| // 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> |
| |
| |
| // define XNN_ENABLE_FULL_BENCHMARKS=1 to enable all microkernel benchmarks. |
| |
| static void GEMMEnd2EndBenchmark( |
| benchmark::State& state, |
| models::ExecutionPlanFactory model_factory, |
| xnn_qs8_gemm_minmax_ukernel_function gemm, |
| xnn_qs8_igemm_minmax_ukernel_function igemm, |
| xnn_qs8_gemm_minmax_ukernel_function gemm1, |
| xnn_qs8_igemm_minmax_ukernel_function igemm1, |
| xnn_init_qs8_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.qs8.gemm because it breaks older gcc. |
| xnn_params.qs8.gemm.minmax.gemm = xnn_init_hmp_gemm_ukernel(xnn_gemm_ukernel_function(gemm)); |
| xnn_params.qs8.gemm.minmax.igemm = xnn_init_hmp_igemm_ukernel(xnn_igemm_ukernel_function(igemm)); |
| xnn_params.qs8.gemm.minmax.gemm1 = xnn_init_hmp_gemm_ukernel(xnn_gemm_ukernel_function(gemm1)); |
| xnn_params.qs8.gemm.minmax.igemm1 = xnn_init_hmp_igemm_ukernel(xnn_igemm_ukernel_function(igemm1)); |
| xnn_params.qs8.gemm.init.qs8 = init_params; |
| xnn_params.qs8.gemm.mr = mr; |
| xnn_params.qs8.gemm.nr = nr; |
| xnn_params.qs8.gemm.log2_kr = log2_kr; |
| xnn_params.qs8.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 |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x16c4_gemmlowp__aarch64_neondot_ld32(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld32, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld32, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_1x16c4_gemmlowp__aarch64_neondot_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_cortex_a53, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_cortex_a53, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_cortex_a55(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_cortex_a55, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_cortex_a55, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld32(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_ld32, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld32, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_ld64, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16c4_fp32__aarch64_neondot_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_fp32_neonv8_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_ld128, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c4__aarch64_neondot_ld128, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16c4_fp32__aarch64_neondot_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x16c4__aarch64_neondot_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c4__aarch64_neondot_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_fp32_neonv8_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| static void qs8_gemm_4x16_gemmlowp__aarch64_neon_mlal_lane_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16__aarch64_neon_mlal_lane_cortex_a53, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_4x16_gemmlowp__aarch64_neon_mlal_lane_prfm_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16__aarch64_neon_mlal_lane_prfm_cortex_a53, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm_cortex_a53(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__aarch64_neon_mlal_padal_prfm_cortex_a53, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| static void qs8_gemm_2x8c16_gemmlowp__aarch64_neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c16__aarch64_neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c16__aarch64_neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_cortex_a53) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm_cortex_a53) |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c4_gemmlowp__aarch64_neondot_ld32) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c4_gemmlowp__aarch64_neondot_ld64) |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_cortex_a55) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld32) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld64) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_gemmlowp__aarch64_neondot_ld128) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_fp32__aarch64_neondot_ld64) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_fp32__aarch64_neondot_ld128) |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm_cortex_a53) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__aarch64_neon_mlal_padal_cortex_a53) |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16_gemmlowp__aarch64_neon_mlal_lane_prfm_cortex_a53) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16_gemmlowp__aarch64_neon_mlal_lane_cortex_a53) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mlal_padal_prfm) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__aarch64_neon_mull_padal) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c16_gemmlowp__aarch64_neon_mlal_padal) |
| #endif // XNN_ARCH_ARM64 && XNN_ENABLE_ASSEMBLY |
| |
| #if XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_6x8_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_6x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_6x8__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 6 /* mr */, 8 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_6x16_gemmlowp__neon_mlal_lane(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_6x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_6x16__neon_mlal_lane, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16__neon_mlal_lane, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 6 /* mr */, 16 /* nr */, 0 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16c2_gemmlowp__neon_mlal_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c2__neon_mlal_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mlal_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16c2_gemmlowp__neon_mull_padal_dup(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c2__neon_mull_padal_dup, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c2__neon_mull_padal_dup, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_1x16c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_4x8c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_4x16c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_6x8c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_6x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_6x8c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 6 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_6x16c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_6x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_6x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 6 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_8x8c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_8x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_8x8c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 8 /* mr */, 8 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| static void qs8_gemm_8x16c4_gemmlowp__neondot(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_8x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_8x16c4__neondot, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c4__neondot, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 8 /* mr */, 16 /* nr */, 2 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEONDOT); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16c8_gemmlowp__neon_mull_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c8__neon_mull_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mull_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16c16_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c16__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c16__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 4 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_1x16c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 1 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x8c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_2x16c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x16c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 2 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x8c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x8c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_3x16c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x16c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 3 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x8c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x8c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x8c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| static void qs8_gemm_4x16c8_gemmlowp__neon_mlal_padal(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x16c8__neon_mlal_padal, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x16c8__neon_mlal_padal, |
| xnn_init_qs8_conv_minmax_gemmlowp_neon_params, |
| 4 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckNEON); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c4_gemmlowp__neondot); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_6x8c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_6x16c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_8x8c4_gemmlowp__neondot); |
