Python for High Performance Computing

Description

Python for High Performance Computing

Presented by William Scullin

Python is becoming increasingly popular within the high performance computing community. While it initially gained traction as a scripting language, Python's role has continued to expand with Python applications for science scaling to hundreds of thousands of cores and bindings to high performance libraries becoming commonplace. This talk is meant as an overview of Python's role in the HPC space.

Abstract

This talk is focused on raising awareness of Python in the high performance computing space. Specific topics include:

  • building the Python interpreter for speed
  • an overview of bindings to numerical libraries
  • using GPUs and accelerators with Python
  • scaling codes with MPI
  • issues when scaling on very large systems
  • an overview of successful science codes
  • a live demonstration of Python running on 163,840 cores