A simple but very effective method for parallelizing Lemke's algorithm for the solution of linear complementarity problems is presented. Implementation details on a 32-node Intel iPSC/2 hypercube for problems of dimension up to 1000 are discussed. A speedup efficiency as high as 76% is achieved with 32 processing nodes for a problem with 500 variables and 250,000 nonzero elements. By combining the effects of concurrency and vectorization the computing time on the Intel iPSC/2 in some cases is reduced by a factor of 100.
Parallel implementation of Lemke's algorithm on the hypercube
DE LEONE, Renato;
1991-01-01
Abstract
A simple but very effective method for parallelizing Lemke's algorithm for the solution of linear complementarity problems is presented. Implementation details on a 32-node Intel iPSC/2 hypercube for problems of dimension up to 1000 are discussed. A speedup efficiency as high as 76% is achieved with 32 processing nodes for a problem with 500 variables and 250,000 nonzero elements. By combining the effects of concurrency and vectorization the computing time on the Intel iPSC/2 in some cases is reduced by a factor of 100.File in questo prodotto:
Non ci sono file associati a questo prodotto.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.