| BENCHMARK_QS8_END2END(qs8_gemm_8x16c4_gemmlowp__neondot); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c8_gemmlowp__neon_mlal_padal); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c8_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c8_gemmlowp__neon_mlal_padal); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c8_gemmlowp__neon_mull_padal); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c8_gemmlowp__neon_mull_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c8_gemmlowp__neon_mull_padal); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c16_gemmlowp__neon_mlal_padal); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c16_gemmlowp__neon_mlal_padal); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c16_gemmlowp__neon_mlal_padal); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c2_gemmlowp__neon_mlal_padal_dup); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c2_gemmlowp__neon_mlal_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c2_gemmlowp__neon_mlal_padal_dup); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c2_gemmlowp__neon_mull_padal_dup); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8c2_gemmlowp__neon_mull_padal_dup); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c2_gemmlowp__neon_mull_padal_dup); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16_gemmlowp__neon_mlal_lane); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x8_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_6x8_gemmlowp__neon_mlal_lane); |
| BENCHMARK_QS8_END2END(qs8_gemm_6x16_gemmlowp__neon_mlal_lane); |
| #endif // XNN_ARCH_ARM || XNN_ARCH_ARM64 |
| |
| #if XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_init_qs8_conv_minmax_fp32_avx512_params, |
| 1 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX512F); |
| } |
| static void qs8_gemm_2x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x16c8__avx512skx, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_init_qs8_conv_minmax_fp32_avx512_params, |
| 2 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX512F); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_3x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x16c8__avx512skx, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_init_qs8_conv_minmax_fp32_avx512_params, |
| 3 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX512F); |
| } |
| static void qs8_gemm_4x16c8__avx512skx(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x16c8__avx512skx, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x16c8__avx512skx, |
| xnn_init_qs8_conv_minmax_fp32_avx512_params, |
| 4 /* mr */, 16 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX512F); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x8c8__avx2(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_init_qs8_conv_minmax_fp32_avx2_params, |
| 1 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX2); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x8c8__avx2(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x8c8__avx2, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_init_qs8_conv_minmax_fp32_avx2_params, |
| 2 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX2); |
| } |
| static void qs8_gemm_3x8c8__avx2(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x8c8__avx2, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x8c8__avx2, |
| xnn_init_qs8_conv_minmax_fp32_avx2_params, |
| 3 /* mr */, 8 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX2); |
| } |
| |
| static void qs8_gemm_2x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__xop_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_2x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__xop_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_3x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__xop_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_3x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__xop_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_4x4c2__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__xop_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_4x4c2__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__xop_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__xop_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| |
| static void qs8_gemm_2x4c8__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__xop_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_3x4c8__xop_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__xop_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| |
| static void qs8_gemm_2x4c8__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__xop_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| static void qs8_gemm_3x4c8__xop_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__xop_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__xop_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__xop_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckXOP); |
| } |
| |
| static void qs8_gemm_2x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__avx_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_2x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__avx_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_3x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__avx_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_3x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__avx_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_4x4c2__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__avx_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_4x4c2__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__avx_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__avx_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| |
| |
| static void qs8_gemm_2x4c8__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__avx_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_2x4c8__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__avx_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_3x4c8__avx_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__avx_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| static void qs8_gemm_3x4c8__avx_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__avx_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__avx_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__avx_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckAVX); |
| } |
| |
| static void qs8_gemm_2x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__sse41_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_2x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__sse41_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_3x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__sse41_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_3x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__sse41_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_4x4c2__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__sse41_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_4x4c2__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__sse41_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse41_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| |
| static void qs8_gemm_2x4c8__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__sse41_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_2x4c8__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__sse41_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_3x4c8__sse41_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__sse41_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| static void qs8_gemm_3x4c8__sse41_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__sse41_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse41_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse41_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse4_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSE41); |
| } |
| |
| |
| static void qs8_gemm_2x4c8__ssse3_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__ssse3_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__ssse3_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__ssse3_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__ssse3_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSSE3); |
| } |
| static void qs8_gemm_2x4c8__ssse3_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__ssse3_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__ssse3_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__ssse3_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__ssse3_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSSE3); |
| } |
| static void qs8_gemm_3x4c8__ssse3_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__ssse3_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__ssse3_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__ssse3_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__ssse3_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSSE3); |
| } |
| static void qs8_gemm_3x4c8__ssse3_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__ssse3_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__ssse3_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__ssse3_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__ssse3_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */, |
| benchmark::utils::CheckSSSE3); |
| } |
| |
