HPCN-Europe 1996: High-Performance Computing and Networking pp 66-73 | Cite as
A comparison of parallel paradigms applied to an industrial code: RADIOSS
Abstract
Crash simulation has always been one of the most CPU intensive application of automotive industry supercomputing centers. It is then of paramount importance both for program developers, hardware vendors and of course end users to tune these softwares. RADIOSS is one of the leading packages available worldwide on the crashworthiness analysis market, running on the best computing platforms available. During the last few years, parallel computers have brought both great promises and some headaches to industrial software developers. The three main parallel programming paradigms (Shared Memory,Message Passing and Data Parallel) have been successfully implemented within RADIOSS and run on various platforms. This paper summarizes the key points of the development of these versions and the performance level obtained.
Keywords
Shared Memory Message Passing Dynamic Schedule Single Instruction Multiple Data Crash SimulationPreview
Unable to display preview. Download preview PDF.
References
- [1]Linpack http://performance.netlib.org/performance/html/PDSreports.htmlGoogle Scholar
- [2]Top 500 list http://parallel.rz.uni-mannheim.de/top500/top500.htmlGoogle Scholar
- [3]G. Deghilage, Architectures et Programmations Paralleles, Addison Wesley 1992Google Scholar
- [4]Danny Hillis, The Connection Machine, MIT Press 1985Google Scholar
- [5]C. Farhat and H.D. Simon, “TOP/DOMDEC: A software tool for mesh partitioning and parallel processing”, Center for Space Structure and Controls, University of Colorado at Boulder, Report CU-CSSC-93-11, 1993.Google Scholar
- [6]Horst Simon “Partitioning of unstructured problems for parallel processing”, Computing Systems in Engineering, 2 (1991) pp 135–148.Google Scholar