Yi-Hong Lyu 6b6a62fb40 Add vectorized AVX512F kernel for ReduceMaximumF32Kernel (#20268) ... ### Description <!-- Describe your changes. --> This commit introduces a new vectorized AVX512F kernel, MlasReduceMaximumF32KernelAvx512F, which efficiently computes the maximum value of the supplied buffer. Additionally, microbenchmarks have been added for MlasComputeSoftmax (inplace), MlasReduceMaximumF32KernelAvx, MlasComputeSumExpF32KernelAvx512F, and MlasComputeSoftmaxOutputF32KernelAvx. ### Motivation and Context <!-- - Why is this change required? What problem does it solve? - If it fixes an open issue, please link to the issue here. --> The goal of this commit is to enhance the performance of ReduceMaximumF32Kernel on CPUs with AVX512F instruction support.   | AVX |   |   | AVX512 |   |   |   -- | -- | -- | -- | -- | -- | -- | -- name | iterations | real_time | cpu_time | iterations | real_time | cpu_time | time_unit REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:3/real_time | 271277304 | 2.58095 | 2.58091 | 263338132 | 2.65661 | 2.65661 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:3/real_time | 271220477 | 2.58095 | 2.58095 | 263509929 | 2.65652 | 2.65649 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:3/real_time | 271240587 | 2.58064 | 2.58064 | 263479542 | 2.65671 | 2.65665 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:3/real_time | 271227745 | 2.58083 | 2.58079 | 263402506 | 2.65657 | 2.65657 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:3/real_time | 271255069 | 2.58073 | 2.58071 | 263463858 | 2.65682 | 2.65682 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:3/real_time | 271257174 | 2.58058 | 2.58052 | 263460120 | 2.65682 | 2.65682 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:4/real_time | 174395051 | 4.01401 | 4.01401 | 197330481 | 3.5465 | 3.54636 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:4/real_time | 174645502 | 3.99691 | 3.99691 | 197474831 | 3.54298 | 3.54278 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:4/real_time | 174523308 | 4.01391 | 4.01386 | 197389981 | 3.54518 | 3.54506 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:4/real_time | 174779200 | 3.99874 | 3.99874 | 197519075 | 3.54227 | 3.54209 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:4/real_time | 174642874 | 4.00645 | 4.00641 | 197642101 | 3.54195 | 3.54188 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:4/real_time | 174546754 | 4.0061 | 4.00608 | 197621033 | 3.54296 | 3.54281 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:5/real_time | 162752651 | 4.30119 | 4.30114 | 215552503 | 3.24767 | 3.24752 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:5/real_time | 162717463 | 4.30123 | 4.30116 | 215541082 | 3.24711 | 3.24695 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:5/real_time | 162718819 | 4.3016 | 4.30153 | 215589239 | 3.24725 | 3.24708 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:5/real_time | 162719596 | 4.30151 | 4.30145 | 215563846 | 3.24956 | 3.24949 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:5/real_time | 162753333 | 4.30125 | 4.30125 | 215537315 | 3.24924 | 3.24908 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:5/real_time | 162752258 | 4.3014 | 4.30141 | 215526482 | 3.24744 | 3.24735 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:7/real_time | 143579660 | 4.87526 | 4.87516 | 100000000 | 5.25767 | 5.25752 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:7/real_time | 143585097 | 4.87476 | 4.87467 | 100000000 | 5.41583 | 5.41567 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:7/real_time | 143571011 | 4.87506 | 4.87503 | 182359467 | 3.83773 | 3.83764 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:7/real_time | 143587142 | 4.87487 | 4.8748 | 182397261 | 3.83807 | 3.8379 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:7/real_time | 143578465 | 4.87525 | 4.87521 | 182428602 | 3.83777 | 3.83768 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:7/real_time | 143588555 | 4.87491 | 4.87488 | 125280452 | 5.59791 | 5.59766 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:9/real_time | 284851058 | 2.43476 | 2.43476 | 156879863 | 4.42895 | 4.42884 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:9/real_time | 270700898 | 2.59031 | 2.59024 | 157953114 | 4.42995 | 4.42968 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:9/real_time | 282871172 | 2.45385 | 2.45385 | 157801156 | 4.42817 | 4.42804 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:9/real_time | 285307738 | 2.47009 | 2.47005 | 158058507 | 4.4279 | 4.42786 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:9/real_time | 285709536 | 2.45481 | 2.45476 | 158070961 | 4.42809 | 4.42799 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:9/real_time | 285449733 | 2.47495 | 