Chapter 8: Parallel scaling
Ed Bueler · Society for Industrial and Applied Mathematics eBooks · 2020
Many users arrive at PETSc wanting to do large-scale parallel computations, knowing that the mathematical concepts in this library can lead to high performance parallel solutions of nonlinear PDEs and multiphysics models. Thus we need a vocabulary for quantifying parallel performance, and this chapter defines several measures of how solver performance scales as the number of processes increases. However, the scaling of algorithmic work done by the solver, as the number of degrees of freedom increases, i.e., the solver complexity (Chapter 7), is at least as important and will remain prominent.