Lagged Fibonacci random number generators for distributed memory parallel computers (Q1373786): Difference between revisions

In order to provide an efficient and good quality parallel random number generator, the author studies the parallelization of lagged Fibonacci generators for distributed memory parallel computers, using the contiguous subsequence technique and the leapfrog technique. By use of the contiguous subsequence technique a parallelization of the lagged Fibonacci plus/minus generators is presented. Applying the leapfrog technique, the author shows that lagged Fibonacci generators with the exclusive or operator can be efficiently parallelized without any communication overhead if the number of processors is a power of 2, and that it is not possible to parallelize other lagged Fibonacci generators efficiently in a communication-free manner. Furthermore, for the proposed algorithms, the author also discusses issues arising in the algorithm implementations and comments on their practical efficiency.