Integration of CUDA kernels #298
Description
@Chiil and collaborators have built CUDA versions of the RTE and RRTMGP kernels. We'd like to get these to interoperate with the Fortran front-end (and the C++ front end too).
What's still needed:
- Get CI running in the CUDA kernels library
- Add a CI workflow to this repo that runs the Fortran tests using the CUDA kernels
- Add OpenACC data statements so that all data is moved to the device by the Fortran front end (so no acc data pragmas are needed in the CUDA kernels)