Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes
Pure is a new programming model and runtime system explicitly designed to take advantage of shared memory within nodes in the context of a mostly message passing interface enhanced with the ability to use tasks to make use of idle cores. Pure leverages shared memory in two ways: (a) by allowing core...
Main Authors: | , |
---|---|
Format: | Article |
Language: | English |
Published: |
ACM
2024
|
Online Access: | https://hdl.handle.net/1721.1/153627 |
_version_ | 1811079825242193920 |
---|---|
author | Psota, James Solar-Lezama, Armando |
author_facet | Psota, James Solar-Lezama, Armando |
author_sort | Psota, James |
collection | MIT |
description | Pure is a new programming model and runtime system explicitly designed to take advantage of shared memory within nodes in the context of a mostly message passing interface enhanced with the ability to use tasks to make use of idle cores. Pure leverages shared memory in two ways: (a) by allowing cores to steal work from each other while waiting on messages to arrive, and, (b) by leveraging efficient lock-free data structures in shared memory to achieve high-performance messaging and collective operations between the ranks within nodes. We use microbenchmarks to evaluate Pure's key messaging and collective features and also show application speedups up to 2.1× on the CoMD molecular dynamics and the miniAMR adaptive mesh refinement applications scaling up to 4,096 cores. |
first_indexed | 2024-09-23T11:21:04Z |
format | Article |
id | mit-1721.1/153627 |
institution | Massachusetts Institute of Technology |
language | English |
last_indexed | 2024-09-23T11:21:04Z |
publishDate | 2024 |
publisher | ACM |
record_format | dspace |
spelling | mit-1721.1/1536272024-03-02T03:44:39Z Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes Psota, James Solar-Lezama, Armando Pure is a new programming model and runtime system explicitly designed to take advantage of shared memory within nodes in the context of a mostly message passing interface enhanced with the ability to use tasks to make use of idle cores. Pure leverages shared memory in two ways: (a) by allowing cores to steal work from each other while waiting on messages to arrive, and, (b) by leveraging efficient lock-free data structures in shared memory to achieve high-performance messaging and collective operations between the ranks within nodes. We use microbenchmarks to evaluate Pure's key messaging and collective features and also show application speedups up to 2.1× on the CoMD molecular dynamics and the miniAMR adaptive mesh refinement applications scaling up to 4,096 cores. 2024-03-01T16:41:58Z 2024-03-01T16:41:58Z 2024-02-20 2024-03-01T08:46:32Z Article http://purl.org/eprint/type/ConferencePaper 979-8-4007-0435-2 https://hdl.handle.net/1721.1/153627 Psota, James and Solar-Lezama, Armando. 2024. "Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes." PUBLISHER_POLICY en 10.1145/3627535.3638503 Article is made available in accordance with the publisher's policy and may be subject to US copyright law. Please refer to the publisher's site for terms of use. The author(s) application/pdf ACM Association for Computing Machinery |
spellingShingle | Psota, James Solar-Lezama, Armando Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title | Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title_full | Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title_fullStr | Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title_full_unstemmed | Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title_short | Pure: Evolving Message Passing To Better Leverage Shared Memory Within Nodes |
title_sort | pure evolving message passing to better leverage shared memory within nodes |
url | https://hdl.handle.net/1721.1/153627 |
work_keys_str_mv | AT psotajames pureevolvingmessagepassingtobetterleveragesharedmemorywithinnodes AT solarlezamaarmando pureevolvingmessagepassingtobetterleveragesharedmemorywithinnodes |