Performance Visualization of SLALOM

Performance visua1,ization provides insights about the complex operation of concurrent computer systems. SLAL O W M is a scalable, fuced-time coinputer benchmark. Each corresponds to U method of computer performance evaluation: monitoring and benchmarking, respectively. Whereas benchmark programs typically report singlenumber performance naetrics for ease of comparison among different machines, a perforfinance monitor (via instrumentation and visualization) gives (a detailed account of the dynamks of program execution. Using sofrware tools developed for the nCCBE 2 and the MasPar MP-1 distributed memory machines and applied to the SLALOM program, we demonstrate the utility of performance visualization for fine-tuning algorithms and understanding phenomena. The tools include PICL and ParaGraph and custom VISTA components.

[1]  G. A. Geist,et al.  A machine-independent communication library , 1989 .

[2]  Michael T. Heath,et al.  Visualizing the performance of parallel programs , 1991, IEEE Software.

[3]  Alva L. Couch,et al.  Monitoring ParalIel Executions in Real Time , 1990, Proceedings of the Fifth Distributed Memory Computing Conference, 1990..

[4]  Gurpur M. Prabhu,et al.  Visualization of Program Performance on Concurrent Computers , 1989, Great Lakes Computer Science Conference.

[5]  John Gustafson,et al.  The Design of a Scalable, Fixed-Time Computer Benchmark , 1991, J. Parallel Distributed Comput..

[6]  Daniel A. Reed,et al.  Scalable Performance Environments for Parallel Systems , 1991, The Sixth Distributed Memory Computing Conference, 1991. Proceedings.