Lines Matching full:benchmark
13 #include <benchmark/benchmark.h>
23 static void SpMMBenchmark(benchmark::State& state, in SpMMBenchmark()
25 benchmark::utils::IsaCheckFunction isa_check = nullptr) in SpMMBenchmark()
55 benchmark::utils::DivideRoundUp<size_t>(benchmark::utils::GetMaxCacheSize(), in SpMMBenchmark()
138 benchmark::utils::PrefetchToL1(a.data(), a.size() * sizeof(float)); in SpMMBenchmark()
151 const uint64_t cpu_frequency = benchmark::utils::GetCurrentCpuFrequency(); in SpMMBenchmark()
156 state.counters["FLOPS"] = benchmark::Counter( in SpMMBenchmark()
157 uint64_t(state.iterations()) * 2 * mc * num_nonzeroes, benchmark::Counter::kIsRate); in SpMMBenchmark()
159 state.counters["EffFLOPS"] = benchmark::Counter( in SpMMBenchmark()
160 uint64_t(state.iterations()) * 2 * mc * nc * kc, benchmark::Counter::kIsRate); in SpMMBenchmark()
165 static void spmm80_4x1__neon(benchmark::State& state, const char* net) { in spmm80_4x1__neon()
166 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_4x1__neon, 4, 1, 0.8f, benchmark::utils::CheckNEO… in spmm80_4x1__neon()
169 static void spmm80_8x1__neon(benchmark::State& state, const char* net) { in spmm80_8x1__neon()
170 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_8x1__neon, 8, 1, 0.8f, benchmark::utils::CheckNEO… in spmm80_8x1__neon()
173 static void spmm80_12x1__neon(benchmark::State& state, const char* net) { in spmm80_12x1__neon()
174 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_12x1__neon, 12, 1, 0.8f, benchmark::utils::CheckN… in spmm80_12x1__neon()
177 static void spmm80_16x1__neon(benchmark::State& state, const char* net) { in spmm80_16x1__neon()
178 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_16x1__neon, 16, 1, 0.8f, benchmark::utils::CheckN… in spmm80_16x1__neon()
181 static void spmm80_32x1__neon(benchmark::State& state, const char* net) { in spmm80_32x1__neon()
182 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_32x1__neon, 32, 1, 0.8f, benchmark::utils::CheckN… in spmm80_32x1__neon()
185 static void spmm80_4x1__neon_x2(benchmark::State& state, const char* net) { in spmm80_4x1__neon_x2()
186 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_4x1__neon_x2, 4, 1, 0.8f, benchmark::utils::Check… in spmm80_4x1__neon_x2()
189 static void spmm80_8x1__neon_x2(benchmark::State& state, const char* net) { in spmm80_8x1__neon_x2()
190 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_8x1__neon_x2, 8, 1, 0.8f, benchmark::utils::Check… in spmm80_8x1__neon_x2()
193 static void spmm80_16x1__neon_x2(benchmark::State& state, const char* net) { in spmm80_16x1__neon_x2()
194 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_16x1__neon_x2, 16, 1, 0.8f, benchmark::utils::Che… in spmm80_16x1__neon_x2()
197 static void spmm80_32x1__neon_x2(benchmark::State& state, const char* net) { in spmm80_32x1__neon_x2()
198 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_32x1__neon_x2, 32, 1, 0.8f, benchmark::utils::Che… in spmm80_32x1__neon_x2()
201 static void spmm80_4x1__neon_pipelined(benchmark::State& state, const char* net) { in spmm80_4x1__neon_pipelined()
202 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_4x1__neon_pipelined, 4, 1, 0.8f, benchmark::utils… in spmm80_4x1__neon_pipelined()
205 static void spmm80_8x1__neon_pipelined(benchmark::State& state, const char* net) { in spmm80_8x1__neon_pipelined()
206 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_8x1__neon_pipelined, 8, 1, 0.8f, benchmark::utils… in spmm80_8x1__neon_pipelined()
209 static void spmm80_16x1__neon_pipelined(benchmark::State& state, const char* net) { in spmm80_16x1__neon_pipelined()
210 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_16x1__neon_pipelined, 16, 1, 0.8f, benchmark::uti… in spmm80_16x1__neon_pipelined()
213 static void spmm80_32x1__neon_pipelined(benchmark::State& state, const char* net) { in spmm80_32x1__neon_pipelined()
214 …SpMMBenchmark(state, xnn_f32_spmm_minmax_ukernel_32x1__neon_pipelined, 32, 1, 0.8f, benchmark::uti… in spmm80_32x1__neon_pipelined()
233 static void spmm80_4x1__neonfma(benchmark::State& state, const char* net) { in BENCHMARK_SPMM()
237 static void spmm80_4x2__neonfma(benchmark::State& state, const char* net) { in spmm80_4x2__neonfma()