2.47491 | 158069718 | 4.45026 | 4.45017 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:11/real_time | 189213618 | 3.79684 | 3.79676 | 139459497 | 5.01882 | 5.01871 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:11/real_time | 185600468 | 3.76394 | 3.76376 | 139444892 | 5.01922 | 5.01905 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:11/real_time | 184968668 | 3.80636 | 3.80636 | 139470834 | 5.01948 | 5.01936 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:11/real_time | 183867226 | 3.80432 | 3.80427 | 139481986 | 5.01975 | 5.01944 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:11/real_time | 184301650 | 3.81634 | 3.81634 | 139452846 | 5.01983 | 5.01972 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:11/real_time | 186215795 | 3.82659 | 3.82654 | 139497736 | 5.02119 | 5.02113 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:13/real_time | 135622415 | 5.16256 | 5.16252 | 124661337 | 5.61227 | 5.61194 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:13/real_time | 135618907 | 5.15967 | 5.1596 | 124805224 | 5.6088 | 5.60854 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:13/real_time | 135612192 | 5.15506 | 5.15501 | 124803221 | 5.60901 | 5.60869 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:13/real_time | 135906082 | 5.15818 | 5.15818 | 124776601 | 5.60898 | 5.60886 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:13/real_time | 135369523 | 5.15709 | 5.15682 | 124790370 | 5.60927 | 5.60902 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:13/real_time | 135596827 | 5.1603 | 5.1603 | 124792145 | 5.61637 | 5.61614 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:15/real_time | 110947137 | 5.96511 | 5.96495 | 112861522 | 6.20035 | 6.20014 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:15/real_time | 118004792 | 6.22645 | 6.22628 | 112909900 | 6.20073 | 6.20073 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:15/real_time | 112630319 | 6.25564 | 6.25552 | 112874563 | 6.19932 | 6.19924 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:15/real_time | 117403034 | 6.17263 | 6.17258 | 112927318 | 6.19866 | 6.19842 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:15/real_time | 108921863 | 6.48624 | 6.48612 | 112927746 | 6.20057 | 6.20026 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:15/real_time | 110358148 | 6.66805 | 6.66789 | 112907312 | 6.19938 | 6.19908 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:16/real_time | 203419574 | 3.4415 | 3.44137 | 237134525 | 2.95649 | 2.95638 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:16/real_time | 203414035 | 3.4411 | 3.44099 | 237129564 | 2.95178 | 2.95171 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:16/real_time | 203404068 | 3.44157 | 3.44151 | 236981704 | 2.9518 | 2.95167 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:16/real_time | 203391471 | 3.44146 | 3.44137 | 237108807 | 2.95203 | 2.95196 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:16/real_time | 203393801 | 3.44131 | 3.44127 | 237126460 | 2.95278 | 2.95272 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:16/real_time | 203407476 | 3.44181 | 3.44162 | 237154444 | 2.95293 | 2.9528 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:500/real_time | 37551439 | 18.6407 | 18.6407 | 39222534 | 17.858 | 17.8571 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:500/real_time | 37544097 | 18.6404 | 18.6401 | 39174151 | 17.8539 | 17.8536 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:500/real_time | 37549837 | 18.6391 | 18.6391 | 39233956 | 17.8507 | 17.8505 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:500/real_time | 45996345 | 15.2157 | 15.2153 | 39285929 | 17.848 | 17.8474 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:500/real_time | 46012429 | 15.2184 | 15.2179 | 65664865 | 10.7366 | 10.7364 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:500/real_time | 45912375 | 15.2349 | 15.2346 | 65205908 | 10.8498 | 10.8492 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:4/D:2000/real_time | 9493955 | 73.7232 | 73.7203 | 10188090 | 68.7931 | 68.7908 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:8/D:2000/real_time | 9495562 | 73.7173 | 73.7173 | 10180895 | 68.7533 | 68.7511 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:16/D:2000/real_time | 9487371 | 73.7852 | 73.7831 | 10164473 | 68.7279 | 68.725 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:32/D:2000/real_time | 10816047 | 64.7322 | 64.7287 | 10168481 | 68.8109 | 68.8096 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:64/D:2000/real_time | 10808802 | 64.7232 | 64.721 | 19478320 | 36.1471 | 36.1461 | ns REDUCEMAXIMUMF32KERNEL[]/ByteAligned:128/D:2000/real_time | 10818192 | 64.7304 | 64.728 | 19419672 | 35.9635 | 35.9635 | ns		2024-04-16 13:52:43 -07:00
..
