Parallel solution of viscous incompressible flow on multi-block structured grids using MPI

Abstract

The aim of this chapter is to develop a parallel code based on an existing unsteady three-dimensional incompressible viscous flow solver for simulation of flows past complex configurations using multi-block structured grids and using message passing interface (MPI) for message passing. The linearized implicit solution algorithm used is modified for execution in parallel using a block-decoupled sub-iterative strategy, and a heuristic performance estimate is developed to guide the parallel problem definition. The chapter provides extensive evaluations of both algorithmic and parallel performance for a test problem consisting of axisymmetric flow past an unappended submarine hull with a Reynolds number of 12 million. Other results are given for three-dimensional appended submarine cases having 0.6M and 3.3M grid points. The decoupled sub-iteration algorithm allows the convergence rate of the sequential algorithm to be recovered at reasonable cost by using a sufficient number of sub-iterations, and allows decompositions having multiple subdivisions across boundary layers. The chapter concludes that the present approach gives reasonable parallel performance for large scale steady flow simulations wherein the grid and decomposition are appropriately sized for the targeted computer architecture.