logo
科言猫学术研究的AI总结
language
手机端手机端
  • Phone Feature微信小程序
简体中文简体中文
notice消息通知
博客博客
delete
arrow
返回

Scalable Block-Sparse Matrix Multiplication Using Template Task Graphs

delete2026-01-01
delete0
PRE
AI
J
Joseph Schuchart *
A
Aurélien Bouteiller
T
Thomas Hérault
V
Valeev, Edvard
G
George Bosilca
R
Robert J. Harrison
DOI：10.1007/978-3-031-97196-9_10delete
delete原文链接
delete原文求助
delete分享
delete收藏
摘要

摘要

En 中文
块稀疏矩阵运算是一般稀疏代数的一种特殊情况，其中矩阵由密集块稀疏分布组成，例如在量子化学的稀疏张量代数中。在实现分布式矩阵乘法 C = A × B 的一般情况下，一个挑战是通信流的管理，因为两个输入矩阵 A 和 B 都易于获取，必须分布到计算 C 相关块的进程上。在本文中，我们提出了对模板任务图编程模型的扩展，允许应用程序通过灵活的 API 约束任务的执行。我们表明，此类约束可以在纯数据流模型中使用，以用更结构化的方法替代人工控制流。在稀疏矩阵乘法的背景下，我们发现约束允许我们限制并发通信的数量，从而避免在网络上创建瓶颈。
Keyword:
Template Task Graph
Sparse Matrix Multiplication
Scheduling Constraints

期刊

A
ASYNCHRONOUS MANY-TASK SYSTEMS AND APPLICATIONS, WAMTA 2025
IF:
0
论文数:
13
被引数:
0

机构

S
stony brook university
学者数:
1.4W
论文数: 1.0W
被引数: 20
S
state university of new york (suny) system
学者数:
6.5W
论文数: 5.8W
被引数: 65
University of Tennessee System 封面图
University of Tennessee System
学者数:
2.9W
论文数: 2.7W
被引数: 115
学者 查看更多机构
引用论文

引用论文

Distributed-memory multi-GPU block-sparse tensor contraction for electronic structure
err2021 IEEE International Parallel and Distributed Processing Symposium (IPDPS)
IF0
err2021-05-01
err0
errOAAI
errThomas Herault; Yves Robert; George Bosilca; Robert J. Harrison; Cannada A. Lewis; Edward F. Valeev; Jack J. Dongarra
err分享
err收藏
err
IF0
err
err0
PREAI
err
err分享
err收藏
StarPU: a unified platform for task scheduling on heterogeneous multicore architectures
errConcurrency and Computation: Practice and Experience
IF0
err2010-11-04
err0
errOAAI
errCédric Augonnet; Samuel Thibault; Raymond Namyst; Pierre‐André Wacrenier
err分享
err收藏
Dynamic task discovery in PaRSEC
errProceedings of the 8th Workshop on Latest Advances in Scalable Algorithms for Large-Scale Systems
IF0
err2017-11-12
err0
PREAI
errReazul Hoque; Thomas Herault; George Bosilca; Jack Dongarra
err分享
err收藏
Dagger: combining benefits of synchronous and asynchronous communication stylesDagger：结合同步和异步通信模式的优点
errProceedings of 8th International Parallel Processing Symposium
IF0
err2024-10-23
err0
PREAI
errA. Gursoy; L.V. Kale
err分享
err收藏
Legion: Expressing locality and independence with logical regions
err2012 International Conference for High Performance Computing, Networking, Storage and Analysis
IF0
err2012-11-01
err0
errOAAI
errMichael Bauer; Sean Treichler; Elliott Slaughter; Alex Aiken
err分享
err收藏
The Foundations for Scalable Multicore Software in Intel Threading Building Blocks
errIntel Technology Journal
IF0
err2007-11-16
err0
PREAI
errAlexey Kukanov
err分享
err收藏
HPX
errProceedings of the 8th International Conference on Partitioned Global Address Space Programming Models
IF0
err2014-10-06
err0
PREAI
errHartmut Kaiser; Thomas Heller; Bryce Adelstein-Lelbach; Adrian Serio; Dietmar Fey
err分享
err收藏
SUMMA: scalable universal matrix multiplication algorithm
errConcurrency: Practice and Experience
IF0
err1997-04-01
err0
PREAI
errR. A. Van De Geijn; J. Watts
err分享
err收藏
PTG: An Abstraction for Unhindered Parallelism
err2014 Fourth International Workshop on Domain-Specific Languages and High-Level Frameworks for High Performance Computing
IF0
err2014-11-01
err0
PREAI
errAnthony Danalis; George Bosilca; Aurelien Bouteiller; Thomas Herault; Jack Dongarra
err分享
err收藏
学者 查看更多内容