| static void qs8_gemm_2x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__sse2_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_2x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c2__sse2_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_3x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__sse2_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_3x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c2__sse2_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_4x4c2__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__sse2_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_4x4c2__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_4x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_4x4c2__sse2_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c2__sse2_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 4 /* mr */, 4 /* nr */, 1 /* log2_kr */, 0 /* log2_sr */); |
| } |
| |
| static void qs8_gemm_2x4c8__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__sse2_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_2x4c8__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__sse2_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_3x4c8__sse2_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__sse2_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld64, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */); |
| } |
| static void qs8_gemm_3x4c8__sse2_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__sse2_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__sse2_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__sse2_ld128, |
| xnn_init_qs8_conv_minmax_fp32_sse2_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */, 0 /* log2_sr */); |
| } |
| |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x16c8__avx512skx); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x16c8__avx512skx); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_3x16c8__avx512skx); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x16c8__avx512skx); |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x8c8__avx2); |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x8c8__avx2); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x8c8__avx2); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__xop_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__xop_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__xop_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__xop_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__xop_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__xop_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__xop_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__xop_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__xop_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__xop_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__avx_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__avx_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__avx_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__avx_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__avx_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__avx_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__avx_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__avx_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__avx_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__avx_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__sse41_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__sse41_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__sse41_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__sse41_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__sse41_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__sse41_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__sse41_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__sse41_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__sse41_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__sse41_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__ssse3_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__ssse3_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__ssse3_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__ssse3_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__sse2_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c2__sse2_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__sse2_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c2__sse2_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__sse2_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4c2__sse2_ld128); |
| |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__sse2_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__sse2_ld128); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__sse2_ld64); |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__sse2_ld128); |
| #endif // XNN_ARCH_X86 || XNN_ARCH_X86_64 |
| |
| #if XNN_ARCH_WASMSIMD |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x4c8__wasmsimd_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 1 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| |
| static void qs8_gemm_1x4c8__wasmsimd_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 1 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x4c8__wasmsimd_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__wasmsimd_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| |
| static void qs8_gemm_2x4c8__wasmsimd_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_2x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_2x4c8__wasmsimd_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 2 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| |
| static void qs8_gemm_3x4c8__wasmsimd_ld64(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__wasmsimd_ld64, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld64, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| |
| static void qs8_gemm_3x4c8__wasmsimd_ld128(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_fp32_ukernel_3x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_3x4c8__wasmsimd_ld128, |
| xnn_qs8_gemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_qs8_igemm_minmax_fp32_ukernel_1x4c8__wasmsimd_ld128, |
| xnn_init_qs8_conv_minmax_fp32_wasmsimd_params, |
| 3 /* mr */, 4 /* nr */, 3 /* log2_kr */); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x4c8__wasmsimd_ld64) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x4c8__wasmsimd_ld128) |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__wasmsimd_ld64) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4c8__wasmsimd_ld128) |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__wasmsimd_ld64) |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4c8__wasmsimd_ld128) |
| #endif // XNN_ARCH_WASMSIMD |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| static void qs8_gemm_1x2_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 1 /* mr */, 2 /* nr */); |
| } |
| |
| static void qs8_gemm_1x4_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 1 /* mr */, 4 /* nr */); |
| } |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| |
| static void qs8_gemm_2x2_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x2__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 2 /* mr */, 2 /* nr */); |
| } |
| |
| static void qs8_gemm_3x2_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x2__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 3 /* mr */, 2 /* nr */); |
| } |
| |
| static void qs8_gemm_4x2_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x2__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x2__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 4 /* mr */, 2 /* nr */); |
| } |
| |
| static void qs8_gemm_2x4_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_2x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_2x4__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 2 /* mr */, 4 /* nr */); |
| } |
| |
| static void qs8_gemm_3x4_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_3x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_3x4__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 3 /* mr */, 4 /* nr */); |
| } |
| |
| static void qs8_gemm_4x4_gemmlowp__scalar(benchmark::State& state, models::ExecutionPlanFactory model) { |
| GEMMEnd2EndBenchmark(state, model, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_4x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_4x4__scalar, |
| xnn_qs8_gemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_qs8_igemm_minmax_gemmlowp_ukernel_1x4__scalar, |
| xnn_init_qs8_conv_minmax_gemmlowp_scalar_params, |
| 4 /* mr */, 4 /* nr */); |
| } |
| |
| #if XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_1x2_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_1x4_gemmlowp__scalar) |
| #endif // XNN_ENABLE_FULL_BENCHMARKS |
| BENCHMARK_QS8_END2END(qs8_gemm_2x2_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_3x2_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x2_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_2x4_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_3x4_gemmlowp__scalar) |
| BENCHMARK_QS8_END2END(qs8_gemm_4x4_gemmlowp__scalar) |
| |
| #ifndef XNNPACK_BENCHMARK_NO_MAIN |
| BENCHMARK_MAIN(); |
| #endif |