Skip to content

[libtorch] [xnnpack] [qnnpack] [cpuinfo] #46649

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom
Open
wants to merge 124 commits into
base: master
Choose a base branch
from
Open

[libtorch] [xnnpack] [qnnpack] [cpuinfo] #46649

wants to merge 124 commits into from

Conversation

EvilMcStevil
Copy link
Contributor

@EvilMcStevil EvilMcStevil commented Jul 29, 2025 •
edited
Loading

  • Changes comply with the maintainer guide.
  • SHA512s are updated for each updated download.
  • The "supports" clause reflects platforms that may be fixed by this new version.
  • Any fixed CI baseline entries are removed from that file.
  • Any patches that are no longer applied are deleted from the port's directory.
  • The version database is fixed by rerunning ./vcpkg x-add-version --all and committing the result.
  • Only one version is added to each modified port's versions file.

Steve and others added 30 commits July 29, 2025 15:30
updated cpuinfo for to latest version
9563cdc
added baseline
e43405a
updated qnnpack to build again with new cpuinfo
6592e67
update qnnpack baseline
f812df8
added xnnpack change
47757b9
add xnnpack baseline
79876b6
added clog dep to cpuinfo port file
1c30a7a
updated qnnpack to use clog compiled from cpuinfo
05abf18
format manifests
fe4678f
updated version baselines
52c2b63
added python to xnnpack build requirements
c01f539
added kliedai as dependecy to arm64
b9e9d25
diabled cpu info clog tests
61738eb
fixed spelling mistake on kleidiai dependency
f4d165d
Update fbgemm to 1.0.0
6b0f748
update fbgemm version database
bf258da
removed failure mark for fbgemm 64-android which passed ci
de67d9e
updated to simplify patch
eb49ccf
removed bad spacing in diff to simplify again
a5fbdce
updated baseline
7e3c42c
initial patches for libtorch
8de7a89
some cmake deps fixed
0868e2f
updated nvidia cutlass and added utils tools as needed by libtorch 2.7.1
2d0f2ee
Added baseline
8c686c4
removed unneeded patch after building
a4432dd
fixed baseline
f0e6817
updated libtorch
0f86d75
added nvtx3
e546f9e
updated libtorch port to compile a shared object
54c572c
moved plibtorch patches back
a7daa4c
Steve and others added 25 commits August 7, 2025 11:28
linux feature fail
82d903e
updated fbgem cmake bindings to export dependecies and correctly add …
c1daa8a 
…FGEMM_STATIC to them
added fbgemm baseline
d23358a
added patch back in
446732a
xnnpack debug cmake install fixed
886c58a
updated baselines
79937ed
fixed magma baseline file
c6b83a5
fixed onnx dependency
6b76e25
osx fix
3741239
@EvilMcStevil
Build output dir fixes
b64387e
updated libtorch binding
afe97ec
@EvilMcStevil
added vulkan fix patch back, fixed error in CUDA port
5bd952f
@EvilMcStevil
updated version binding
7e6f590
Removed vulkan cascade error
30b8528
added vulkan fix
359267a
marking windows cuda as expected to fail in ci,
7f4a210 
[1441/1606] C:\PROGRA~1\NVIDIA~2\CUDA\v12.9\bin\nvcc.exe -forward-unknown-to-host-compiler -DABSL_CONSUME_DLL -DAT_PER_OPERATOR_HEADERS -DCUTLASS_ENABLE_CUBLAS=1 -DCUTLASS_ENABLE_CUDNN=1 -DEXPORT_AOTI_FUNCTIONS -DFMT_HEADER_ONLY=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DNOMINMAX -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DPROTOBUF_USE_DLLS -DTORCH_CUDA_BUILD_MAIN_LIB -DTORCH_CUDA_USE_NVTX3 -DUSE_CUDA -DUSE_EXTERNAL_MZCRC -DUSE_MEM_EFF_ATTENTION -DUSE_MIMALLOC -DWIN32_LEAN_AND_MEAN -D_CRT_SECURE_NO_DEPRECATE=1 -D_UCRT_LEGACY_INFINITY -Dtorch_cuda_EXPORTS -ID:\b\libtorch\x64-windows-release-rel\aten\src -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src -ID:\b\libtorch\x64-windows-release-rel -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\THC -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\cuda -ID:\b\libtorch\x64-windows-release-rel\caffe2\aten\src -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\c10\cuda\..