Skip Navigation

The Computer Journal 1997 40(6):373-387; doi:10.1093/comjnl/40.6.373
© 1997 by British Computer Society
This Article
Right arrow Full Text (PDF)
Right arrow Alert me when this article is cited
Right arrow Alert me if a correction is posted
Services
Right arrow Email this article to a friend
Right arrow Similar articles in this journal
Right arrow Similar articles in ISI Web of Science
Right arrow Alert me to new issues of the journal
Right arrow Add to My Personal Archive
Right arrow Download to citation manager
Right arrowRequest Permissions
Google Scholar
Right arrow Articles by Kulkarni, D.
Right arrow Articles by Stumm, M.
Right arrow Search for Related Content
Social Bookmarking
Add to CiteULike   Add to Connotea   Add to Del.icio.us  
What's this?

Linear and Extended Linear Transformations for Shared-Memory Multiprocessors

D. Kulkarni1 and M. Stumm2

1 Parallel Compiler Development, IBM Toronto Laboratory, Toronto, Canada M3C 1H7 Email: dkulki{at}vnet.ibm.com, 2 Department of Electrical and Computer Engineering, University of Toronto, Toronto, Canada M5S 3G4

Advances in program transformation frameworks have significantly advanced compiler technology over the past few years. Program transformation frameworks provide mathematical abstractions of loop and data structures and formal methods for manipulating these structures. It is these frameworks that have allowed the development of algorithms capable of automatically tailoring an application for a target architecture. In this paper, we focus on the utility of these frameworks in improving the performance of mainly parallel applications on shared-memory multiprocessors. Data locality-oriented program optimizations are a key to good performance on shared-memory multiprocessors, since these optimizations can often improve performance by a factor of 10 or more. In this paper, we show the effectiveness of three key loop and data transformation frameworks in optimizing parallel programs on shared-memory multiprocessors. In particular, we describe our computation decomposition and alignment (CDA) framework, which can modify both the composition and the execution order of the re-composed iterations. We show how fine-grain transformations within the CDA framework enable new optimizations such as local optimizations that are otherwise achieved by global data transformations.

Received October, 1996, 1997. revised September, 1997.


Add to CiteULike CiteULike   Add to Connotea Connotea   Add to Del.icio.us Del.icio.us    What's this?




Disclaimer:
Please note that abstracts for content published before 1996 were created through digital scanning and may therefore not exactly replicate the text of the original print issues. All efforts have been made to ensure accuracy, but the Publisher will not be held responsible for any remaining inaccuracies. If you require any further clarification, please contact our Customer Services Department.