CI: make sure to use Clang as CUDA compiler on Clang+CUDA builds

- set CMAKE_CUDA_COMPILER=clang++ in CI script
- changed CUDA arch specification with SPIRIT_CUDA_ARCH to number only (dropped prefix "sm_") in order to be able to set CMAKE_CUDA_ARCHITECTURES needed for Clang properly (which only expects a number)

Author: muellan

.github/workflows/ci.yml

@@ -158,16 +158,16 @@ jobs:
         sudo update-alternatives --install /usr/bin/clang clang /usr/bin/clang-16 16
         sudo update-alternatives --install /usr/bin/clang++ clang++ /usr/bin/clang++-16 16
 
-    - name: 📚 Install NVIDIA HPC SDK
+    - name: 📚 Install CUDA Toolkit
       if: matrix.parallelisation == 'cuda'
       shell: bash
       run: |
-        curl https://developer.download.nvidia.com/hpc-sdk/ubuntu/DEB-GPG-KEY-NVIDIA-HPC-SDK | sudo gpg --dearmor -o /usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg
-        echo 'deb [signed-by=/usr/share/keyrings/nvidia-hpcsdk-archive-keyring.gpg] https://developer.download.nvidia.com/hpc-sdk/ubuntu/amd64 /' | sudo tee /etc/apt/sources.list.d/nvhpc.list
-        sudo apt update -y
-        sudo apt install -y nvhpc-23-3
-        echo "/opt/nvidia/hpc_sdk/Linux_x86_64/23.3/compilers/bin" >> $GITHUB_PATH
-        echo "/opt/nvidia/hpc_sdk/Linux_x86_64/23.3/math_libs/lib64" >> $GITHUB_PATH
+        wget https://developer.download.nvidia.com/compute/cuda/repos/ubuntu2204/x86_64/cuda-keyring_1.0-1_all.deb
+        sudo dpkg -i cuda-keyring_1.0-1_all.deb
+        sudo apt-get update
+        sudo apt-get -y install cuda
+        echo "/usr/local/cuda-12.1/bin" >> $GITHUB_PATH
+        echo "/usr/local/cuda-12.1/lib64" >> $GITHUB_PATH
 
     - name: 📚 Install LLVM (MacOS)
       if: matrix.os == 'macos-13'
@@ -199,7 +199,7 @@ jobs:
       if: matrix.parallelisation == 'cuda'
       shell: bash
       working-directory: ${{runner.workspace}}/build
-      run: cmake $GITHUB_WORKSPACE -DCMAKE_BUILD_TYPE=$BUILD_TYPE $CMAKE_FLAGS -DSPIRIT_USE_CUDA=ON -DCUDA_TOOLKIT_ROOT_DIR=/opt/nvidia/hpc_sdk/Linux_x86_64/23.3/cuda/
+      run: cmake $GITHUB_WORKSPACE -DCMAKE_CUDA_COMPILER=clang++ -DSPIRIT_CUDA_ARCH=70 -DCMAKE_BUILD_TYPE=$BUILD_TYPE $CMAKE_FLAGS -DSPIRIT_USE_CUDA=ON -DCUDA_TOOLKIT_ROOT_DIR=/usr/local/cuda-12.1/
 
     - name: ⚙ Configure with OpenMP
       if: matrix.parallelisation == 'openmp'
@@ -478,7 +478,7 @@ jobs:
           - os: ubuntu-22.04
     env:
       BUILD_TYPE: Release
-      CMAKE_FLAGS: -DSPIRIT_UI_USE_IMGUI=OFF -DSPIRIT_UI_CXX_USE_QT=OFF -DSPIRIT_USE_CUDA=ON -DSPIRIT_CUDA_ARCH=sm_70
+      CMAKE_FLAGS: -DSPIRIT_UI_USE_IMGUI=OFF -DSPIRIT_UI_CXX_USE_QT=OFF -DSPIRIT_USE_CUDA=ON -DSPIRIT_CUDA_ARCH=70
 
     steps:
     - uses: actions/checkout@v3
@@ -763,4 +763,4 @@ jobs:
         tag_name: ${{ github.ref }}
         release_name: Release ${{ github.ref }}
         body: ${{ steps.changelog.outputs.changelog }}
-        draft: true
+        draft: true

CMakeLists.txt

@@ -26,8 +26,8 @@ set( SPIRIT_USE_THREADS       OFF  CACHE BOOL "Use std threads to speed up certa
 set( SPIRIT_USE_FFTW          ON   CACHE BOOL "If available, use the FFTW library instead of kissFFT." )
 ### Set the scalar type used in the Spirit library
 set( SPIRIT_SCALAR_TYPE "double" CACHE STRING "The scalar type to be used in the Spirit library." )
-### Set the compute capability for CUDA compilation
-set( SPIRIT_CUDA_ARCH   "sm_60"  CACHE STRING "The CUDA compute architecture to use in case of a CUDA build." )
+### Set the compute capability for CUDA compilation (note: specify without prefix "sm_")
+set( SPIRIT_CUDA_ARCH   "60"  CACHE STRING "The CUDA compute architecture to use in case of a CUDA build." )
 ####################################################################
 ### CMake Verbosity
 option( SPIRIT_PRINT_SOURCES    "Print Headers and Sources from Cmake."                          OFF )
@@ -303,4 +303,4 @@ install( FILES ${CMAKE_CURRENT_LIST_DIR}/ui-cpp/ui-imgui/thirdparty/stb/LICENSE
 