\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\c10\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\torch\csrc\api -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\torch\csrc\api\include -isystem D:\installed\x64-windows-release\include -isystem D:\installed\x64-windows-release\include\eigen3 -isystem "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9\include" -isystem D:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\cmake\..\third_party\cudnn_frontend\include -DLIBCUDACXX_ENABLE_SIMPLIFIED_COMPLEX_OPERATIONS -Xcompiler  /Zc:__cplusplus -Xcompiler /w -w -Xcompiler /FS -Xfatbin -compress-all -DONNX_NAMESPACE=onnx --use-local-env -gencode arch=compute_50,code=sm_50 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_89,code=sm_89 -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90a,code=sm_90a -gencode arch=compute_100,code=sm_100 -gencode arch=compute_100a,code=sm_100a -gencode arch=compute_101a,code=sm_101a -gencode arch=compute_120,code=sm_120 -gencode arch=compute_120a,code=sm_120a -Xcudafe --diag_suppress=cc_clobber_ignored,--diag_suppress=field_without_dll_interface,--diag_suppress=base_class_has_different_dll_interface,--diag_suppress=dll_interface_conflict_none_assumed,--diag_suppress=dll_interface_conflict_dllexport_assumed,--diag_suppress=bad_friend_decl --Werror cross-execution-space-call --no-host-device-move-forward --expt-relaxed-constexpr --expt-extended-lambda  -Xcompiler=/wd4819,/wd4503,/wd4190,/wd4244,/wd4251,/wd4275,/wd4522 -Wno-deprecated-gpu-targets --expt-extended-lambda -DCUB_WRAPPED_NAMESPACE=at_cuda_detail -DCUDA_HAS_FP16=1 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -Xcompiler="-O2 -Ob2" -DNDEBUG -Xcompiler /MD -std=c++17 -Xcompiler=-MD -MD -MT caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj -MF caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj.d -x cu -c D:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\native\cuda\SegmentReduce.cu -o caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj -Xcompiler=-Fdcaffe2\CMakeFiles\torch_cuda.dir\,-FS
FAILED: caffe2/CMakeFiles/torch_cuda.dir/__/aten/src/ATen/native/cuda/SegmentReduce.cu.obj
C:\PROGRA~1\NVIDIA~2\CUDA\v12.9\bin\nvcc.exe -forward-unknown-to-host-compiler -DABSL_CONSUME_DLL -DAT_PER_OPERATOR_HEADERS -DCUTLASS_ENABLE_CUBLAS=1 -DCUTLASS_ENABLE_CUDNN=1 -DEXPORT_AOTI_FUNCTIONS -DFMT_HEADER_ONLY=1 -DMINIZ_DISABLE_ZIP_READER_CRC32_CHECKS -DNOMINMAX -DONNXIFI_ENABLE_EXT=1 -DONNX_ML=1 -DONNX_NAMESPACE=onnx -DPROTOBUF_USE_DLLS -DTORCH_CUDA_BUILD_MAIN_LIB -DTORCH_CUDA_USE_NVTX3 -DUSE_CUDA -DUSE_EXTERNAL_MZCRC -DUSE_MEM_EFF_ATTENTION -DUSE_MIMALLOC -DWIN32_LEAN_AND_MEAN -D_CRT_SECURE_NO_DEPRECATE=1 -D_UCRT_LEGACY_INFINITY -Dtorch_cuda_EXPORTS -ID:\b\libtorch\x64-windows-release-rel\aten\src -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src -ID:\b\libtorch\x64-windows-release-rel -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\THC -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\cuda -ID:\b\libtorch\x64-windows-release-rel\caffe2\aten\src -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\c10\cuda\..