AN INTEGRATED PERFORMANCE ANALYSIS TOOL FOR SPMD DATA-PARALLEL PROGRAMS

A new generation of data parallel languages have been proposed whereby a user specifies how data structures are to be distributed amongst th e processor nodes of a distributed-memory machine. Based on this infor mation, the compiler then generates code for the parallel application. Although this approach significantly simplifies the development of th e initial version of a parallel application, selection of good data di stributions leading to efficient computations is often quite difficult . Therefore, performance debuggers are needed to yield insights into t he data distribution effects. On the other hand, most of the existing approaches to performance debugging are very general and thus do not p rovide the user feedback in terms of the high level programming model or the source code of the parallel application. In this paper, we desc ribe a novel approach to performance debugging of data parallel progra ms. The design and implementation of a visual performance debugger is described that is specifically targeted to meet the performance debugg ing requirements of a data-parallel programming model based on user-sp ecified data distributions. The performance debugger is part of an int egrated programming environment, called EPPP, which also supports a da ta parallel compiler and a parallel architecture simulator. Thus devel opment and performance debugging of an application may be done either on the real hardware or by using the simulator. The code for EPPP can be obtained free of cost (contact web address, http://www.crim.ca/apar ).