Multitasking on the Cray-2 and Cray Y-MP: An Experimental Study

摘要

Abstract. This paper presents the results of an experiment to study the performance of multitasking techniques on the Cray-2 and Cray Y-MP using a single algorithm. The algorithm is a compact difference scheme for the solution of the incompressible, two-dimensional, time dependent Navier-Stokes equations. Two implementations of multitasking on both machines are considered. These are: macrotasking (parallelism at the subroutine level) and microtasking (parallelism at the do-loop level). These two techniques are briefly described. The imple- mentation of the algorithm is discussed in relation to these techniques and the results for three problem sizes are presented. The timing results for both tech- niques using few processors are in general comparable on both machines. For the eight processor case on the Cray Y-MP, microtasking outperformed macro- tasking for the small problems while the latter outperformed the former for the larger problem. The best achieved processing rate on the Cray Y-MP is 1.2 GFLOPS using macrotasking. Finally comparisons are made and conclusions are drawn.