Parallel Implicit Time-Accurate Navier-Stokes Computations Using Coarse Grid Correction

Abstract

A parallel implicit Navier ‐Stokes solver is constructed using a block-by-block inversion of the resulting linear system of equations followed by a global coarse grid correction to account for the block coupling. This solver is used to compute two different cases of vortex shedding e ows. The efe ciency of our method is compared to the efe ciency of the method based on pseudotime stepping and multigrid acceleration. This comparison is based on direct comparison of the two methods applied to identical test cases, as well as results reported in the literature. Emphasis is put on scalability as well as mesh sensitivity. It is concluded that the solver is not sensitive to grid stretching and scales well with both problem size and number of processors, and it is shown that measured CPU times compare well with the multigrid approach.