• https://doi.org/10.1007/s10766-017-0499-9Copy DOI Icon

Software Speculation on Caching DSMs

Show More
  • Abstract
  • Literature Map
  • References
  • Similar Papers
Abstract

Clusters with caching DSMs deliver programmability and performance by supporting shared-memory programming model and tolerating communication latency of remote fetches via caching. The input of a data parallel program is partitioned across machines in the cluster while the DSM transparently fetches and caches remote data as needed by the application. Irregular applications are challenging to parallelize because the input related data dependences that manifest at runtime require the use of speculation for efficiently exploiting parallelism. By speculating that there are no cross iteration dependences, multiple iterations of a data parallel loop are executed in parallel using locally cached copies of data; the absence of dependences is validated before committing the speculatively computed results. In this paper we show that in irregular data-parallel applications, while caching helps tolerate long communication latencies, using a value read from the cache in a computation can lead to misspeculation, and thus aggressive caching can degrade performance due to increased misspeculation rate. To limit misspeculation rate we present optimizations for distributed speculation on caching based DSMs that decrease the cost of misspeculation check and speed up the re-execution of misspeculated recomputations. These optimizations give speedups of 2.24\({\times }\) for graph coloring, 1.71\({\times }\) for connected components, 1.88\({\times }\) for community detection, 1.32\({\times }\) for shortest path, and 1.74\({\times }\) for pagerank over baseline parallel executions.

Similar Papers
  • Conference Article
  • Citations1

Optimizing Caching DSM for Distributed Software Speculation

  • Sep 01, 2015
  • Sai Charan Koduru +2
  • Research Article
  • Citations86

Optimistic parallelism requires abstractions

  • Jun 10, 2007
  • ACM SIGPLAN Notices
  • Milind Kulkarni +5
  • Conference Article
  • Citations460

Optimistic parallelism requires abstractions

  • Jun 10, 2007
  • Milind Kulkarni +5
  • Conference Article
  • Citations70

Scheduling strategies for optimistic parallel execution of irregular programs

  • Jun 14, 2008
  • Milind Kulkarni +6
  • Conference Article
  • Citations5

Integrating task and data parallelism in an irregular application: a case study

  • Oct 23, 1996
  • K Mcpherson +1
  • Research Article
  • Citations1

Two-Level Task Scheduling for Irregular Applications on GPU Platform

  • Nov 04, 2015
  • International Journal of Parallel Programming
  • Jing Li +4
  • Research Article
  • Citations15

Optimistic parallelism benefits from data partitioning

  • Mar 01, 2008
  • ACM SIGARCH Computer Architecture News
  • Milind Kulkarni +5
  • Conference Article
  • Citations81

Optimistic parallelism benefits from data partitioning

  • Mar 01, 2008
  • Milind Kulkarni +5
  • Research Article
  • Citations6

Optimistic parallelism benefits from data partitioning

  • Mar 01, 2008
  • ACM SIGOPS Operating Systems Review
  • Milind Kulkarni +5
  • Research Article
  • Citations1

Optimistic parallelism benefits from data partitioning

  • Mar 01, 2008
  • ACM SIGPLAN Notices
  • Milind Kulkarni +5
  • Conference Article
  • Citations2

Maximizing the performance of irregular applications on multithreaded, NUMA systems

  • Nov 17, 2013
  • Guojing Cong +1
  • Conference Article
  • Citations2

Mapping applications for high performance on multithreaded, NUMA systems

  • May 14, 2013
  • Guojing Cong +1
  • Conference Article
  • Citations5

A data parallel programming model based on distributed objects

  • Jun 26, 2003
  • R Diaconescu +1
  • Conference Article
  • Citations2

Speculation with Little Wasting: Saving Cost in Software Speculation through Transparent Learning

  • Jan 01, 2009
  • Yunlian Jiang +2
  • Conference Article
  • Citations26

Scaling irregular parallel codes with minimal programming effort

  • Nov 10, 2001
  • Dimitrios S Nikolopoulos +2
Cactus Communications logo

Copyright 2026 Cactus Communications. All rights reserved.