Edinburgh Research Explorer

A study on the performance of reproducible computations

Research output: Chapter in Book/Report/Conference proceedingChapter (peer-reviewed)

Original languageEnglish
Title of host publicationHigh Performance Computing
Subtitle of host publicationISC High Performance 2019 International Workshops, Frankfurt/Main, Germany, June 16-20, 2019, Revised Selected Papers
EditorsMichele Weiland, Guido Juckeland, Sadaf Alam, Heike Jagode
PublisherSpringer
Pages441-451
Number of pages12
ISBN (Electronic)978-3-030-34356-9
ISBN (Print)978-3-030-34355-2
DOIs
Publication statusPublished - 3 Dec 2019
EventISC High Performance 2019 - Frankfurt, Germany
Duration: 16 Jun 201920 Jun 2019
https://www.isc-hpc.com/research-papers-2019.html

Publication series

NameLecture Notes in Computer Science
PublisherSpringer
Volume11887
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

ConferenceISC High Performance 2019
Abbreviated titleISC HPC 2019
CountryGermany
CityFrankfurt
Period16/06/1920/06/19
Internet address

Abstract

Parallel computations are intrinsically non-reproducible, due to a combined effect of non-deterministic parallel reductions and nonassociative floating point operations. Different strategies have been proposed in literature to alleviate this issue or eliminate it altogether, however at present there is no study on the performance impact of associative floating point operations on large scale applications. In this work, we implement associative operations using binned doubles in MiniFE, and perform various performance tests on Cirrus and Fulhame, two state-ofthe-art HPC systems.

    Research areas

  • Reproducibility, Binned doubles, Performance

Event

ISC High Performance 2019

16/06/1920/06/19

Frankfurt, Germany

Event: Conference

Download statistics

No data available

ID: 104502287