blob: da6c7b731a8ba62650e65e761c5478dbfdf87bf8 [file] [log] [blame]
// Copyright 2021 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.
#include <algorithm>
#include <cmath>
#include <functional>
#include <random>
#include <vector>
#include <xnnpack.h>
#include <benchmark/benchmark.h>
#include "bench/end2end.h"
#include "bench/utils.h"
#include "models/models.h"
#include <xnnpack/gemm.h>
#include <xnnpack/igemm.h>
#include <xnnpack/params.h>
#include <xnnpack/params-init.h>
static void GEMMEnd2EndBenchmark(
benchmark::State& state,
models::ExecutionPlanFactory model_factory,
xnn_qu8_gemm_minmax_ukernel_function gemm,
xnn_qu8_igemm_minmax_ukernel_function igemm,
xnn_qu8_gemm_minmax_ukernel_function gemm1,
xnn_qu8_igemm_minmax_ukernel_function igemm1,
xnn_init_qu8_conv_minmax_params_fn init_params,
uint8_t mr, uint8_t nr, uint8_t log2_kr = 0, uint8_t log2_sr = 0,
benchmark::utils::IsaCheckFunction isa_check = nullptr)
{
if (isa_check && !isa_check(state)) {
return;
}
if (xnn_initialize(nullptr /* allocator */) != xnn_status_success) {
state.SkipWithError("failed to initialize XNNPACK");
return;
}
// Override microkernels chosen in xnn_initialize
// Note: do not directly assign to xnn_params.qu8.gemm because it breaks older gcc.
xnn_params.qu8.gemm.minmax.gemm = xnn_init_hmp_gemm_ukernel(xnn_gemm_ukernel_function(gemm));
xnn_params.qu8.gemm.minmax.igemm = xnn_init_hmp_igemm_ukernel(xnn_igemm_ukernel_function(igemm));
xnn_params.qu8.gemm.minmax.gemm1 = xnn_init_hmp_gemm_ukernel(xnn_gemm_ukernel_function(gemm1));
xnn_params.qu8.gemm.minmax.igemm1 = xnn_init_hmp_igemm_ukernel(xnn_igemm_ukernel_function(igemm1));
xnn_params.qu8.gemm.init.qu8 = init_params;
xnn_params.qu8.gemm.mr = mr;
xnn_params.qu8.gemm.nr = nr;
xnn_params.qu8.gemm.log2_kr = log2_kr;
xnn_params.qu8.gemm.log2_sr = log2_sr;
auto execution_plan = model_factory(nullptr);
if (execution_plan.empty()) {
state.SkipWithError("failed to create a model");
return;
}
for (auto _ : state) {
for (const std::unique_ptr<xnn_operator, decltype(&xnn_delete_operator)>& op : execution_plan) {
xnn_status status = xnn_run_operator(op.get(), nullptr);
if (status != xnn_status_success) {
state.SkipWithError("failed to run a model");
return;
}
}
}
const uint64_t cpu_frequency = benchmark::utils::GetCurrentCpuFrequency();
if (cpu_frequency != 0) {
state.counters["cpufreq"] = cpu_frequency;
}
}
#if XNN_ARCH_ARM64 && XNN_ENABLE_ASSEMBLY
static void qu8_gemm_4x16c4__aarch64_neondot_cortex_a55(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_cortex_a55,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_cortex_a55,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x16c4__aarch64_neondot_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_ld128,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16c4__aarch64_neondot_ld128,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x8c4__aarch64_neondot_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x8c4__aarch64_neondot_ld128,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x8c4__aarch64_neondot_ld128,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x8c4__aarch64_neondot_cortex_a55(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x8c4__aarch64_neondot_cortex_a55,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x8c4__aarch64_neondot_cortex_a55,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_cortex_a75(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a75,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a75,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_cortex_a75(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a75,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a75,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a53,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a53,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_ld64,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_ld64,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_ld64,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__aarch64_neon_mlal_lane_prfm_ld64,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
BENCHMARK_QU8_END2END(qu8_gemm_4x8c4__aarch64_neondot_cortex_a55);
BENCHMARK_QU8_END2END(qu8_gemm_4x16c4__aarch64_neondot_cortex_a55);
BENCHMARK_QU8_END2END(qu8_gemm_4x8c4__aarch64_neondot_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x16c4__aarch64_neondot_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_cortex_a75);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_cortex_a75);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_cortex_a53);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__aarch64_neon_mlal_lane_prfm_ld64);
#endif // XNN_ARCH_ARM64 && XNN_ENABLE_ASSEMBLY
#if XNN_ARCH_ARM || XNN_ARCH_ARM64
static void qu8_gemm_4x8__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x8__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x8__neon_mlal_lane,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_4x16__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16__neon_mlal_lane,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16__neon_mlal_lane,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEON);
}
static void qu8_gemm_1x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
1 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_2x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_2x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_2x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
2 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_3x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_3x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_3x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
3 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_5x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_5x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_5x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
5 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_6x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_6x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_6x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
6 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_8x8c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_8x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_8x8c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x8c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
8 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_1x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
1 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_2x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_2x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_2x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
2 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_3x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_3x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_3x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
3 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_4x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_4x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_4x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_5x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_5x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_5x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
5 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_6x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_6x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_6x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
