Array form representation of idiom recognition system for numerical programs

  • Authors:
  • Sato Hiroyuki

  • Affiliations:
  • The University of Tokyo, 2-11-16, Yayoi, Bunkyo-ku, Tokyo, 113-8658, Japan

  • Venue:
  • Proceedings of the 2001 conference on APL: an arrays odyssey
  • Year:
  • 2000

Quantified Score

Hi-index 0.00

Visualization

Abstract

Today, PSE emerges as a promising approach for providing programmers with high-level abstract functions Particularly for numerical programs which essentially handle large arrays, abstract representation of array-handling functions is very important. Idiom recognition can be a bridge between abstract source programs and concrete architectures. It can remove obstacles of gaps between the two that cannot be bridged by today's state of the art compilers. A major problem in idiom recognition is intermediate representation. This paper applies term rewriting theory---a very general framework --- to an idiom recognition system. Specifically, we apply it to the idiom recognition problem of Fortran90/95 matrix manipulations, and show that the Fortran90/95 system is never a trivial system. Straight forms are introduced to make the idiom recognition system more powerful. We make a completion procedure based on this Fortran90/95 system, and show some non-trivial idiom recognition using our theory.