\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\c10\.. -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\torch\csrc\api -ID:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\torch\csrc\api\include -isystem D:\installed\x64-windows-release\include -isystem D:\installed\x64-windows-release\include\eigen3 -isystem "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.9\include" -isystem D:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\cmake\..\third_party\cudnn_frontend\include -DLIBCUDACXX_ENABLE_SIMPLIFIED_COMPLEX_OPERATIONS -Xcompiler  /Zc:__cplusplus -Xcompiler /w -w -Xcompiler /FS -Xfatbin -compress-all -DONNX_NAMESPACE=onnx --use-local-env -gencode arch=compute_50,code=sm_50 -gencode arch=compute_80,code=sm_80 -gencode arch=compute_86,code=sm_86 -gencode arch=compute_89,code=sm_89 -gencode arch=compute_90,code=sm_90 -gencode arch=compute_90a,code=sm_90a -gencode arch=compute_100,code=sm_100 -gencode arch=compute_100a,code=sm_100a -gencode arch=compute_101a,code=sm_101a -gencode arch=compute_120,code=sm_120 -gencode arch=compute_120a,code=sm_120a -Xcudafe --diag_suppress=cc_clobber_ignored,--diag_suppress=field_without_dll_interface,--diag_suppress=base_class_has_different_dll_interface,--diag_suppress=dll_interface_conflict_none_assumed,--diag_suppress=dll_interface_conflict_dllexport_assumed,--diag_suppress=bad_friend_decl --Werror cross-execution-space-call --no-host-device-move-forward --expt-relaxed-constexpr --expt-extended-lambda  -Xcompiler=/wd4819,/wd4503,/wd4190,/wd4244,/wd4251,/wd4275,/wd4522 -Wno-deprecated-gpu-targets --expt-extended-lambda -DCUB_WRAPPED_NAMESPACE=at_cuda_detail -DCUDA_HAS_FP16=1 -D__CUDA_NO_HALF_OPERATORS__ -D__CUDA_NO_HALF_CONVERSIONS__ -D__CUDA_NO_HALF2_OPERATORS__ -D__CUDA_NO_BFLOAT16_CONVERSIONS__ -Xcompiler="-O2 -Ob2" -DNDEBUG -Xcompiler /MD -std=c++17 -Xcompiler=-MD -MD -MT caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj -MF caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj.d -x cu -c D:\b\libtorch\src\v2.7.1-2650bb1bd5.clean\aten\src\ATen\native\cuda\SegmentReduce.cu -o caffe2\CMakeFiles\torch_cuda.dir\__\aten\src\ATen\native\cuda\SegmentReduce.cu.obj -Xcompiler=-Fdcaffe2\CMakeFiles\torch_cuda.dir\,-FS
LLVM ERROR: out of memory
made fbgemm only for x64 builds
dbf1ad3
uopdated baseline
3879342
fixed use fbgemm override
a6cde9e
added kleidai as dependecy in cmake binding of xnnpack
bc84d17
@EvilMcStevil
added fix glog back - needed on osx builds
93817f2
renamed fix build to fix cmake to make comparisons with luncliffs por…
cbc4358 
…t easier
added correct find library to xnnpack for kleidiai
cbf6355
removed cuda from default windows build
723b749
@EvilMcStevil
removed cascade failures for arm64 osx
54f94c4
@EvilMcStevil EvilMcStevil marked this pull request as ready for review August 10, 2025 01:28
@EvilMcStevil
Copy link
Contributor Author

Also credit to @luncliff as i shamelessly stole the OSX fixes from #46700

luncliff and others added 2 commits August 10, 2025 14:15
@luncliff @EvilMcStevil
[libtorch] patch for CUDA build in Windows
9aac013 
# Conflicts:
#	ports/libtorch/fix-cmake.patch
#	ports/libtorch/portfile.cmake
#	versions/l-/libtorch.json
@EvilMcStevil
made cuda builds expect to complete
cd9e721
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Reviewers

@dg0yt dg0yt dg0yt left review comments

At least 1 approving review is required to merge this pull request.

Assignees
No one assigned
Labels
depends:vm-update PR contains changes to the VM provisioning scripts
Projects
None yet
Milestone
No milestone
Development

Successfully merging this pull request may close these issues.
4 participants
@EvilMcStevil @BillyONeal @dg0yt @luncliff