SqueezeNet 1.1", SqueezeNetV11GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, vgg, "VGG", VGGGemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, srcnn915, "SRCNN (9-1-5)", SRCNN915GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, srcnn935, "SRCNN (9-3-5)", SRCNN935GemmArguments, blockwise); \
BENCHMARK_CASE(gemm_fn, llm, "LLM", LLMGemmArguments, blockwise);
// Removed due to OOM SEGFAULT on 32 bit ARM.
// BENCHMARK_CAPTURE(gemm_fn, srcnn955, "SRCNN (9-5-5)")->Apply(SRCNN955GemmArguments)->UseRealTime();
#define BENCHMARK_GEMM(gemm_fn) BENCHMARK_GEMM_BASE(gemm_fn, false)
#define BENCHMARK_GEMM_BL(gemm_fn) BENCHMARK_GEMM_BASE(gemm_fn, true)
// ShuffleNet v1 with 1 group.
//static void ShuffleNetV1G1GemmArguments(benchmark::internal::Benchmark* b) {
static void ShuffleNetV1G1GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 36, 24 * 1 * 1});
b->Args({ 28 * 28, 120, 36 * 1 * 1});
b->Args({ 28 * 28, 36, 144 * 1 * 1});
b->Args({ 28 * 28, 144, 36 * 1 * 1});
b->Args({ 28 * 28, 72, 144 * 1 * 1});
b->Args({ 14 * 14, 144, 72 * 1 * 1});
b->Args({ 14 * 14, 72, 288 * 1 * 1});
b->Args({ 14 * 14, 288, 72 * 1 * 1});
b->Args({ 14 * 14, 144, 288 * 1 * 1});
b->Args({ 7 * 7, 288, 144 * 1 * 1});
b->Args({ 7 * 7, 144, 576 * 1 * 1});
b->Args({ 7 * 7, 576, 144 * 1 * 1});
}
// ShuffleNet v1 with 2 groups.
static void ShuffleNetV1G2GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 50, 24 * 1 * 1});
b->Args({ 28 * 28, 88, 25 * 1 * 1});
b->Args({ 28 * 28, 25, 100 * 1 * 1});
b->Args({ 28 * 28, 100, 25 * 1 * 1});
b->Args({ 28 * 28, 50, 100 * 1 * 1});
b->Args({ 14 * 14, 100, 50 * 1 * 1});
b->Args({ 14 * 14, 50, 200 * 1 * 1});
b->Args({ 14 * 14, 200, 50 * 1 * 1});
b->Args({ 14 * 14, 100, 200 * 1 * 1});
b->Args({ 7 * 7, 200, 100 * 1 * 1});
b->Args({ 7 * 7, 100, 400 * 1 * 1});
b->Args({ 7 * 7, 400, 100 * 1 * 1});
}
// ShuffleNet v1 with 3 groups.
static void ShuffleNetV1G3GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 60, 24 * 1 * 1});
b->Args({ 28 * 28, 72, 20 * 1 * 1});
b->Args({ 28 * 28, 20, 80 * 1 * 1});
b->Args({ 28 * 28, 80, 20 * 1 * 1});
b->Args({ 28 * 28, 40, 80 * 1 * 1});
b->Args({ 14 * 14, 80, 40 * 1 * 1});
b->Args({ 14 * 14, 40, 160 * 1 * 1});
b->Args({ 14 * 14, 160, 40 * 1 * 1});
b->Args({ 14 * 14, 80, 160 * 1 * 1});
b->Args({ 7 * 7, 160, 80 * 1 * 1});
b->Args({ 7 * 7, 80, 320 * 1 * 1});
b->Args({ 7 * 7, 320, 80 * 1 * 1});
}
// ShuffleNet v1 with 4 groups.