6 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_8x16c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_8x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_8x16c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x16c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
8 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_2x32c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_2x32c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_2x32c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x32c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x32c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
2 /* mr */, 32 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
static void qu8_gemm_3x32c4__neondot(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_rndnu_ukernel_3x32c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_3x32c4__neondot,
xnn_qu8_gemm_minmax_rndnu_ukernel_1x32c4__neondot,
xnn_qu8_igemm_minmax_rndnu_ukernel_1x32c4__neondot,
xnn_init_qu8_conv_minmax_rndnu_neon_params,
3 /* mr */, 32 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_QU8_END2END(qu8_gemm_1x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_2x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_3x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_4x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_5x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_6x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_8x8c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_1x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_2x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_3x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_4x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_5x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_6x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_8x16c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_2x32c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_3x32c4__neondot);
BENCHMARK_QU8_END2END(qu8_gemm_4x8__neon_mlal_lane);
BENCHMARK_QU8_END2END(qu8_gemm_4x16__neon_mlal_lane);
#endif // XNN_ARCH_ARM || XNN_ARCH_ARM64
#if XNN_ARCH_X86 || XNN_ARCH_X86_64
static void qu8_gemm_2x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_2x16c8__avx512skx,
xnn_qu8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_init_qu8_conv_minmax_fp32_avx512_params,
2 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX512F);
}
static void qu8_gemm_3x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_3x16c8__avx512skx,
xnn_qu8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_init_qu8_conv_minmax_fp32_avx512_params,
3 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX512F);
}
static void qu8_gemm_4x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_4x16c8__avx512skx,
xnn_qu8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_qu8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx,
xnn_init_qu8_conv_minmax_fp32_avx512_params,
4 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX512F);
}
static void qu8_gemm_2x8c8__avx2(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x8c8__avx2,
xnn_qu8_igemm_minmax_fp32_ukernel_2x8c8__avx2,
xnn_qu8_gemm_minmax_fp32_ukernel_1x8c8__avx2,
xnn_qu8_igemm_minmax_fp32_ukernel_1x8c8__avx2,
xnn_init_qu8_conv_minmax_fp32_avx2_params,
2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX2);
}
static void qu8_gemm_3x8c8__avx2(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x8c8__avx2,
xnn_qu8_igemm_minmax_fp32_ukernel_3x8c8__avx2,
xnn_qu8_gemm_minmax_fp32_ukernel_1x8c8__avx2,
xnn_qu8_igemm_minmax_fp32_ukernel_1x8c8__avx2,
xnn_init_qu8_conv_minmax_fp32_avx2_params,
3 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX2);
}
static void qu8_gemm_2x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__xop_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_2x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__xop_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_3x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__xop_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_3x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__xop_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_4x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__xop_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_4x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__xop_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_2x4c8__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__xop_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_3x4c8__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__xop_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_2x4c8__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__xop_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_3x4c8__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__xop_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckXOP);
}
static void qu8_gemm_2x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__avx_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_2x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__avx_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_3x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__avx_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_3x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__avx_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_4x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__avx_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_4x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__avx_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_2x4c8__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__avx_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_2x4c8__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__avx_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_3x4c8__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__avx_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_3x4c8__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__avx_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckAVX);
}
static void qu8_gemm_2x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__sse41_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_2x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__sse41_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_3x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__sse41_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_3x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__sse41_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_4x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__sse41_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_4x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__sse41_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_2x4c8__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__sse41_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_2x4c8__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__sse41_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_3x4c8__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__sse41_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_3x4c8__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__sse41_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */,
benchmark::utils::CheckSSE41);
}
