A formal framework for orthogonal data and control parallelism handling

  • Authors:
  • Sonia Campa

  • Affiliations:
  • Dept. Computer Science, University of Pisa, PISA, Italy

  • Venue:
  • ICCS'05 Proceedings of the 5th international conference on Computational Science - Volume Part II
  • Year:
  • 2005

Quantified Score

Hi-index 0.00

Visualization

Abstract

We propose a semantic framework for parallel programming based on the orthogonalization of data access and control concerns by means of set of abstraction mechanisms. Such mechanisms regard the description of how data has to be accessed, the description of how data has to be computed and the description of how coupling data accesses and patterns of control. Each description is represented by an abstraction mechanism formalized through a formal semantics. The set of semantics specifications defines a method to investigate the structure of the whole application. We demonstrate how this semantics provides a formal, provable method to statically or dynamically evaluate the overall performance of the application and, eventually, apply optimization rules.