static void ShuffleNetV1G4GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 68, 24 * 1 * 1});
b->Args({ 28 * 28, 62, 17 * 1 * 1});
b->Args({ 28 * 28, 17, 68 * 1 * 1});
b->Args({ 28 * 28, 68, 17 * 1 * 1});
b->Args({ 28 * 28, 34, 68 * 1 * 1});
b->Args({ 14 * 14, 68, 34 * 1 * 1});
b->Args({ 14 * 14, 34, 136 * 1 * 1});
b->Args({ 14 * 14, 136, 34 * 1 * 1});
b->Args({ 14 * 14, 68, 136 * 1 * 1});
b->Args({ 7 * 7, 136, 68 * 1 * 1});
b->Args({ 7 * 7, 68, 272 * 1 * 1});
b->Args({ 7 * 7, 272, 68 * 1 * 1});
}
// ShuffleNet v1 with 8 groups.
static void ShuffleNetV1G8GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 96, 24 * 1 * 1});
b->Args({ 28 * 28, 45, 12 * 1 * 1});
b->Args({ 28 * 28, 12, 48 * 1 * 1});
b->Args({ 28 * 28, 48, 12 * 1 * 1});
b->Args({ 28 * 28, 24, 48 * 1 * 1});
b->Args({ 14 * 14, 48, 24 * 1 * 1});
b->Args({ 14 * 14, 24, 96 * 1 * 1});
b->Args({ 14 * 14, 96, 24 * 1 * 1});
b->Args({ 14 * 14, 48, 96 * 1 * 1});
b->Args({ 7 * 7, 96, 48 * 1 * 1});
b->Args({ 7 * 7, 48, 192 * 1 * 1});
b->Args({ 7 * 7, 192, 48 * 1 * 1});
}
// ShuffleNet v2 (0.5X scale)
static void ShuffleNetV2X05GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 24, 24 * 1 * 1});
b->Args({ 28 * 28, 24, 24 * 1 * 1});
b->Args({ 28 * 28, 48, 48 * 1 * 1});
b->Args({ 14 * 14, 48, 48 * 1 * 1});
b->Args({ 14 * 14, 96, 96 * 1 * 1});
b->Args({ 7 * 7, 96, 96 * 1 * 1});
b->Args({ 7 * 7, 1024, 192 * 1 * 1});
}
// ShuffleNet v2 (1.0X scale)
static void ShuffleNetV2X10GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 58, 24 * 1 * 1});
b->Args({ 28 * 28, 58, 24 * 1 * 1});
b->Args({ 28 * 28, 58, 58 * 1 * 1});
b->Args({ 14 * 14, 116, 116 * 1 * 1});
b->Args({ 14 * 14, 116, 116 * 1 * 1});
b->Args({ 14 * 14, 232, 232 * 1 * 1});
b->Args({ 7 * 7, 232, 232 * 1 * 1});
b->Args({ 7 * 7, 1024, 464 * 1 * 1});
}
// ShuffleNet v2 (1.5X scale)
static void ShuffleNetV2X15GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 88, 24 * 1 * 1});
b->Args({ 28 * 28, 88, 24 * 1 * 1});
b->Args({ 28 * 28, 88, 88 * 1 * 1});
b->Args({ 28 * 28, 176, 176 * 1 * 1});
b->Args({ 14 * 14, 176, 176 * 1 * 1});
b->Args({ 14 * 14, 352, 352 * 1 * 1});
b->Args({ 7 * 7, 352, 352 * 1 * 1});
b->Args({ 7 * 7, 1024, 704 * 1 * 1});
}
// ShuffleNet v2 (2.0X scale)
static void ShuffleNetV2X20GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 24, 3 * 3 * 3});
b->Args({ 56 * 56, 122, 24 * 1 * 1});
b->Args({ 28 * 28, 122, 24 * 1 * 1});
b->Args({ 28 * 28, 122, 122 * 1 * 1});
b->Args({ 28 * 28, 244, 244 * 1 * 1});
b->Args({ 14 * 14, 244, 244 * 1 * 1});
b->Args({ 14 * 14, 488, 488 * 1 * 1});
b->Args({ 7 * 7, 488, 488 * 1 * 1});
