IDEALS Home University of Illinois at Urbana-Champaign logo The Alma Mater The Main Quad

The dependence analysis and synchronization of parallel programs

Show full item record

Bookmark or cite this item:

Files in this item

File Description Format
PDF 9216669.pdf (6MB) Restricted to U of Illinois (no description provided) PDF
Title: The dependence analysis and synchronization of parallel programs
Author(s): Midkiff, Samuel Pratt
Doctoral Committee Chair(s): Padua, David
Department / Program: Computer Science
Discipline: Computer Science
Degree Granting Institution: University of Illinois at Urbana-Champaign
Degree: Ph.D.
Genre: Dissertation
Subject(s): Computer Science
Abstract: The optimization of programs with explicit--i.e. user specified--parallelism requires the computation of the data dependence relation if optimizations performed by the compiler are to preserve sequential consistency. Shasha and Snir investigated the problem for straight-line code with data accesses performed on scalars. We extend their results to programs with sequential loops and cobegin or doall loop parallelism and data accesses performed on arrays. The array subscript information is used to increase the accuracy of the test in a way analogous to the data dependence tests for sequential programs.Once data dependences have been identified synchronization can be used to enforce the orderings specified by the data dependence. The automatic generation of synchronization, and the optimization of synchronization, by a compiler is investigated. Experimental results comparing the effectiveness of different optimization techniques are presented.
Issue Date: 1992
Type: Text
Language: English
Rights Information: Copyright 1992 Midkiff, Samuel Pratt
Date Available in IDEALS: 2011-05-07
Identifier in Online Catalog: AAI9216669
OCLC Identifier: (UMI)AAI9216669

This item appears in the following Collection(s)

Show full item record

Item Statistics

  • Total Downloads: 2
  • Downloads this Month: 0
  • Downloads Today: 0


My Account


Access Key