Fix clang vla warning

[cyyever]

Fix the error

/var/lib/jenkins/workspace/third_party/fbgemm/src/FbgemmI8Spmdm.cc:188:41: error: variable length arrays in C++ are a Clang extension [-Werror,-Wvla-cxx-extension]

[netlify]

✅ Deploy Preview for pytorch-fbgemm-docs ready!

Name	Link
🔨 Latest commit	130b2f3
🔍 Latest deploy log	https://app.netlify.com/sites/pytorch-fbgemm-docs/deploys/67afde6dbff8b800084f6b37
😎 Deploy Preview	https://deploy-preview-2736--pytorch-fbgemm-docs.netlify.app
📱 Preview on mobile	Toggle QR Code... Use your smartphone camera to open QR code link.

---

To edit notification comments on pull requests, go to your Netlify site configuration.

[jamesETsmith]

Thanks for working on this @cyyever. I'm still seeing the issue even when manually adding add the flags this PR implements (using clang/clang++ 18.1.8). Do you know if the fixs works for clang 18?

[cyyever]

@jamesETsmith Can you help check compile_commands.json and make sure that the flag is indeed added to the problematic files' command lines?

[jamesETsmith]

@cyyever great idea, it looks like the flag is there (and in stdout during compilation):

From compile_commands.json:

{
  "directory": "/home/james.smith/apps/pytorch/build_qcog",
  "command": "/usr/bin/clang++ -DFBGEMM_STATIC -I/home/james.smith/apps/pytorch/third_party/cpuinfo/include -I/home/james.smith/apps/pytorch/third_party/fbgemm/third_party/asmjit/src -I/home/james.smith/apps/pytorch/third_party/fbgemm/include -I/home/james.smith/apps/pytorch/third_party/fbgemm -isystem /home/james.smith/apps/pytorch/third_party/protobuf/src -isystem /opt/intel/oneapi/mkl/latest/include -isystem /home/james.smith/apps/pytorch/third_party/XNNPACK/include -Wno-vla-cxx-extension -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp=libomp -Wall -Wextra -Werror -Wno-deprecated-declarations -Wimplicit-fallthrough -g -std=c++17 -fPIC -fvisibility=hidden -DMKL_HAS_SBGEMM -DMKL_HAS_SHGEMM -o third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/EmbeddingSpMDM.cc.o -c /home/james.smith/apps/pytorch/third_party/fbgemm/src/EmbeddingSpMDM.cc",
  "file": "/home/james.smith/apps/pytorch/third_party/fbgemm/src/EmbeddingSpMDM.cc",
  "output": "third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/EmbeddingSpMDM.cc.o"
}

From STDOUT:

[26/2892] Building CXX object third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/FbgemmI8Spmdm.cc.o
FAILED: third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/FbgemmI8Spmdm.cc.o 
/usr/bin/ccache /usr/bin/clang++ -DFBGEMM_STATIC -I/home/james.smith/apps/pytorch/third_party/cpuinfo/include -I/home/james.smith/apps/pytorch/third_party/fbgemm/third_party/asmjit/src -I/home/james.smith/apps/pytorch/third_party/fbgemm/include -I/home/james.smith/apps/pytorch/third_party/fbgemm -isystem /home/james.smith/apps/pytorch/third_party/protobuf/src -isystem /opt/intel/oneapi/mkl/latest/include -isystem /home/james.smith/apps/pytorch/third_party/XNNPACK/include -Wno-vla-cxx-extension -D_GLIBCXX_USE_CXX11_ABI=1 -fvisibility-inlines-hidden -DUSE_PTHREADPOOL -fopenmp=libomp -Wall -Wextra -Werror -Wno-deprecated-declarations -Wimplicit-fallthrough -g -std=c++17 -fPIC -fvisibility=hidden -DMKL_HAS_SBGEMM -DMKL_HAS_SHGEMM -MD -MT third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/FbgemmI8Spmdm.cc.o -MF third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/FbgemmI8Spmdm.cc.o.d -o third_party/fbgemm/CMakeFiles/fbgemm_generic.dir/src/FbgemmI8Spmdm.cc.o -c /home/james.smith/apps/pytorch/third_party/fbgemm/src/FbgemmI8Spmdm.cc
/home/james.smith/apps/pytorch/third_party/fbgemm/src/FbgemmI8Spmdm.cc:86:32: error: variable length arrays in C++ are a Clang extension [-Werror,-Wvla-cxx-extension]
   86 |   alignas(64) uint8_t A_buffer[K * 32];

[cyyever]

@jamesETsmith That flag is overwritten by '-Wall -Wextra -Werror' later. That means there is a flag order problem. Check where you added the flag and move it behind Wall

[jamesETsmith]

Thanks for the heads up @cyyever. I'm building FBGEMM through PyTorch and it doesn't seem like I can change the ordering of the CMAKE_CXX_FLAGS from the command line without modifying the cmake source like you have for FBGEMM or PyTorch. I replicated the changes from your pytorch PR and that seems to have worked.

Thanks for the work on this, I hope it gets merged.

[facebook-github-bot]

@q10 has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.