b->Args({ 7 * 7, 2048, 976 * 1 * 1});
}
static void MobileNetV1GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 32, 3 * 3 * 3});
b->Args({112 * 112, 64, 32 * 1 * 1});
b->Args({ 56 * 56, 128, 64 * 1 * 1});
b->Args({ 56 * 56, 128, 128 * 1 * 1});
b->Args({ 28 * 28, 256, 128 * 1 * 1});
b->Args({ 28 * 28, 256, 256 * 1 * 1});
b->Args({ 14 * 14, 512, 256 * 1 * 1});
b->Args({ 14 * 14, 512, 512 * 1 * 1});
b->Args({ 7 * 7, 1024, 512 * 1 * 1});
b->Args({ 7 * 7, 1024, 1024 * 1 * 1});
}
static void MobileNetV2GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/*********** Initial Stage ************/
/* M N K */
b->Args({112 * 112, 32, 3 * 3 * 3});
/************ Bottleneck 1 ************/
/* M N K */
b->Args({112 * 112, 16, 32 * 1 * 1});
/************ Bottleneck 2 ************/
/* M N K */
b->Args({112 * 112, 96, 16 * 1 * 1});
b->Args({ 56 * 56, 24, 96 * 1 * 1});
b->Args({ 56 * 56, 144, 24 * 1 * 1});
b->Args({ 56 * 56, 24, 144 * 1 * 1});
/************ Bottleneck 3 ************/
/* M N K */
b->Args({ 28 * 28, 32, 144 * 1 * 1});
b->Args({ 28 * 28, 192, 32 * 1 * 1});
b->Args({ 28 * 28, 32, 192 * 1 * 1});
/************ Bottleneck 4 ************/
/* M N K */
b->Args({ 14 * 14, 64, 192 * 1 * 1});
b->Args({ 14 * 14, 384, 64 * 1 * 1});
b->Args({ 14 * 14, 64, 384 * 1 * 1});
/************ Bottleneck 5 ************/
/* M N K */
b->Args({ 14 * 14, 96, 384 * 1 * 1});
b->Args({ 14 * 14, 576, 96 * 1 * 1});
b->Args({ 14 * 14, 96, 576 * 1 * 1});
/************ Bottleneck 6 ************/
/* M N K */
b->Args({ 7 * 7, 160, 576 * 1 * 1});
b->Args({ 7 * 7, 960, 160 * 1 * 1});
b->Args({ 7 * 7, 160, 960 * 1 * 1});
/************ Bottleneck 7 ************/
/* M N K */
b->Args({ 7 * 7, 320, 960 * 1 * 1});
/********* Pre-pooling Conv2D *********/
/* M N K */
b->Args({ 7 * 7, 1280, 320 * 1 * 1});
/******** Post-pooling Conv2D *********/
/* M N K */
b->Args({ 1 * 1, 1000, 1280 * 1 * 1});
}
static void MobileNetV3SmallGemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/************ Initial Stage ************/
/* M N K */
b->Args({112 * 112, 16, 3 * 3 * 3});
/************* Bottleneck 1 ************/
/* M N K */
b->Args({ 1 * 1, 8, 16 * 1 * 1});
b->Args({ 1 * 1, 16, 8 * 1 * 1});
b->Args({ 56 * 56, 16, 16 * 1 * 1});
/************* Bottleneck 2 ************/
/* M N K */
b->Args({ 56 * 56, 72, 16 * 1 * 1});
b->Args({ 28 * 28, 24, 72 * 1 * 1});
/************* Bottleneck 3 ************/
/* M N K */
b->Args({ 28 * 28, 88, 24 * 1 * 1});
b->Args({ 28 * 28, 24, 88 * 1 * 1});
/************* Bottleneck 4 ************/
/* M N K */
b->Args({ 28 * 28, 96, 24 * 1 * 1});
