Another View on Parallel Speedup

X.-H. Sun; L. Ni

doi:10.1109/SUPERC.1990.130037

Authors: X.-H. Sun, L. Ni
Venue: Proceedings of the 1990 ACM/IEEE Conference on Supercomputing (SC'90)
Date: 1990
Type: Conference
DOI: 10.1109/SUPERC.1990.130037 (opens in a new tab)

Abstract

This paper presents an alternative view on the speedup of parallel processing. Three models of parallel speedup are studied: fixed-size speedup, fixed-time speedup, and a new memory-bounded speedup model. The memory-bounded model considers the relationship between problem size scaling and available memory, providing a more realistic measure of parallel system performance than traditional fixed-size speedup.