Note that the performance warnings from numba.cuda are potentially helpful but not especially concerning. We know we are under utilizing the GPU in some areas which is why we are using CUDA MPS to have multiple MPI ranks share GPUs. Numba is not aware of that.
This PR addresses #69 and #70.
Note that the performance warnings from numba.cuda are potentially helpful but not especially concerning. We know we are under utilizing the GPU in some areas which is why we are using CUDA MPS to have multiple MPI ranks share GPUs. Numba is not aware of that.