b->Args({ 1 * 1, 24, 96 * 1 * 1});
b->Args({ 1 * 1, 96, 24 * 1 * 1});
b->Args({ 14 * 14, 40, 96 * 1 * 1});
/************* Bottleneck 5 ************/
/* M N K */
b->Args({ 14 * 14, 240, 40 * 1 * 1});
b->Args({ 1 * 1, 64, 240 * 1 * 1});
b->Args({ 1 * 1, 240, 64 * 1 * 1});
b->Args({ 14 * 14, 40, 240 * 1 * 1});
/************* Bottleneck 6 ************/
/* M N K */
//b->Args({ 14 * 14, 240, 40 * 1 * 1});
//b->Args({ 1 * 1, 64, 240 * 1 * 1});
//b->Args({ 1 * 1, 240, 64 * 1 * 1});
//b->Args({ 14 * 14, 40, 240 * 1 * 1});
/************* Bottleneck 7 ************/
/* M N K */
b->Args({ 14 * 14, 120, 40 * 1 * 1});
b->Args({ 1 * 1, 32, 120 * 1 * 1});
b->Args({ 1 * 1, 120, 32 * 1 * 1});
b->Args({ 14 * 14, 48, 120 * 1 * 1});
/************* Bottleneck 8 ************/
/* M N K */
b->Args({ 14 * 14, 144, 48 * 1 * 1});
b->Args({ 1 * 1, 40, 144 * 1 * 1});
b->Args({ 1 * 1, 144, 40 * 1 * 1});
b->Args({ 14 * 14, 48, 144 * 1 * 1});
/************* Bottleneck 9 ************/
/* M N K */
b->Args({ 14 * 14, 288, 48 * 1 * 1});
b->Args({ 1 * 1, 72, 288 * 1 * 1});
b->Args({ 1 * 1, 288, 72 * 1 * 1});
b->Args({ 7 * 7, 96, 288 * 1 * 1});
/************ Bottleneck 10 ************/
/* M N K */
b->Args({ 7 * 7, 576, 96 * 1 * 1});
b->Args({ 1 * 1, 144, 576 * 1 * 1});
b->Args({ 1 * 1, 576, 144 * 1 * 1});
b->Args({ 7 * 7, 96, 576 * 1 * 1});
/************ Bottleneck 11 ************/
/* M N K */
//b->Args({ 7 * 7, 576, 96 * 1 * 1});
//b->Args({ 1 * 1, 144, 576 * 1 * 1});
//b->Args({ 1 * 1, 576, 144 * 1 * 1});
//b->Args({ 7 * 7, 96, 576 * 1 * 1});
/************* Last Stage *************/
/* M N K */
//b->Args({ 7 * 7, 576, 96 * 1 * 1});
b->Args({ 1 * 1, 1024, 576 * 1 * 1});
b->Args({ 1 * 1, 1001, 1024 * 1 * 1});
}
static void MobileNetV3LargeGemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/************ Initial Stage ************/
/* M N K */
b->Args({112 * 112, 16, 3 * 3 * 3});
/************* Bottleneck 1 ************/
/* M N K */
b->Args({112 * 112, 16, 16 * 1 * 1});
/************* Bottleneck 2 ************/
/* M N K */
b->Args({112 * 112, 64, 16 * 1 * 1});
b->Args({ 56 * 56, 24, 64 * 1 * 1});
/************* Bottleneck 3 ************/
/* M N K */
b->Args({ 56 * 56, 72, 24 * 1 * 1});
b->Args({ 56 * 56, 24, 72 * 1 * 1});
/************* Bottleneck 4 ************/
/* M N K */
//b->Args({ 56 * 56, 72, 24 * 1 * 1});
b->Args({ 1 * 1, 24, 72 * 1 * 1});
b->Args({ 1 * 1, 72, 24 * 1 * 1});
b->Args({ 28 * 28, 40, 72 * 1 * 1});
/************* Bottleneck 5 ************/
/* M N K */
b->Args({ 28 * 28, 120, 40 * 1 * 1});
b->Args({ 1 * 1, 32, 120 * 1 * 1});
b->Args({ 1 * 1, 120, 32 * 1 * 1});
