Benefits of parallel processing in stochastic simulations of fish breeding programs
Ingrid Olesen, Jørn Amundsen · Bulletin - International Bull Evaluation Service/Interbull bulletin · 1999
Experiences with adapting and running a Fortran 77 program for stochastic simulation of a mass selection program in fish breeding for parallel processing are presented. Little rewriting of the program was required to assign different replicates of the simulations to different processors. In order to study the speedup according to number of processors, a case of 20 replicates, requiring 2 000 s using one processor, was run. Wall clock time for the program (without initial input) using 1, 2, 4, 10 and 20 processors were obtained. The saved computer time increased considerably with increased number of processors. Wall clock time was 17 times lower on a Cray T3E using 20 parallel processes compared to sequential execution (116 s versus 2000 s). This is expected to be improved for a more realistic and demanding case.