next up previous contents index
Next: 13.5.1 Degrees of Difficulty Up: 13 Data Parallel C Previous: 13.4 Optimizing Compilers by

13.5 ASPAR


ASPAR was developed by Ikudome from CP [Ikudome:90a] in collaboration with ParaSoft. It was aimed at aiding the conversion of existing Fortran codes and embodies the experience especially of Flower and Kolawa. ASPAR is aimed at those applications involving particular stencil operation on arrays-noting that many sequential stencils need modification for parallel execution. In this way, ASPAR involves a collaboration between user and compiler in the parallelization process. The discussion in this section is due to Flower and Kolawa, and we include some of the introductory material as a contrast to the discussion given in the introductory sections of each chapter in this book, which largely reflect Fox's prejudice.

It is now a widely accepted fact that parallel computing is a successful technology. It has been applied to problems in many fields and has achieved excellent results on projects ranging in scope from academic demonstrations to complete commercial applications, as shown by other sections of this book.

Despite this success, however, parallel computing is still considered something of a ``black art'' to be undertaken only by those with intimate knowledge of hardware, software, physics, computer science and a wealth of other complex areas. To the uninitiated there is something frightening about the strange incantations that abound in parallel processing circles-not just the ``buzz words'' that come up in polite conversation but the complex operations carried out on a once elegant piece of sequential code in order for it to successfully run on a parallel processing system.

Guy Robinson
Wed Mar 1 10:19:35 EST 1995