The performance realities of massively parallel processors: A case study

Conference · Wed Jul 01 04:00:00 EDT 1992

DOI:https://doi.org/10.1109/SUPERC.1992.236663· OSTI ID:10149504

Lubeck, O M; Simmons, M L; Wasserman, H J

This paper presents the results of an architectural comparison of SIMD massive parallelism, as implemented in the Thinking Machines Corp. CM-2 computer, and vector or concurrent-vector processing, as implemented in the Cray Research Inc. Y-MP/8. The comparison is based primarily upon three application codes that represent Los Alamos production computing. Tests were run by porting optimized CM Fortran codes to the Y-MP, so that the same level of optimization was obtained on both machines. The results for fully-configured systems, using measured data rather than scaled data from smaller configurations, show that the Y-MP/8 is faster than the 64k CM-2 for all three codes. A simple model that accounts for the relative characteristic computational speeds of the two machines, and reduction in overall CM-2 performance due to communication or SIMD conditional execution, is included. The model predicts the performance of two codes well, but fails for the third code, because the proportion of communications in this code is very high. Other factors, such as memory bandwidth and compiler effects, are also discussed. Finally, the paper attempts to show the equivalence of the CM-2 and Y-MP programming models, and also comments on selected future massively parallel processor designs.

• Research Organization: Los Alamos National Lab., NM (United States)
• Sponsoring Organization: USDOE, Washington, DC (United States)
• DOE Contract Number: W-7405-ENG-36
• OSTI ID: 10149504
• Report Number(s): LA-UR--92-1463; CONF-921125--4; ON: DE92015211
• Country of Publication: United States
• Language: English

Related Subjects

99 GENERAL AND MISCELLANEOUS
990200
ARRAY PROCESSORS
BENCHMARKS
COMPARATIVE EVALUATIONS
COMPUTER ARCHITECTURE
MATHEMATICS AND COMPUTERS
PARALLEL PROCESSING
PERFORMANCE
SUPERCOMPUTERS