external	Integration with ONNX 1.16.0 (#19745)	2024-04-12 09:46:49 -07:00
patches	Integration with ONNX 1.16.0 (#19745)	2024-04-12 09:46:49 -07:00
tensorboard
adjust_global_compile_flags.cmake	Support xcframework for mac catalyst builds. (#19534)	2024-03-20 10:55:19 -07:00
arm64x.cmake	Build onnxruntime.dll as arm64x (#18633)	2023-12-06 16:49:00 -08:00
CMakeLists.txt	[QNN EP] refactor QNN deps/copy logic. start copying deps to target python loc… (#20317)	2024-04-15 22:33:12 -07:00
CMakeSettings.json
codeconv.runsettings
deps.txt	Integration with ONNX 1.16.0 (#19745)	2024-04-12 09:46:49 -07:00
deps_update_and_upload.py	Update google benchmark to 1.8.3. (#19734)	2024-03-01 11:01:58 -08:00
EnableVisualStudioCodeAnalysis.props
gdk_toolchain.cmake
Info.plist.in
libonnxruntime.pc.cmake.in
linux_arm32_crosscompile_toolchain.cmake	Add a build validation for Linux ARM64 cross-compile (#18200)	2023-11-08 13:03:18 -08:00
linux_arm64_crosscompile_toolchain.cmake	Add a build validation for Linux ARM64 cross-compile (#18200)	2023-11-08 13:03:18 -08:00
maccatalyst_prepare_objects_for_prelink.py	Support xcframework for mac catalyst builds. (#19534)	2024-03-20 10:55:19 -07:00
nuget_helpers.cmake
onnxruntime.cmake	Support xcframework for mac catalyst builds. (#19534)	2024-03-20 10:55:19 -07:00
onnxruntime_codegen_tvm.cmake
onnxruntime_common.cmake	Fix build errors from date/date.h C++20 compatibility (#20139)	2024-04-02 22:10:25 -07:00
onnxruntime_compile_triton_kernel.cmake
onnxruntime_config.h.in	Enabling c++ 20 in MacOS build (#16187)	2023-09-26 11:27:02 -07:00
onnxruntime_csharp.cmake
onnxruntime_flatbuffers.cmake
onnxruntime_framework.cmake
onnxruntime_framework.natvis
onnxruntime_fuzz_test.cmake
onnxruntime_graph.cmake	[Apple framework] Fix minimal build with training enabled. (#19858)	2024-03-12 11:33:30 -07:00
onnxruntime_ios.toolchain.cmake
onnxruntime_java.cmake
onnxruntime_java_unittests.cmake
onnxruntime_kernel_explorer.cmake
onnxruntime_language_interop_ops.cmake
onnxruntime_mlas.cmake	Add vectorized AVX512F kernel for ReduceMaximumF32Kernel (#20268)	2024-04-16 13:52:43 -07:00
onnxruntime_nodejs.cmake	Support building Windows CUDA with Ninja (#20176)	2024-04-03 11:19:31 +08:00
onnxruntime_objectivec.cmake
onnxruntime_opschema_lib.cmake
onnxruntime_optimizer.cmake	[ROCm] Fix hipify error: fast_divmod.h: No such file or directory (#19060)	2024-01-10 14:49:19 +08:00
onnxruntime_providers.cmake	Add initial support for CoreML ML Program to the CoreML EP. (#19347)	2024-02-15 08:46:03 +10:00
onnxruntime_providers_acl.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_armnn.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_azure.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_cann.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_coreml.cmake	Make partitioning utils QDQ aware so it does not break up QDQ node units (#19723)	2024-03-12 10:55:49 +10:00
