Implementation of a cell-vertex FV code for turbulent transonic flows on a Meiko computing surface

Abstract

A parallel, domain-decomposed version of a finite-volume scheme for the calculation of turbulent transonic flow has been developed and implemented on a T-800 Meiko Computing Surface. The code combines a cell-vertex-storage arrangement of flow variables with a Lax-Wendroff time-marching scheme which advances the solution to the steady state. Approaches to dealing with the extensive data dependencies arising from the above discretisation method are considered, and related implementation issues are discussed. Parallel speed-up and efficiency levels achieved with a tile-decomposition strategy are reported for inviscid and turbulent transonic flow over a bump in a two-dimensional channel. Efficiency values are in the range 75–98%, depending on problem type, mesh size and granularity of the decomposed domain.