Productivity, Portability, Performance: Data-Centric Python
Abstract
Python has become the de facto language for scientific computing. Programming in Python is highly productive, mainly due to its rich science-oriented software ecosystem built around the NumPy module. As a result, the demand for Python support in High Performance Computing (HPC) has skyrocketed. However, the Python language itself does not necessarily offer high performance. In this work, we present a workflow that retains Python's high productivity while achieving portable performance across different architectures. The workflow's key features are HPC-oriented language extensions and a set of automatic optimizations powered by a data-centric intermediate representation. We show performance results and scaling across CPU, GPU, FPGA, and the Piz Daint supercomputer (up to 23,328 cores), with 2.47x and 3.75x speedups over previous-best solutions, first-ever Xilinx and Intel FPGA results of annotated Python, and up to 93.16% scaling efficiency on 512 nodes.
Cite
@article{arxiv.2107.00555,
title = {Productivity, Portability, Performance: Data-Centric Python},
author = {Alexandros Nikolaos Ziogas and Timo Schneider and Tal Ben-Nun and Alexandru Calotoiu and Tiziano De Matteis and Johannes de Fine Licht and Luca Lavarini and Torsten Hoefler},
journal= {arXiv preprint arXiv:2107.00555},
year = {2021}
}