Build error on Windows: CUDA_ARCHITECTURES is empty for target "cmTC_0c70f"

[useronym]

Visual Studio Community 2022 (MSVC 143)
Cuda 11.6
CMake 3.22.1

Lots of compile errors
CMakeError.log
CMakeOutput.log

Am I using a too-new buildchain?

[liguanglin]

https://stackoverflow.com/questions/67966258/cuda-architectures-is-empty-for-target-cmtc-28d80

[useronym]

I have tried to set the architecture manually according to that, but it has no effect. I suspect the error message is not completely relevant here, as there's a lot of compile errors (seen in CMakeError.log) when compiling include files:

#$ cicc --microsoft_version=1930 --msvc_target_version=1930 --compiler_bindir "C:/Program Files/Microsoft Visual Studio/2022/Community/VC/Tools/MSVC/14.30.30705/bin/HostX86/x86/../../../../../../.." --sdk_dir "C:/Program Files (x86)/Windows Kits/10/" --display_error_number --orig_src_file_name "CMakeCUDACompilerId.cu" --orig_src_path_name "E:\instant-ngp\build\CMakeFiles\3.22.1\CompilerIdCUDA\CMakeCUDACompilerId.cu" --allow_managed  -arch compute_52 -m64 --no-version-ident -ftz=0 -prec_div=1 -prec_sqrt=1 -fmad=1 --include_file_name "CMakeCUDACompilerId.fatbin.c" -tused --gen_module_id_file --module_id_file_name "tmp/CMakeCUDACompilerId.module_id" --gen_c_file_name "tmp/CMakeCUDACompilerId.cudafe1.c" --stub_file_name "tmp/CMakeCUDACompilerId.cudafe1.stub.c" --gen_device_file_name "tmp/CMakeCUDACompilerId.cudafe1.gpu"  "tmp/CMakeCUDACompilerId.cpp1.ii" -o "tmp/CMakeCUDACompilerId.ptx"
C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime.h(197): error: invalid redeclaration of type name "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(48): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(53): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(59): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(64): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(166): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new.h(181): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new_debug.h(26): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\vcruntime_new_debug.h(34): error: first parameter of allocation function must be of type "size_t"

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\type_traits(381): error: class template "std::_Is_memfunptr" has already been defined

C:\Program Files\Microsoft Visual Studio\2022\Community\VC\Tools\MSVC\14.30.30705\include\type_traits(381): error: class template "std::_Is_memfunptr" has already been defined

...

Compiling the CUDA compiler identification source file "CMakeCUDACompilerId.cu" failed.

[somisawa]

I solved the same error by setting CMAKE_CUDA_COMPILER. When running cmake, I ran cmake . -B build -DCMAKE_CUDA_COMPILER=/usr/local/cuda-<your cuda version>/bin/nvcc and then, this was solved in my case.

[useronym]

Unfortunately that has no effect either, in fact nvcc is already on my path anyway

[jaymefosa]

I had this error on Ubuntu 18.04 and although NVCC was in the path, setting it directly as somisawa recommended worked

[somisawa]

@jaymefosa
That's nice. And if you face other warnings such as GPU not detected, it is also possibility caused by environmental arguments. Then, see Issue #28 .

[useronym]

Managed to get it to build. Instead of using the VS2022 developer powershell to run CMake directly, I imported the folder into VS and use the CMake integration. VS managed to do the build without any extra settings. Sounds like maybe there is something wrong with the VS2022 developer shell ¯_(ツ)_/¯

[liuchangf]

I solved the problem by following code
cmake . -B build -DCMAKE_CUDA_COMPILER="C:/Program Files/NVIDIA GPU Computing Toolkit/CUDA/v10.2/bin/nvcc.exe" -G"Visual Studio 15 2017 Win64"

[limberc]

You may installed the nvidia driver and nvcc in some way. For latest version, you can install nvidia driver using the 3rd party softwares. They may not identify you as a programmer and modify the CUDA path to somewhere else.