b->Args({ 28 * 28, 40, 120 * 1 * 1});
/************* Bottleneck 6 ************/
/* M N K */
//b->Args({ 28 * 28, 120, 40 * 1 * 1});
//b->Args({ 1 * 1, 32, 120 * 1 * 1});
//b->Args({ 1 * 1, 120, 32 * 1 * 1});
//b->Args({ 28 * 28, 40, 120 * 1 * 1});
/************* Bottleneck 7 ************/
/* M N K */
b->Args({ 28 * 28, 240, 40 * 1 * 1});
b->Args({ 14 * 14, 80, 240 * 1 * 1});
/************* Bottleneck 8 ************/
/* M N K */
b->Args({ 14 * 14, 200, 80 * 1 * 1});
b->Args({ 14 * 14, 80, 200 * 1 * 1});
/************* Bottleneck 9 ************/
/* M N K */
b->Args({ 14 * 14, 184, 80 * 1 * 1});
b->Args({ 14 * 14, 80, 184 * 1 * 1});
/************ Bottleneck 10 ************/
/* M N K */
b->Args({ 14 * 14, 184, 80 * 1 * 1});
b->Args({ 14 * 14, 80, 184 * 1 * 1});
/************ Bottleneck 11 ************/
/* M N K */
b->Args({ 14 * 14, 480, 80 * 1 * 1});
b->Args({ 1 * 1, 120, 480 * 1 * 1});
b->Args({ 1 * 1, 480, 120 * 1 * 1});
b->Args({ 14 * 14, 112, 480 * 1 * 1});
/************ Bottleneck 12 ************/
/* M N K */
b->Args({ 14 * 14, 672, 112 * 1 * 1});
b->Args({ 1 * 1, 168, 672 * 1 * 1});
b->Args({ 1 * 1, 672, 168 * 1 * 1});
b->Args({ 14 * 14, 112, 672 * 1 * 1});
/************ Bottleneck 13 ************/
/* M N K */
//b->Args({ 14 * 14, 672, 112 * 1 * 1});
//b->Args({ 1 * 1, 168, 672 * 1 * 1});
//b->Args({ 1 * 1, 672, 168 * 1 * 1});
b->Args({ 7 * 7, 160, 672 * 1 * 1});
/************ Bottleneck 14 ************/
/* M N K */
b->Args({ 7 * 7, 960, 160 * 1 * 1});
b->Args({ 1 * 1, 240, 960 * 1 * 1});
b->Args({ 1 * 1, 960, 240 * 1 * 1});
b->Args({ 7 * 7, 160, 960 * 1 * 1});
/************ Bottleneck 15 ************/
/* M N K */
//b->Args({ 7 * 7, 960, 160 * 1 * 1});
//b->Args({ 1 * 1, 240, 960 * 1 * 1});
//b->Args({ 1 * 1, 960, 240 * 1 * 1});
//b->Args({ 7 * 7, 160, 960 * 1 * 1});
/************* Last Stage *************/
/* M N K */
//b->Args({ 7 * 7, 960, 160 * 1 * 1});
b->Args({ 1 * 1, 1280, 960 * 1 * 1});
b->Args({ 1 * 1, 1001, 1280 * 1 * 1});
}
// SqueezeNet 1.0
static void SqueezeNetV10GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/************** Conv 1 ***************/
/* M N K */
b->Args({111 * 111, 96, 3 * 7 * 7});
/************** Fire 2 ***************/
/* M N K */
b->Args({ 55 * 55, 16, 96 * 1 * 1});
b->Args({ 55 * 55, 64, 16 * 1 * 1});
b->Args({ 55 * 55, 64, 16 * 3 * 3});
/************** Fire 3 ***************/
/* M N K */
b->Args({ 55 * 55, 16, 128 * 1 * 1});
/************** Fire 4 ***************/
/* M N K */
b->Args({ 55 * 55, 32, 128 * 1 * 1});
b->Args({ 55 * 55, 128, 32 * 1 * 1});
b->Args({ 55 * 55, 128, 32 * 3 * 3});
