GENERATION OF QUASI-RANDOM (LP(TAU)) VECTORS FOR PARALLEL COMPUTATION

A procedure for generating sequences of LP tau vectors that are uniformly distributed in the multi-dimensional unit cube is presented here together with its implementation in ANSI-C. The high performance and the efficiency of the generating algorithm make sequences of LP tau vectors preferable to cope with some problems traditionally associated with Monte Carlo methods. The suggested implementation uses the recurrent algorithm whenever it is possible, or executes the algorithm producing LP tau vectors from their numbers otherwise. This approach shortens the average time of vector generation during massive computation and enables the procedure to be used for parallel computation. Some of the fundamental properties of sequences of LP tau vectors are exemplified in the paper in comparison with Faure sequences.