Analyzing suitability of contemporary 3D stacked PIM architectures for HPC scientific applications

Ivy B. Peng, Jeffrey S. Vetter, Shirley Moore, Rakshit Joydeep, Stefano Markidis

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

3 Scopus citations

Abstract

Scaling off-chip bandwidth is challenging due to fundamental limitations, such as a fixed pin count and plateauing signaling rates. Recently, vendors have turned to 2.5D and 3D stacking to closely integrate system components. Interestingly, these technologies can integrate a logic layer under multiple memory dies, enabling computing capability inside a memory stack. This trend in stacking is making PIM architectures commercially viable. In this work, we investigate the suitability of offloading kernels in scientific applications onto 3D stacked PIM architectures. We evaluate several hardware constraints resulted from the stacked structure. We perform extensive simulation experiments and in-depth analysis to quantify the impact of application locality in TLBs, data caches, and memory stacks. Our results also identify design optimization areas in software and hardware for HPC scientific applications.

Original languageEnglish
Title of host publicationACM International Conference on Computing Frontiers 2019, CF 2019 - Proceedings
PublisherAssociation for Computing Machinery, Inc
Pages256-262
Number of pages7
ISBN (Electronic)9781450366854
DOIs
StatePublished - Apr 30 2019
Event16th ACM International Conference on Computing Frontiers, CF 2019 - Alghero, Sardinia, Italy
Duration: Apr 30 2019May 2 2019

Publication series

NameACM International Conference on Computing Frontiers 2019, CF 2019 - Proceedings

Conference

Conference16th ACM International Conference on Computing Frontiers, CF 2019
Country/TerritoryItaly
CityAlghero, Sardinia
Period04/30/1905/2/19

Keywords

  • 3D Stacked Memory
  • PIM
  • Processing-In-Memory

Fingerprint

Dive into the research topics of 'Analyzing suitability of contemporary 3D stacked PIM architectures for HPC scientific applications'. Together they form a unique fingerprint.

Cite this