/************** Fire 5 ***************/
/* M N K */
b->Args({ 27 * 27, 32, 256 * 1 * 1});
b->Args({ 27 * 27, 128, 32 * 1 * 1});
b->Args({ 27 * 27, 128, 32 * 3 * 3});
/************** Fire 6 ***************/
/* M N K */
b->Args({ 27 * 27, 48, 256 * 1 * 1});
b->Args({ 27 * 27, 192, 48 * 1 * 1});
b->Args({ 27 * 27, 192, 48 * 3 * 3});
/************** Fire 7 ***************/
/* M N K */
b->Args({ 27 * 27, 48, 384 * 1 * 1});
/************** Fire 8 ***************/
/* M N K */
b->Args({ 27 * 27, 64, 384 * 1 * 1});
b->Args({ 27 * 27, 256, 64 * 1 * 1});
b->Args({ 27 * 27, 256, 64 * 3 * 3});
/************** Fire 9 ***************/
/* M N K */
b->Args({ 13 * 13, 64, 512 * 1 * 1});
b->Args({ 13 * 13, 256, 64 * 1 * 1});
b->Args({ 13 * 13, 256, 64 * 3 * 3});
/************** Conv 10 **************/
/* M N K */
b->Args({ 13 * 13, 1000, 512 * 1 * 1});
}
// SqueezeNet 1.1
static void SqueezeNetV11GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/************** Conv 1 ***************/
/* M N K */
b->Args({111 * 111, 64, 3 * 3 * 3});
/************** Fire 2 ***************/
/* M N K */
b->Args({ 55 * 55, 16, 64 * 1 * 1});
b->Args({ 55 * 55, 64, 16 * 1 * 1});
b->Args({ 55 * 55, 64, 16 * 3 * 3});
/************** Fire 3 ***************/
/* M N K */
b->Args({ 55 * 55, 16, 128 * 1 * 1});
/************** Fire 4 ***************/
/* M N K */
b->Args({ 27 * 27, 32, 128 * 1 * 1});
b->Args({ 27 * 27, 128, 32 * 1 * 1});
b->Args({ 27 * 27, 128, 32 * 3 * 3});
/************** Fire 5 ***************/
/* M N K */
b->Args({ 27 * 27, 32, 256 * 1 * 1});
/************** Fire 6 ***************/
/* M N K */
b->Args({ 13 * 13, 48, 256 * 1 * 1});
b->Args({ 13 * 13, 192, 48 * 1 * 1});
b->Args({ 13 * 13, 192, 48 * 3 * 3});
/************** Fire 7 ***************/
/* M N K */
b->Args({ 13 * 13, 48, 384 * 1 * 1});
/************** Fire 8 ***************/
/* M N K */
b->Args({ 13 * 13, 64, 384 * 1 * 1});
b->Args({ 13 * 13, 256, 64 * 1 * 1});
b->Args({ 13 * 13, 256, 64 * 3 * 3});
/************** Fire 9 ***************/
/* M N K */
b->Args({ 13 * 13, 64, 512 * 1 * 1});
/************** Conv 10 **************/
/* M N K */
b->Args({ 13 * 13, 1000, 512 * 1 * 1});
}
static void InceptionV3GemmArguments(BenchmarkWrapper* b) {
/* M N K */
b->Args({150 * 150, 32, 3 * 3 * 3});
b->Args({149 * 149, 32, 32 * 3 * 3});
b->Args({149 * 149, 64, 32 * 3 * 3});
b->Args({ 75 * 75, 80, 64 * 1 * 1});
b->Args({ 73 * 73, 192, 80 * 3 * 3});
b->Args({ 37 * 37, 64, 192 * 1 * 1});
b->Args({ 37 * 37, 48, 192 * 1 * 1});
b->Args({ 37 * 37, 64, 48 * 5 * 5});
b->Args({ 37 * 37, 96, 64 * 3 * 3});
