Results from Measuring the Performance of the NAS Benchmarks on the Current Generation of Parallel Computers and Observations Drawn from these Measurements

The NAS Division at NASA Ames Research Center has developed a highly respected and widely used set of benchmarks for parallel computers. These benchmarks are based on the needs of computational fluid dynamics applications, but appear to have relevance to other disciplines as well. Unfortunately, NAS...

Full description

Saved in:
Bibliographic Details
Main Author: Pressel, Daniel M
Format: Report
Language:English
Subjects:
COMPONENT REPORTS
COMPUTATIONAL FLUID DYNAMICS
COMPUTER BENCHMARKING
Computer Systems Management and Standards
MEASUREMENT
NAS(NASA ADVANCED SUPERCOMPUTING)
PARALLEL PROCESSORS
PERFORMANCE(ENGINEERING)
RESEARCH FACILITIES
SUPERCOMPUTERS
SYMPOSIA
Online Access:Request full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:The NAS Division at NASA Ames Research Center has developed a highly respected and widely used set of benchmarks for parallel computers. These benchmarks are based on the needs of computational fluid dynamics applications, but appear to have relevance to other disciplines as well. Unfortunately, NAS last published a collection of results for a wide range of systems in November of 1997. Given the rapid level of change in the field of computers in general, and high performance parallel computing, it seemed appropriate to try and fill in some of the gaps. As such, the major computational assets of the ARL-MSRC is being benchmarked, with the intent to publish the results on an ARL website. While making these measurements, several observations and conclusions were drawn. These observations and conclusions will be discussed at length in this document. Presented at the DoD High Performance Computing Modernization Program (HPCMP) Users Group Conference (UGC), held in Williamsburg, Virginia on 7-11 June 2004. The original document contains color images. All DTIC reproductions will be in black and white. This article is from ADA492363 Proceedings of the HPCMP Users Group Conference 2004.