pyamgx – Accelerated Python Library

Summary

Interpreted languages typified by Python are very quickly becoming popular for science, and Python is already the most popular language for Deep Learning. A number of libraries and extensions have been written for Python, some of which are accelerated with the use of GPUs – but some of which are not.

Ashwin Srinath’s Python interface to Nvidia’s AmgX algebraic multigrid library that runs on GPUs has a wide range of applicability in fluid flow, physics, and astrophysics.

He applied the Python interfaces to the problem of fluid diffusion with the use of FiPy and showed how easy it was to use the AmgX library through the Python interface to get a 3.0x speedup.