b->Args({ 37 * 37, 96, 96 * 3 * 3});
b->Args({ 37 * 37, 32, 192 * 1 * 1});
b->Args({ 37 * 37, 64, 256 * 1 * 1});
b->Args({ 37 * 37, 48, 256 * 1 * 1});
b->Args({ 37 * 37, 64, 288 * 1 * 1});
b->Args({ 37 * 37, 48, 288 * 1 * 1});
b->Args({ 18 * 18, 384, 288 * 3 * 3});
b->Args({ 18 * 18, 96, 96 * 3 * 3});
b->Args({ 19 * 19, 192, 768 * 1 * 1});
b->Args({ 19 * 19, 128, 768 * 1 * 1});
b->Args({ 19 * 19, 128, 128 * 1 * 7});
b->Args({ 19 * 19, 192, 128 * 7 * 1});
b->Args({ 19 * 19, 128, 128 * 7 * 1});
b->Args({ 19 * 19, 192, 128 * 1 * 7});
b->Args({ 19 * 19, 160, 768 * 1 * 1});
b->Args({ 19 * 19, 160, 160 * 1 * 7});
b->Args({ 19 * 19, 192, 160 * 7 * 1});
b->Args({ 19 * 19, 160, 160 * 7 * 1});
b->Args({ 19 * 19, 192, 160 * 1 * 7});
b->Args({ 19 * 19, 192, 192 * 1 * 7});
b->Args({ 19 * 19, 192, 192 * 7 * 1});
b->Args({ 9 * 9, 320, 192 * 3 * 3});
b->Args({ 9 * 9, 192, 192 * 3 * 3});
b->Args({ 10 * 10, 320, 1280 * 1 * 1});
b->Args({ 10 * 10, 384, 1280 * 1 * 1});
b->Args({ 10 * 10, 384, 384 * 1 * 3});
b->Args({ 10 * 10, 384, 384 * 3 * 1});
b->Args({ 10 * 10, 448, 1280 * 1 * 1});
b->Args({ 10 * 10, 384, 448 * 3 * 3});
b->Args({ 10 * 10, 192, 1280 * 1 * 1});
b->Args({ 10 * 10, 320, 2048 * 1 * 1});
b->Args({ 10 * 10, 384, 2048 * 1 * 1});
b->Args({ 10 * 10, 448, 2048 * 1 * 1});
b->Args({ 10 * 10, 192, 2048 * 1 * 1});
b->Args({ 3 * 3, 1001, 2048 * 1 * 1});
}
static void ResNet18GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({112 * 112, 64, 3 * 7 * 7});
b->Args({ 56 * 56, 64, 64 * 3 * 3});
b->Args({ 28 * 28, 128, 64 * 3 * 3});
b->Args({ 28 * 28, 128, 128 * 3 * 3});
b->Args({ 28 * 28, 128, 64 * 1 * 1});
b->Args({ 14 * 14, 256, 128 * 3 * 3});
b->Args({ 14 * 14, 256, 256 * 3 * 3});
b->Args({ 14 * 14, 256, 128 * 1 * 1});
b->Args({ 7 * 7, 512, 256 * 3 * 3});
b->Args({ 7 * 7, 512, 512 * 3 * 3});
b->Args({ 7 * 7, 512, 256 * 1 * 1});
}
static void ResNet50GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/*************** Conv 1 ***************/
/* M N K */
b->Args({112 * 112, 64, 3 * 7 * 7});
/************** Conv 2.X **************/
/* M N K */
b->Args({ 56 * 56, 64, 64 * 1 * 1});
b->Args({ 56 * 56, 64, 64 * 3 * 3});
b->Args({ 56 * 56, 256, 64 * 1 * 1});
b->Args({ 56 * 56, 64, 256 * 1 * 1});
/************** Conv 3.X **************/
/* M N K */
b->Args({ 56 * 56, 128, 256 * 1 * 1});
b->Args({ 28 * 28, 128, 128 * 3 * 3});
b->Args({ 28 * 28, 512, 128 * 1 * 1});
b->Args({ 28 * 28, 512, 256 * 1 * 1});
b->Args({ 28 * 28, 128, 512 * 1 * 1});
/************** Conv 4.X **************/
/* M N K */
b->Args({ 28 * 28, 256, 512 * 1 * 1});
b->Args({ 14 * 14, 256, 256 * 3 * 3});