241 static void spmm80_4x4__neonfma(benchmark::State& state, const char* net) { in spmm80_4x4__neonfma()
245 static void spmm80_8x1__neonfma(benchmark::State& state, const char* net) { in spmm80_8x1__neonfma()
249 static void spmm80_8x2__neonfma(benchmark::State& state, const char* net) { in spmm80_8x2__neonfma()
253 static void spmm80_8x4__neonfma(benchmark::State& state, const char* net) { in spmm80_8x4__neonfma()
257 static void spmm80_12x1__neonfma(benchmark::State& state, const char* net) { in spmm80_12x1__neonfma()
261 static void spmm80_12x2__neonfma(benchmark::State& state, const char* net) { in spmm80_12x2__neonfma()
265 static void spmm80_12x4__neonfma(benchmark::State& state, const char* net) { in spmm80_12x4__neonfma()
269 static void spmm80_16x1__neonfma(benchmark::State& state, const char* net) { in spmm80_16x1__neonfma()
273 static void spmm80_16x2__neonfma(benchmark::State& state, const char* net) { in spmm80_16x2__neonfma()
277 static void spmm80_16x4__neonfma(benchmark::State& state, const char* net) { in spmm80_16x4__neonfma()
281 static void spmm80_32x1__neonfma(benchmark::State& state, const char* net) { in spmm80_32x1__neonfma()
285 static void spmm80_32x2__neonfma(benchmark::State& state, const char* net) { in spmm80_32x2__neonfma()
289 static void spmm80_32x4__neonfma(benchmark::State& state, const char* net) { in spmm80_32x4__neonfma()
293 static void spmm80_4x1__neonfma_x2(benchmark::State& state, const char* net) { in spmm80_4x1__neonfma_x2()
297 static void spmm80_8x1__neonfma_x2(benchmark::State& state, const char* net) { in spmm80_8x1__neonfma_x2()
301 static void spmm80_16x1__neonfma_x2(benchmark::State& state, const char* net) { in spmm80_16x1__neonfma_x2()
305 static void spmm80_32x1__neonfma_x2(benchmark::State& state, const char* net) { in spmm80_32x1__neonfma_x2()
309 static void spmm80_4x1__neonfma_pipelined(benchmark::State& state, const char* net) { in spmm80_4x1__neonfma_pipelined()
313 static void spmm80_8x1__neonfma_pipelined(benchmark::State& state, const char* net) { in spmm80_8x1__neonfma_pipelined()
317 static void spmm80_16x1__neonfma_pipelined(benchmark::State& state, const char* net) { in spmm80_16x1__neonfma_pipelined()
321 static void spmm80_32x1__neonfma_pipelined(benchmark::State& state, const char* net) { in spmm80_32x1__neonfma_pipelined()
351 static void spmm80_4x1__sse(benchmark::State& state, const char* net) { in BENCHMARK_SPMM()
355 static void spmm80_8x1__sse(benchmark::State& state, const char* net) { in spmm80_8x1__sse()
359 static void spmm80_16x1__sse(benchmark::State& state, const char* net) { in spmm80_16x1__sse()
363 static void spmm80_32x1__sse(benchmark::State& state, const char* net) { in spmm80_32x1__sse()
373 static void spmm80_1x1__scalar(benchmark::State& state, const char* net) { in BENCHMARK_SPMM()
377 static void spmm80_2x1__scalar(benchmark::State& state, const char* net) { in spmm80_2x1__scalar()
381 static void spmm80_4x1__scalar(benchmark::State& state, const char* net) { in spmm80_4x1__scalar()
385 static void spmm80_8x1__scalar(benchmark::State& state, const char* net) { in spmm80_8x1__scalar()
389 static void spmm80_8x2__scalar(benchmark::State& state, const char* net) { in spmm80_8x2__scalar()
393 static void spmm80_8x4__scalar(benchmark::State& state, const char* net) { in spmm80_8x4__scalar()
397 static void spmm80_1x1__scalar_pipelined(benchmark::State& state, const char* net) { in spmm80_1x1__scalar_pipelined()
401 static void spmm80_2x1__scalar_pipelined(benchmark::State& state, const char* net) { in spmm80_2x1__scalar_pipelined()
405 static void spmm80_4x1__scalar_pipelined(benchmark::State& state, const char* net) { in spmm80_4x1__scalar_pipelined()
409 static void spmm80_8x1__scalar_pipelined(benchmark::State& state, const char* net) { in spmm80_8x1__scalar_pipelined()
425 static void spmm80_4x1__wasmsimd_arm(benchmark::State& state, const char* net) { in BENCHMARK_SPMM()
429 static void spmm80_8x1__wasmsimd_arm(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_arm()
433 static void spmm80_16x1__wasmsimd_arm(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_arm()