onnxruntime_providers_cpu.cmake	Revert "Revert NeuralSpeed code for x64 MatMulNBits (#19382)" (#19474)	2024-02-09 09:24:54 -08:00
onnxruntime_providers_cuda.cmake	Enable CUDA EP unit testing on Windows (#20039)	2024-03-27 13:32:36 -07:00
onnxruntime_providers_dml.cmake	Delay load dxcore.dll in addition to ext-ms-win-dxcore-l1-1-0.dll (#18913)	2023-12-26 12:33:42 -08:00
onnxruntime_providers_dnnl.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_js.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_migraphx.cmake	CUDA EP vs ROCM EP hipify audit (#17776)	2023-10-13 10:13:53 +08:00
onnxruntime_providers_nnapi.cmake	Make partitioning utils QDQ aware so it does not break up QDQ node units (#19723)	2024-03-12 10:55:49 +10:00
onnxruntime_providers_openvino.cmake	Ort openvino npu 1.17 master (#19966)	2024-03-21 18:44:00 -07:00
onnxruntime_providers_qnn.cmake	Make partitioning utils QDQ aware so it does not break up QDQ node units (#19723)	2024-03-12 10:55:49 +10:00
onnxruntime_providers_rknpu.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_rocm.cmake	CUDA EP vs ROCM EP hipify audit (#17776)	2023-10-13 10:13:53 +08:00
onnxruntime_providers_tensorrt.cmake	Use CMake's find package for CUDA libs (#19673)	2024-02-27 11:26:48 -08:00
onnxruntime_providers_tvm.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_vitisai.cmake	[VitisAI] Refactor the VAIEP to use MSFT's standalone API (#19058)	2024-01-31 21:08:26 -08:00
onnxruntime_providers_webnn.cmake	Split onnxruntime_providers.cmake to multiple (#17853)	2023-10-09 20:33:44 -07:00
onnxruntime_providers_xnnpack.cmake	Make partitioning utils QDQ aware so it does not break up QDQ node units (#19723)	2024-03-12 10:55:49 +10:00
onnxruntime_pyop.cmake
onnxruntime_python.cmake	[QNN EP] refactor QNN deps/copy logic. start copying deps to target python loc… (#20317)	2024-04-15 22:33:12 -07:00
onnxruntime_rocm_hipify.cmake	add QMoE (#20108)	2024-03-29 10:24:19 -07:00
onnxruntime_session.cmake
onnxruntime_snpe_provider.cmake
onnxruntime_training.cmake
onnxruntime_unittests.cmake	[QNN EP] refactor QNN deps/copy logic. start copying deps to target python loc… (#20317)	2024-04-15 22:33:12 -07:00
onnxruntime_util.cmake
onnxruntime_webassembly.cmake	[WebNN EP] Fixed build issue with disable_rtti (#19173)	2024-01-16 21:35:13 -08:00
precompiled_header.cmake
riscv64.toolchain.cmake	Enable RISC-V 64-bit Cross-Compiling Support for ONNX Runtime on Linux (#19238)	2024-01-24 16:27:05 -08:00
Sdl.ruleset
set_winapi_family_desktop.h
target_delayload.cmake
uwp_stubs.h
wcos_rules_override.cmake	Stop using apiset in OneCore build: use onecoreuap.lib instead of onecoreuap_apiset.lib (#19632)	2024-02-23 22:31:57 -08:00
winml.cmake	[CP] Fix for xfgcheck and Fix WAI ARM64 build (#19634) (#19644)	2024-03-13 17:54:06 -07:00
winml_cppwinrt.cmake
winml_sdk_helpers.cmake
winml_unittests.cmake	Update C/C++ dependencies: abseil, date, nsync, googletest, wil, mp11, cpuinfo and safeint (#15470)	2023-09-08 13:35:04 -07:00