Optimizing Timing and Code Size Using Maximum Direct Loop Fusion

Meilin Liu, Chun Xue, Meikang Qiu, Edwin H.M. Sha

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

1 Scopus citations

Abstract

In this paper, we develop the technique that combines loop distribution with maximum direct loop fusion (LD_MDF), which performs maximum loop distribution, followed by maximum direct loop fusion to optimize timing and code size simultaneously. We illustrate using examples that dependence cycle is not always a restriction for loop distribution for multi-level loops. We map the maximum direct loop fusion problem to the graph partitioning problem. A polynomial graph partitioning algorithm is developed to compute the fusion partitions. We prove that our maximum direct loop fusion algorithm produces the fewest number of resultant loop nests without violating dependence constraints. We also show that the resultant code size of the fused loops by the technique of loop distribution with maximum direct loop fusion is always smaller than the code size of the original loops.

Original languageEnglish
Title of host publication19th International Conference on Parallel and Distributed Computing Systems 2006, PDCS 2006
PublisherInternational Society for Computers and Their Applications (ISCA)
Pages38-43
Number of pages6
ISBN (Electronic)9781604236446
StatePublished - 2006
Externally publishedYes
Event19th International Conference on Parallel and Distributed Computing Systems, PDCS 2006 - San Francisco, United States
Duration: 20 Sep 200622 Sep 2006

Publication series

Name19th International Conference on Parallel and Distributed Computing Systems 2006, PDCS 2006

Conference

Conference19th International Conference on Parallel and Distributed Computing Systems, PDCS 2006
Country/TerritoryUnited States
CitySan Francisco
Period20/09/0622/09/06

Keywords

  • Code Size
  • Embedded DSP
  • Loop Distribution
  • Loop Fusion

Fingerprint

Dive into the research topics of 'Optimizing Timing and Code Size Using Maximum Direct Loop Fusion'. Together they form a unique fingerprint.

Cite this