disable boost.atomic in alpaka to avoid building it on linux

.github/workflows/ci.yaml

@@ -51,7 +51,7 @@ jobs:
       run: |
         mkdir build
         cd build
-        cmake .. -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DCMAKE_BUILD_TYPE=$CONFIG -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=ON -DALPAKA_CXX_STANDARD=17 -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
+        cmake .. -DCMAKE_EXPORT_COMPILE_COMMANDS=ON -DCMAKE_BUILD_TYPE=$CONFIG -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=ON -DALPAKA_ACC_CPU_DISABLE_ATOMIC_REF=ON -DALPAKA_CXX_STANDARD=17 -DCMAKE_TOOLCHAIN_FILE=/usr/local/share/vcpkg/scripts/buildsystems/vcpkg.cmake
         sed -i 's/\(-forward-unknown-to-host-compiler\|--generate-code=arch=[^ ]\+\|--expt-extended-lambda\|--expt-relaxed-constexpr\|--use_fast_math\)//g' compile_commands.json # remove NVCC specific flags which clang cannot handle
         run-clang-tidy-12 -header-filter='^((?!/thirdparty/).)*$' -extra-arg=--no-cuda-version-check -extra-arg=-nocudalib
 
@@ -228,7 +228,7 @@ jobs:
             unset CUDACXX
           fi
           echo "nvcc is here: $CUDACXX"
-          cmake .. -DCMAKE_BUILD_TYPE=$CONFIG -DLLAMA_ENABLE_ASAN_FOR_TESTS=ON -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=${{ !matrix.cuda }} -DALPAKA_ACC_GPU_CUDA_ENABLE=${{ matrix.cuda }} -DALPAKA_CXX_STANDARD=17 -DCMAKE_CUDA_COMPILER=$CUDACXX -DCMAKE_TOOLCHAIN_FILE=$VCPKG_INSTALLATION_ROOT/scripts/buildsystems/vcpkg.cmake
+          cmake .. -DCMAKE_BUILD_TYPE=$CONFIG -DLLAMA_ENABLE_ASAN_FOR_TESTS=ON -DALPAKA_ACC_CPU_B_SEQ_T_SEQ_ENABLE=${{ !matrix.cuda }} -DALPAKA_ACC_CPU_DISABLE_ATOMIC_REF=ON -DALPAKA_ACC_GPU_CUDA_ENABLE=${{ matrix.cuda }} -DALPAKA_CXX_STANDARD=17 -DCMAKE_CUDA_COMPILER=$CUDACXX -DCMAKE_TOOLCHAIN_FILE=$VCPKG_INSTALLATION_ROOT/scripts/buildsystems/vcpkg.cmake
       - name: build tests + examples
         run: |
           if [ ${{ matrix.install_oneapi }} ]; then source /opt/intel/oneapi/setvars.sh; fi