Distributed FFTs on GPU #30
Description
I want to do distributed FFTs on GPUs in Python. I am currently working on porting the mpi4py-fft library to GPUs by using CuPy.
Sanil Rao brought the FFTX library to my attention in his talk at SIAM PP24. He advertised several features, including multi-GPU. As far as I can tell, this Python version, however, is only single GPU.
Do you plan on adding support for multi-GPU in the Python version? I ask mostly out of curiosity. For my personal purposes, mpi4py-fft+CuPy is pretty much sufficient. But I would like to compare to FFTX if possible.