Replace compile_cpp_as_cuda(...) with CUDA runtime header inclusion
authorArtem Zhmurov <zhmurov@gmail.com>
Fri, 12 Jun 2020 08:12:17 +0000 (08:12 +0000)
committerPaul Bauer <paul.bauer.q@gmail.com>
Fri, 12 Jun 2020 08:12:17 +0000 (08:12 +0000)
commit6993ee6992573566e7b7083858e23c92aef1d3dd
tree47b3c7228972d188acc91826f7de44136eedf934
parent766a2367b61fe0424400eb36a19e45e9bab90c94
Replace compile_cpp_as_cuda(...) with CUDA runtime header inclusion

To make CUDA-specific tipes visible, one should either use CUDA compiler
for the files where they are exposed, or include the CUDA runtime header
for the normal compiler. This changes first approach to the second.
src/gromacs/ewald/CMakeLists.txt
src/gromacs/ewald/tests/CMakeLists.txt
src/gromacs/gpu_utils/CMakeLists.txt
src/gromacs/gpu_utils/device_stream.h
src/gromacs/mdtypes/CMakeLists.txt