ZPL vs HPF: A Comparison of Performance and Programming Style

Calvin Lin
Lawrence Snyder
Ruth E. Anderson
Bradford L. Chamberlain
Sung-Eun Choi
George Forman
E Christopher Lewis
W. Derrick Weathersby

University of Washington Technical Report UW-CSE-95-11-05

Abstract: This paper compares two data parallel languages, ZPL and HPF, in terms of programming style and performance. The results show that for eight programs from a number of standard benchmark suites, ZPL generally outperforms HPF, and ZPL expresses problems at higher levels of abstraction, yielding programs that are shorter, less error prone and easier to maintain. ZPL's better performance comes from its clean expression of parallelism that allows for better compiler analysis.

postscript | PDF