Skip to main content
U.S. Department of Energy
Office of Scientific and Technical Information

A performance model for massive parallelism

Technical Report ·
OSTI ID:6964033

A popular argument is that vector and parallel architectures should not be carried to extremes because the scalar or serial portion of the code will eventually dominate. Since pipeline stages and extra processors obviously add hardware cost, a corollary to this argument is that the most cost-effective computer is one based on uniprocessor, scalar principles. For architectures that are both parallel and vector, the argument is compounded, making it appear that near-optimal performance on such architectures is a near-impossibility. A new argument is presented that is based on the assumption that program execution time, not problem size, is constant for various amounts of vectorization and parallelism. This has a dramatic effect on Amdahl Law, revealing that one can be much more optimistic about achieving high speedups on massively-parallel and highly-vectorized machines. The revised argument is supported by recent results of over 1000 times speedup on 1024 processors on several practical scientific applications. This document discusses the results and presents a reevaluation of Amdahl's Law.

Research Organization:
Sandia National Labs., Albuquerque, NM (USA)
DOE Contract Number:
AC04-76DP00789
OSTI ID:
6964033
Report Number(s):
SAND-88-1281; ON: DE88011763
Country of Publication:
United States
Language:
English