Isolating the Reasons for the Performance of Parallel Machines on Numerical Programs
Arno Formella, Silvia M. Müller, Wolfgang J. Paul, Anke Bingert · 1994
In this paper we present a nontrivial set of modules which measure performance parameters of node processors and interconnection networks. With the help of these parameters we explain the mu time of the following algorithms conjugate gradient method , one-dimensional partial differential equation solver and two-dimensional partial differential equation solver on the parallel machine Ncube-2. The iPSC/860 Hypercube and the vector machine VP100 are analyzed in an other paper (see [3]). Our explanations are sometimes within 0.5% and almost always within 5% of the measured run times. These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.