Array organization in parallel memories

Mayez Al-Mouhamed*

*Corresponding author for this work

Research output: Contribution to journalReview articlepeer-review

Abstract

A memory and compiler co-optimization aimed at reducing low-level memory accesses using software and hardware locality optimizations is discussed. The analysis of conflict-free access to parallel memories and alignment networks using algebra of non-singular Boolean matrices is also discussed. An effective compiler heuristic is proposed for finding a storage matrix that minimizes overall memory access time. The results show that a memory utilization above 83% is achievable by the storage scheme for 64 memories.

Original languageEnglish
Pages (from-to)123-163
Number of pages41
JournalInternational Journal of Parallel Programming
Volume32
Issue number2
DOIs
StatePublished - Apr 2004

Keywords

  • Access patterns
  • Compiler optimization
  • Embedded systems
  • Parallel memory
  • Streamed computations

ASJC Scopus subject areas

  • Software
  • Theoretical Computer Science
  • Information Systems

Fingerprint

Dive into the research topics of 'Array organization in parallel memories'. Together they form a unique fingerprint.

Cite this