b->Args({ 14 * 14, 1024, 256 * 1 * 1});
b->Args({ 14 * 14, 1024, 512 * 1 * 1});
b->Args({ 14 * 14, 256, 1024 * 1 * 1});
/************** Conv 5.X **************/
/* M N K */
b->Args({ 14 * 14, 512, 1024 * 1 * 1});
b->Args({ 7 * 7, 512, 512 * 3 * 3});
b->Args({ 7 * 7, 2048, 512 * 1 * 1});
b->Args({ 7 * 7, 2048, 1024 * 1 * 1});
b->Args({ 7 * 7, 512, 2048 * 1 * 1});
}
static void VGGGemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/************** Conv 1.1 *************/
/* M N K */
b->Args({224 * 224, 64, 3 * 3 * 3});
/************** Conv 1.2 *************/
/* M N K */
b->Args({224 * 224, 64, 64 * 3 * 3});
/************** Conv 2.1 *************/
/* M N K */
b->Args({112 * 112, 128, 64 * 3 * 3});
/************** Conv 2.2 *************/
/* M N K */
b->Args({112 * 112, 128, 128 * 3 * 3});
/************** Conv 3.1 *************/
/* M N K */
b->Args({ 56 * 56, 256, 128 * 3 * 3});
/************** Conv 3.3 *************/
/* M N K */
b->Args({ 56 * 56, 256, 256 * 1 * 1});
/************** Conv 4.1 *************/
/* M N K */
b->Args({ 28 * 28, 512, 256 * 3 * 3});
/************** Conv 4.2 *************/
/* M N K */
b->Args({ 28 * 28, 512, 512 * 3 * 3});
/************** Conv 4.3 *************/
/* M N K */
b->Args({ 28 * 28, 512, 512 * 1 * 1});
/************** Conv 5.X *************/
/* M N K */
b->Args({ 14 * 14, 512, 512 * 3 * 3});
/************** Conv 5.3 *************/
/* M N K */
b->Args({ 14 * 14, 512, 512 * 1 * 1});
}
// SRCNN (9-1-5)
static void SRCNN915GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({376 * 376, 64, 1 * 9 * 9});
b->Args({376 * 376, 32, 64 * 1 * 1});
b->Args({372 * 372, 1, 32 * 5 * 5});
}
// SRCNN (9-3-5)
static void SRCNN935GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({376 * 376, 64, 1 * 9 * 9});
b->Args({374 * 374, 32, 64 * 3 * 3});
b->Args({370 * 370, 1, 32 * 5 * 5});
}
// SRCNN (9-5-5)
static void SRCNN955GemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K"});
/* M N K */
b->Args({376 * 376, 64, 1 * 9 * 9});
b->Args({372 * 372, 32, 64 * 5 * 5});
b->Args({368 * 368, 1, 32 * 5 * 5});
}
// Large Language Model (Generic)
static void LLMGemmArguments(BenchmarkWrapper* b) {
b->ArgNames({"M", "N", "K", "Bl"});
b->Args({ 128, 16, 1024, 32 });
b->Args({ 128, 16, 1024, 256 });
b->Args({ 128, 128, 1024, 32 });
b->Args({ 128, 128, 1024, 256 });
b->Args({ 128, 4096, 1024, 32 });
b->Args({ 128, 4096, 1024, 256 });
b->Args({ 128, 11008, 4096, 32 });
b->Args({ 128, 11008, 4096, 256 });
b->Args({ 128, 32000, 4096, 32 });
b->Args({ 128, 32000, 4096, 256 });
}
Reference in New Issue View Git Blame Copy Permalink