437 static void spmm80_32x1__wasmsimd_arm(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_arm()
441 static void spmm80_4x1__wasmsimd_x86(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_x86()
445 static void spmm80_8x1__wasmsimd_x86(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_x86()
449 static void spmm80_16x1__wasmsimd_x86(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_x86()
453 static void spmm80_32x1__wasmsimd_x86(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_x86()
457 static void spmm80_32x1__wasmsimd_arm_x2(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_arm_x2()
461 static void spmm80_4x1__wasmsimd_arm_x2(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_arm_x2()
465 static void spmm80_8x1__wasmsimd_arm_x2(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_arm_x2()
469 static void spmm80_16x1__wasmsimd_arm_x2(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_arm_x2()
473 static void spmm80_32x1__wasmsimd_x86_x2(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_x86_x2()
477 static void spmm80_4x1__wasmsimd_x86_x2(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_x86_x2()
481 static void spmm80_8x1__wasmsimd_x86_x2(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_x86_x2()
485 static void spmm80_16x1__wasmsimd_x86_x2(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_x86_x2()
489 static void spmm80_32x1__wasmsimd_arm_x4(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_arm_x4()
493 static void spmm80_4x1__wasmsimd_arm_x4(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_arm_x4()
497 static void spmm80_8x1__wasmsimd_arm_x4(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_arm_x4()
501 static void spmm80_16x1__wasmsimd_arm_x4(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_arm_x4()
505 static void spmm80_32x1__wasmsimd_x86_x4(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_x86_x4()
509 static void spmm80_4x1__wasmsimd_x86_x4(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_x86_x4()
513 static void spmm80_8x1__wasmsimd_x86_x4(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_x86_x4()
517 static void spmm80_16x1__wasmsimd_x86_x4(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_x86_x4()
520 static void spmm80_4x1__wasmsimd_arm_pipelined(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_arm_pipelined()
524 static void spmm80_8x1__wasmsimd_arm_pipelined(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_arm_pipelined()
528 static void spmm80_16x1__wasmsimd_arm_pipelined(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_arm_pipelined()
532 static void spmm80_32x1__wasmsimd_arm_pipelined(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_arm_pipelined()
536 static void spmm80_4x1__wasmsimd_x86_pipelined(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_x86_pipelined()
540 static void spmm80_8x1__wasmsimd_x86_pipelined(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_x86_pipelined()
544 static void spmm80_16x1__wasmsimd_x86_pipelined(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_x86_pipelined()
548 static void spmm80_32x1__wasmsimd_x86_pipelined(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_x86_pipelined()
552 static void spmm80_32x1__wasmsimd_arm_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_arm_pipelined_x2()
556 static void spmm80_4x1__wasmsimd_arm_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_arm_pipelined_x2()
560 static void spmm80_8x1__wasmsimd_arm_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_arm_pipelined_x2()
564 static void spmm80_16x1__wasmsimd_arm_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_arm_pipelined_x2()
568 static void spmm80_32x1__wasmsimd_x86_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_32x1__wasmsimd_x86_pipelined_x2()
572 static void spmm80_4x1__wasmsimd_x86_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_4x1__wasmsimd_x86_pipelined_x2()
576 static void spmm80_8x1__wasmsimd_x86_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_8x1__wasmsimd_x86_pipelined_x2()
580 static void spmm80_16x1__wasmsimd_x86_pipelined_x2(benchmark::State& state, const char* net) { in spmm80_16x1__wasmsimd_x86_pipelined_x2()