Finite difference wave propagation involves a type of computation that GPUs
(graphics processing units) are suitable for. Creating GPU versions of the
propagators could therefore enable speed-ups for users who have appropriate
GPUs. One of the benefits of basing Deepwave on PyTorch is the ease of using
GPUs. There is a PyTorch tutorial that explains how to create a CUDA extension.
Finite difference wave propagation involves a type of computation that GPUs (graphics processing units) are suitable for. Creating GPU versions of the propagators could therefore enable speed-ups for users who have appropriate GPUs. One of the benefits of basing Deepwave on PyTorch is the ease of using GPUs. There is a PyTorch tutorial that explains how to create a CUDA extension.