heFFTe: Highly efficient fft for exascale

Alan Ayala, Stanimire Tomov, Azzam Haidar, Jack Dongarra

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

39 Scopus citations

Abstract

Exascale computing aspires to meet the increasing demands from large scientific applications. Software targeting exascale is typically designed for heterogeneous architectures; henceforth, it is not only important to develop well-designed software, but also make it aware of the hardware architecture and efficiently exploit its power. Currently, several and diverse applications, such as those part of the Exascale Computing Project (ECP) in the United States, rely on efficient computation of the Fast Fourier Transform (FFT). In this context, we present the design and implementation of heFFTe (Highly Efficient FFT for Exascale) library, which targets the upcoming exascale supercomputers. We provide highly (linearly) scalable GPU kernels that achieve more than 40× speedup with respect to local kernels from CPU state-of-the-art libraries, and over 2× speedup for the whole FFT computation. A communication model for parallel FFTs is also provided to analyze the bottleneck for large-scale problems. We show experiments obtained on Summit supercomputer at Oak Ridge National Laboratory, using up to 24,576 IBM Power9 cores and 6,144 NVIDIA V-100 GPUs.

Original languageEnglish
Title of host publicationComputational Science – ICCS 2020 - 20th International Conference, Proceedings
EditorsValeria V. Krzhizhanovskaya, Gábor Závodszky, Michael H. Lees, Peter M.A. Sloot, Peter M.A. Sloot, Peter M.A. Sloot, Jack J. Dongarra, Sérgio Brissos, João Teixeira
PublisherSpringer Science and Business Media Deutschland GmbH
Pages262-275
Number of pages14
ISBN (Print)9783030503703
DOIs
StatePublished - 2020
Event20th International Conference on Computational Science, ICCS 2020 - Amsterdam, Netherlands
Duration: Jun 3 2020Jun 5 2020

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume12137 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference20th International Conference on Computational Science, ICCS 2020
Country/TerritoryNetherlands
CityAmsterdam
Period06/3/2006/5/20

Keywords

  • Exascale
  • FFT
  • GPUs
  • Scalable algorithm

Fingerprint

Dive into the research topics of 'heFFTe: Highly efficient fft for exascale'. Together they form a unique fingerprint.

Cite this