Compiler setup blocks usage of CLANG for CUDA

[PhilipDeegan]

SAMRAI/cmake/SetupCompilers.cmake

Line 6 in d84e4e7

set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -arch ${CUDA_ARCH} --expt-extended-lambda --expt-relaxed-constexpr")

my cmake config is here

[PhilipDeegan]

would a compiler detection PR be considered here?

if("${CMAKE_CXX_COMPILER_ID}" STREQUAL "PGI")

    # Set specific options for CUDA if enabled
    if (ENABLE_RAJA AND ENABLE_CUDA)
      # RAJA requires some experimental features
      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -arch ${CUDA_ARCH} --expt-extended-lambda --expt-relaxed-constexpr")
    endif ()
    
    # TODO Ensure openmp flags are not enabled twice!
    if (ENABLE_OPENMP AND ENABLE_CUDA)
      # CMAKE_CUDA_FLAGS does not pass OpenMP_CXX_FLAGS to the host compiler by default
      set(CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcompiler ${OpenMP_CXX_FLAGS}")
    endif ()

endif() 

I've commented out both calls to "set" locally and it all still compiles with -DCMAKE_CUDA_FLAGS="-x cuda"

or perhaps here it should detect clang and add "-x cuda"?