Collective affinity aware computation mapping

Mahmut T. Kandemir, Jihyun Ryoo, Hui Zhao, Myoungsoo Jung, Mustafa Karakoy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This work defines the concept of collective affinity. It is claimedthat collective affinity has more potential than single core-centricaffinity, for data locality optimization in manycores. The reason isthat collective affinity captures the potential benefits of transferringcomputations originally assigned to one core to other cores. Next,building upon the collective affinity concept and a cache contentestimation strategy, it presents a computation-to-core mappingstrategy, specifically tuned for exploiting near data computing byreducing distance-to-data.

Original languageEnglish
Title of host publicationPACT 2020 - Proceedings of the ACM International Conference on Parallel Architectures and Compilation Techniques
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages343-344
Number of pages2
ISBN (Electronic)9781450380751
DOIs
Publication statusPublished - 2020 Sep 30
Event2020 ACM International Conference on Parallel Architectures and Compilation Techniques, PACT 2020 - Virtual, Online, United States
Duration: 2020 Oct 32020 Oct 7

Publication series

NameParallel Architectures and Compilation Techniques - Conference Proceedings, PACT
ISSN (Print)1089-795X

Conference

Conference2020 ACM International Conference on Parallel Architectures and Compilation Techniques, PACT 2020
CountryUnited States
CityVirtual, Online
Period20/10/320/10/7

Bibliographical note

Funding Information:
This material is based upon work supported by the National Science Foundation under grants 1763681, 1629129, 1931531, and 1629915.

All Science Journal Classification (ASJC) codes

  • Software
  • Theoretical Computer Science
  • Hardware and Architecture

Fingerprint Dive into the research topics of 'Collective affinity aware computation mapping'. Together they form a unique fingerprint.

Cite this