Abstract
Experimentation aimed at determining the potential efficiency of multi-microprocessor designs of SIMD machines is reported. The experimentation is based on timing measurements made on the PASM system prototype at Purdue. The application used to measure and evaluate this phenomenon was bitonic sorting, which has feasible solutions in both SIMD and MIMD modes of computation, as well as in at least two hybrids of SlMD and MIMD modes. Bitonic sorting was coded in these four ways and experiments were performed that examine the tradeoffs among all of these modes. Also, a new PE mask generation scheme for multiple "of-the-shelf" microprocessor based SIMD systems is proposed, and its performance was measured.
Keywords
Parallel Processing, Performance Evaluation, PASM, Computer Architecture, Hybrid SIMD/MIMD Computing
Date of this Version
11-1-1988
Comments
This work supported by the National Science Foundation under Grant Numbers CCR- 8704826, CCR-8809600, by the NSF Software Engineering Research Center (SERC), and by the Air Force Office of Scientific Research under Grant Number F49620-86-K-0006.