Comparison between a message passing and a global shared memory parallel computer by means of a concrete Monte-Carlo algorithm
A program originally developed to be executed on a CRAY vector computer was parallelized for the message passing parallel computer nCUBE2 and subsequently ported to the KSR1, which is a global shared memory parallel computer. This offers the possibility to compare the two machine types from a user's point of view. Beyond this special implementation a more general strategy has been derived, which allows to port programs almost automatically from the nCUBE2 system to the KSR1 machine. This strategy might possibly be generalized to other message passing and global shared memory systems.
Key wordsparallelization portability applications
Unable to display preview. Download preview PDF.
- 1.W. Lotz und J. Nührenberg, Monte Carlo Simulation of Neoclassical Transport in Axisymmetric and Ripple Tokamaks, Z. Naturforsch. 37a, 899–905 (1982)Google Scholar
- 2.Allen H. Boozer and Gioietta Kuo-Petravic, Monte Carlo evaluation of transport coefficients, Phys. Fluids 24 (5), 851–859 (1981)Google Scholar
- 3.nCUBE2 manualGoogle Scholar
- 4.KSR1 manualGoogle Scholar
- 5.M. Ballico and H. Lederer, “Plasma Fusion Research: Multi-platform serial and parallel computing with one piece of source code on Cray YMP, nCUBE2, workstations using PVM and KSR1”, to appear in Max-Planck-Gesellschaft, Berichte und Mitteilungen (1994)Google Scholar