static void qu8_gemm_2x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__sse2_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_2x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c2__sse2_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_3x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__sse2_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_3x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c2__sse2_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_4x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__sse2_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_4x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4c2__sse2_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_2x4c8__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__sse2_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_2x4c8__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__sse2_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_3x4c8__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__sse2_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld64,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */);
}
static void qu8_gemm_3x4c8__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__sse2_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld128,
xnn_init_qu8_conv_minmax_fp32_sse2_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */);
}
BENCHMARK_QU8_END2END(qu8_gemm_2x16c8__avx512skx);
BENCHMARK_QU8_END2END(qu8_gemm_3x16c8__avx512skx);
BENCHMARK_QU8_END2END(qu8_gemm_4x16c8__avx512skx);
BENCHMARK_QU8_END2END(qu8_gemm_2x8c8__avx2);
BENCHMARK_QU8_END2END(qu8_gemm_3x8c8__avx2);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__xop_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__xop_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__xop_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__xop_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__xop_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__xop_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__xop_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__xop_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__xop_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__xop_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__avx_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__avx_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__avx_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__avx_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__avx_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__avx_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__avx_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__avx_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__avx_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__avx_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__sse41_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__sse41_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__sse41_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__sse41_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__sse41_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__sse41_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__sse41_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__sse41_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__sse41_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__sse41_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__sse2_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c2__sse2_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__sse2_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c2__sse2_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__sse2_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_4x4c2__sse2_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__sse2_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__sse2_ld128);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__sse2_ld64);
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__sse2_ld128);
#endif // XNN_ARCH_X86 || XNN_ARCH_X86_64
#if XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
static void qu8_gemm_2x4c8__wasmsimd_mul32_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__wasmsimd_mul32_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__wasmsimd_mul32_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld64,
xnn_init_qu8_conv_minmax_fp32_wasmsimd_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */);
}
static void qu8_gemm_2x4c8__wasmsimd_mul32_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4c8__wasmsimd_mul32_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4c8__wasmsimd_mul32_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld128,
xnn_init_qu8_conv_minmax_fp32_wasmsimd_params,
2 /* mr */, 4 /* nr */, 3 /* log2_kr */);
}
static void qu8_gemm_3x4c8__wasmsimd_mul32_ld64(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__wasmsimd_mul32_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__wasmsimd_mul32_ld64,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld64,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld64,
xnn_init_qu8_conv_minmax_fp32_wasmsimd_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */);
}
static void qu8_gemm_3x4c8__wasmsimd_mul32_ld128(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4c8__wasmsimd_mul32_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4c8__wasmsimd_mul32_ld128,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld128,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_mul32_ld128,
xnn_init_qu8_conv_minmax_fp32_wasmsimd_params,
3 /* mr */, 4 /* nr */, 3 /* log2_kr */);
}
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__wasmsimd_mul32_ld64)
BENCHMARK_QU8_END2END(qu8_gemm_2x4c8__wasmsimd_mul32_ld128)
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__wasmsimd_mul32_ld64)
BENCHMARK_QU8_END2END(qu8_gemm_3x4c8__wasmsimd_mul32_ld128)
#endif // XNN_ARCH_WASMSIMD || XNN_ARCH_WASMRELAXEDSIMD
static void qu8_gemm_2x2_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_2x2__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
2 /* mr */, 2 /* nr */);
}
static void qu8_gemm_3x2_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_3x2__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
3 /* mr */, 2 /* nr */);
}
static void qu8_gemm_4x2_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_4x2__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x2__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
4 /* mr */, 2 /* nr */);
}
static void qu8_gemm_2x4_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_2x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_2x4__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
2 /* mr */, 4 /* nr */);
}
static void qu8_gemm_3x4_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_3x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_3x4__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
3 /* mr */, 4 /* nr */);
}
static void qu8_gemm_4x4_scalar_fmagic(benchmark::State& state, models::ExecutionPlanFactory model) {
GEMMEnd2EndBenchmark(state, model,
xnn_qu8_gemm_minmax_fp32_ukernel_4x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_4x4__scalar_fmagic,
xnn_qu8_gemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_qu8_igemm_minmax_fp32_ukernel_1x4__scalar_fmagic,
xnn_init_qu8_conv_minmax_fp32_scalar_fmagic_params,
4 /* mr */, 4 /* nr */);
}
BENCHMARK_QU8_END2END(qu8_gemm_2x2_scalar_fmagic)
BENCHMARK_QU8_END2END(qu8_gemm_3x2_scalar_fmagic)
BENCHMARK_QU8_END2END(qu8_gemm_4x2_scalar_fmagic)
BENCHMARK_QU8_END2END(qu8_gemm_2x4_scalar_fmagic)
BENCHMARK_QU8_END2END(qu8_gemm_3x4_scalar_fmagic)
BENCHMARK_QU8_END2END(qu8_gemm_4x4_scalar_fmagic)
#ifndef XNNPACK_BENCHMARK_NO_MAIN
BENCHMARK_MAIN();
#endif