David J. McNamara (presenter),
Brian Q. Brode, James J. Bonang, and Chris R. Warber
Pacific-Sierra Research Corp., 2901 28th Street, Santa Monica,
CA 90405, USA.
The use of the DEEP development environment for HPF programs is described. The full integrated environment contains tools for the analysis, profiling, and debugging of Fortran and C parallel programs for distributed and shared memory parallel systems. The tools present compile-time optimization (parallelization) feedback, run-time profiling information, and debugging information at the parallel source code level.
The first part of the paper describes the program analysis and profiling portion of DEEP, with examples of its use on HPF programs. With DEEP, the HPF programmer can better understand the performance and behavior of an HPF program, and structure the program for greater efficiency. The second part of the paper presents the parallel debugging capabilities of DEEP for distributed memory systems. With DEEP, the HPF program can be debugged with a GUI interface as a single program at the HPF source code level; distributed arrays can be visualized as a single item.