Compile bug: common.cuh defines clash with winAPI #10575
Description
Git commit
Operating systems
Windows
GGML backends
CUDA
Problem description & steps to reproduce
Whilst trying to improve performance on windows, I tried to include nvtx3/nvToolsExt.h in ggml-cuda.cu.
This breaks compilation, because there is a naming clash between common.cuh and oaidl.h(winapi, enum tagCALLCONV @ Line 812).
To be specific, CC_PASCAL is set to 600. At this place, the GPU arch pascal is referenced. In WinAPI, an identical name exists for pascal calling conventions.
On windows, this could create issues for many libraries who interface with winAPI
I would suggest prefixing or renaming CC_PASCAL et al.
First Bad Commit
No response
Relevant log output
FAILED: ggml/src/ggml-cuda/CMakeFiles/ggml-cuda.dir/ggml-cuda.cu.obj
C:\PROGRA~1\NVIDIA~2\CUDA\v12.6\bin\nvcc.exe -forward-unknown-to-host-compiler -DGGML_BACKEND_BUILD -DGGML_BACKEND_SHARED -DGGML_CUDA_PEER_MAX_BATCH_SIZE=128 -DGGML_CUDA_USE_GRAPHS -DGGML_SCHED_MAX_COPIES=4 -DGGML_SHARED -D_CRT_SECURE_NO_WARNINGS -D_XOPEN_SOURCE=600 -Dggml_cuda_EXPORTS -IC:\Users\akieslinger\CLionProjects\llama.cpp\ggml\src\ggml-cuda\. -IC:\Users\akieslinger\CLionProjects\llama.cpp\ggml\src\ggml-cuda\.. -IC:\Users\akieslinger\CLionProjects\llama.cpp\ggml\src\..\include -isystem "C:\Program Files\NVIDIA GPU Computing Toolkit\CUDA\v12.6\include" -D_WINDOWS -Xcompiler=" /GR /EHsc" -Xcompiler="-MD -Zi -O2 -Ob1" -DNDEBUG "--generate-code=arch=compute_89,code=[compute_89,sm_89]" -use_fast_math -MD -MT ggml\src\ggml-cuda\CMakeFiles\ggml-cuda.dir\ggml-cuda.cu.obj -MF ggml\src\ggml-cuda\CMakeFiles\ggml-cuda.dir\ggml-cuda.cu.obj.d -x cu -c C:\Users\akieslinger\CLionProjects\llama.cpp\ggml\src\ggml-cuda\ggml-cuda.cu -o ggml\src\ggml-cuda\CMakeFiles\ggml-cuda.dir\ggml-cuda.cu.obj -Xcompiler=-Fdggml\src\ggml-cuda\CMakeFiles\ggml-cuda.dir\,-FS
C:\Program Files (x86)\Windows Kits\10\include\10.0.22621.0\um\oaidl.h(812): error: expected an identifier
600 = CC_MSCPASCAL,