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/qd8-f16-qc4w-gemm.cc

1611 lines
64 KiB
C++
Raw Blame History

// Copyright 2023 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.
//
// Auto-generated file. Do not edit!
// Specification: test/qd8-f16-qc4w-gemm-minmax.yaml
// Generator: tools/generate-gemm-test.py
#include <benchmark/benchmark.h>
#include "gemm-benchmark.h"
#include "utils.h"
#include "xnnpack/common.h"
#include "xnnpack/gemm.h"
#include "xnnpack/isa-checks.h"
#include "xnnpack/microfnptr.h"
#include "xnnpack/microparams-init.h"
#include "xnnpack/pack.h"
#include "xnnpack/packw.h"
#if XNN_ENABLE_ARM_I8MM && XNN_ARCH_ARM64
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x32c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x16c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x16c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/16, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x16c8__neoni8mm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x32c8__neoni8mm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x32c8__neoni8mm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/32, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckNEONI8MM);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x32c8__neoni8mm)
#endif // XNN_ENABLE_ARM_I8MM && XNN_ARCH_ARM64
#if XNN_ENABLE_AVX256VNNIGFNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnnigfni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNIGFNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnnigfni_prfm)
#endif // XNN_ENABLE_AVX256VNNIGFNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
#if XNN_ENABLE_AVX256VNNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256vnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256VNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256vnni_prfm)
#endif // XNN_ENABLE_AVX256VNNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
#if XNN_ENABLE_AVX256SKX && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/9, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_9x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/10, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_10x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/12, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_12x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/14, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_14x8c8__avx256skx_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx256skx)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX256SKX);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx256skx)
#endif // XNN_ENABLE_AVX256SKX && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
#if XNN_ENABLE_AVXVNNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avxvnni_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVXVNNI);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avxvnni_prfm)
#endif // XNN_ENABLE_AVXVNNI && (XNN_ARCH_X86 || XNN_ARCH_X86_64)
#if XNN_ARCH_X86 || XNN_ARCH_X86_64
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4uw_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2_madd_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/7, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_7x8c8__avx2)
static void qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/8, /*nr=*/8, /*kr=*/8, /*sr=*/1,
benchmark::utils::CheckAVX2);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_8x8c8__avx2)
#endif // XNN_ARCH_X86 || XNN_ARCH_X86_64
#if XNN_ENABLE_ARM_DOTPROD && XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x16c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x16c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x16c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x16c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_5x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_5x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/5, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_5x16c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x8c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x8c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/8, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x8c4__neondotfp16arith)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x16c4__neondotfp16arith(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x16c4__neondotfp16arith,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/16, /*kr=*/4, /*sr=*/1,
benchmark::utils::CheckNEONDOT);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x16c4__neondotfp16arith)
#endif // XNN_ENABLE_ARM_DOTPROD && XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
#if XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane)
static void qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/1, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_1x16__neonfp16arith_mlal_lane_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane)
static void qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/2, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_2x16__neonfp16arith_mlal_lane_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane)
static void qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/3, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_3x16__neonfp16arith_mlal_lane_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane)
static void qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/4, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_4x16__neonfp16arith_mlal_lane_prfm)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane)
static void qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane_prfm(benchmark::State& state, const char* net) {
GEMMBenchmark(state,
xnn_qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane_prfm,
xnn_init_f16_qc4w_minmax_scalar_params,
xnn_pack_qs8_qc4w_gemm_goi_w,
/*mr=*/6, /*nr=*/16, /*kr=*/1, /*sr=*/1,
benchmark::utils::CheckNEONFP16ARITH);
}
BENCHMARK_GEMM(qd8_f16_qc4w_gemm_minmax_ukernel_6x16__neonfp16arith_mlal_lane_prfm)
#endif // XNN_ENABLE_ARM_FP16_VECTOR && (XNN_ARCH_ARM || XNN_ARCH_ARM64)
#ifndef XNNPACK_BENCHMARK_NO_MAIN
BENCHMARK_MAIN();
#endif
Reference in New Issue View Git Blame Copy Permalink