CMU-CS-85-128

Computer Science Department
School of Computer Science, Carnegie Mellon University


CMU-CS-85-128

PIE - A Programming and Instrumentation Environment for Parallel Processing

CMU-CS-85-128

Zary Segall, Larry Rudolph

April 1985

The issues of the efficient development of performance efficient parallel programs is explored. The Programming and Instrumentation Environment for Parallel Processing PIE system's concepts, designs, and preliminary implementation results are presented. The key goal in PIE is semiautomatic generation of performance efficient parallel programs. In PIE, a system intensive rather than programmer intensive programming environment is promoted for supporting users with different experience in parallel programming. Three levels of such support are provided, namely the Modular Programming Metalanguage, the Program Constructor, and the Implementation Assistant. In order to facilitate the task of parallel programming, each component employs a set of new concepts and approaches to integrate functionality with performance concerns. Some of them are: programming for observability, semantic monitoring, relational parallel program representation, constraint driven abstract data types Frames , frame unfolding, programming with templates, and parallel implementation assistant. This paper presents the results of PIE 1, the first of a three phase project.

33 pages


Return to: SCS Technical Report Collection
School of Computer Science homepage

This page maintained by reports@cs.cmu.edu