 ######### Write VERSION.txt ########################################
 file( WRITE "${CMAKE_SOURCE_DIR}/VERSION.txt" "${SPIRIT_META_NAME_VERSION}" )
-####################################################################
+####################################################################

core/CMakeLists.txt

@@ -28,7 +28,7 @@ option( SPIRIT_USE_FFTW         "If available, use the FFTW library instead of k
 ### Set the scalar type used in the Spirit library
 option( SPIRIT_SCALAR_TYPE      "Use std threads to speed up certain parts of the code." "double" )
 ### Set the compute capability for CUDA compilation
-option( SPIRIT_CUDA_ARCH        "Use std threads to speed up certain parts of the code." "sm_60"  )
+option( SPIRIT_CUDA_ARCH        "Use std threads to speed up certain parts of the code." "60"  )
 ####################################################################
 
 
@@ -146,20 +146,28 @@ if( SPIRIT_USE_CUDA )
     set( SPIRIT_CUDA_LIBS "${CUDA_LIBRARIES};${CUDA_CUFFT_LIBRARIES};${CUDA_curand_LIBRARY}" )
     set( FFT_LIB          ${CUDA_CUFFT_LIBRARIES} )
 
-    if( "${CMAKE_CXX_COMPILER_ID}" MATCHES "NVHPC" OR "${CMAKE_CXX_COMPILER_ID}" MATCHES "NVCC" ) 
-        set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -lineinfo -arch=${SPIRIT_CUDA_ARCH} --expt-relaxed-constexpr --expt-extended-lambda" )
+    if( "${CMAKE_CUDA_COMPILER_ID}" MATCHES "NVIDIA" ) 
+        set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -lineinfo -arch=sm_${SPIRIT_CUDA_ARCH} --expt-relaxed-constexpr --expt-extended-lambda" )
         ### Deactivate CUDA warning inside Eigen such as "warning: __host__ annotation is ignored on a function("Quaternion") that is explicitly defaulted on its first declaration"
         set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe=\"--diag_suppress=esa_on_defaulted_function_ignored\"")
         ### Display warning number when writing a warning
         set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -Xcudafe \"--display_error_number\"" )
 
         set( META_COMPILER         "${META_COMPILER} and nvcc" )
         set( META_COMPILER_VERSION "${META_COMPILER_VERSION} and ${CUDA_VERSION}" )
-        set( META_COMPILER_FULL    "${META_COMPILER_FULL} and nvcc (${CUDA_VERSION}) for cuda arch \\\"${SPIRIT_CUDA_ARCH}\\\"" )
+        set( META_COMPILER_FULL    "${META_COMPILER_FULL} and nvcc (${CUDA_VERSION}) for cuda arch \\\"sm_${SPIRIT_CUDA_ARCH}\\\"" )
+
+    elseif( "${CMAKE_CXX_COMPILER_ID}" MATCHES "NVHPC") 
+        set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -cuda" )
+
+        set( META_COMPILER         "${META_COMPILER}" )
+        set( META_COMPILER_VERSION "${META_COMPILER_VERSION} and ${CUDA_VERSION}" )
+        set( META_COMPILER_FULL    "${META_COMPILER_FULL} and nvcc (${CUDA_VERSION}) for cuda arch \\\"sm_${SPIRIT_CUDA_ARCH}\\\"" )
+
     else()
+        set( CMAKE_CUDA_ARCHITECTURES ${SPIRIT_CUDA_ARCH} )
         # set( CMAKE_CUDA_FLAGS "${CMAKE_CUDA_FLAGS} -D__STRICT_ANSI__" )
         # set( SPIRIT_COMPILE_DEFINITIONS "${SPIRIT_COMPILE_DEFINITIONS} -D__STRICT_ANSI__" )
-
         set( META_COMPILER         "${META_COMPILER}" )
         set( META_COMPILER_VERSION "${META_COMPILER_VERSION} and CUDA ${CUDA_VERSION}" )
         set( META_COMPILER_FULL    "${META_COMPILER_FULL} and (CUDA ${CUDA_VERSION}) for cuda arch \\\"${SPIRIT_CUDA_ARCH}\\\"" )
@@ -741,4 +749,4 @@ if( SPIRIT_PRINT_SOURCES )
 endif()
 MESSAGE( STATUS ">> --------------------- Spirit done ---------------------------------- <<" )
 message( STATUS ">> CMake CXX Flags:        ${CMAKE_CXX_FLAGS}" )
-MESSAGE( STATUS ">> -------------------------------------------------------------------- <<" )
+MESSAGE( STATUS ">> -------------------------------------------------------------------- <<" )