CUDA: Don't fail to detect the architecture just because NVCC doesn't support the system's C++ compiler
I'm using cmake 3.23.0-rc2. On a system with GCC 10 being the default C++ compiler and CUDA 10.2.89 installed, trying to configure a CUDA project yields:
CMake Error at /opt/versions/cmake/3.23.0-rc2/share/cmake-3.23/Modules/CMakeDetermineCUDACompiler.cmake:632 (message):
Failed to detect a default CUDA architecture.
Compiler output:
Call Stack (most recent call first):
CMakeLists.txt:19 (PROJECT)
This statement is not true. That is, yes, you failed to detect, but you really don't have to. You can compile a perfectly mundane C++ program that will give you the information you need; or you could try nvidia-smi; etc.
Related to #23322 .
Edited by Brad King