Title :
Generating efficient parallel code for successive over-relaxation
Author_Institution :
Dept. of Math. & Comput., Univ. of Southern Queensland, Toowoomba, Qld., Australia
Abstract :
A complete suite of algorithms for parallelizing compilers to generate efficient SPMD code for SOR problems is presented. By applying unimodular transformation before loop tiling and parallelization, the number of messages per iteration per processor is reduced from 3n -1 in the conventional parallel SOR algorithm to 2n-1, where n is the dimensionality of the data set. To maintain the memory-scalability, a novel approach to use the local dynamic memory of parallel processors to implement the skewed data set is proposed
Keywords :
parallel algorithms; parallelising compilers; SOR problems; SPMD code; local dynamic memory; memory-scalability; parallel code generation; parallel processors; parallelizing compilers; skewed data set; successive over-relaxation; unimodular transformation; Australia Council; Concurrent computing; Data communication; Iterative algorithms; Iterative methods; Jacobian matrices; Mathematics; Parallel processing; Partial differential equations; Scalability;
Conference_Titel :
Algorithms and Architectures for Parallel Processing, 1997. ICAPP 97., 1997 3rd International Conference on
Conference_Location :
Melbourne, Vic.
Print_ISBN :
0-7803-4229-1
DOI :
10.1109/ICAPP.1997.651517
