[Previous [1] [2] [3] [4] [5] [6] [7] [8] [9] [10] [11] [12] [13] [14]

Journal of Inforamtion Science and Engineering, Vol.14 No.1, pp.191-203 (March 1998)
Algorithmic Concept Recognition Support for
Automatic Parallelization: A Case Study on Loop
Optimization and Parallelization*

Beniamino Di Martino
Dipartimento di Informaticae Sistemistica
University "Federico II" - Naples - Italy
and Institute for Software Technology and Parallel Systems
University of Vienna - Austria

Automated algorithmic concept recognition within sequential code can support compilation techniques for program parallelization by allowing the introduction of heuristics and extensive pruning of the search space associated with the code transformation selections, thus enabling application of more aggressive transformations.

        This paper shows, through a case study, how automatic recognition of algorithmic patterns can enable automatic selection of suitable sequences of loop transformations for the implementing code, selection of suitable data and work distributions, and provision for communication optimizations.

Keywords: program analysis, automated program understanding, algorithmic pattern recognition, knowledge based program transformation and optimization, automated code replacement with optimized libraries

Received April 25, 1997; revised November 14, 1997.
Communicated by Chung-Ta King.
*This work has been partly supported by a Marie Curie grant within the European Commission TMR (Training and Mobility of Researchers) Programme.