Optimization of multigrid based elliptic solver for large scale simulations in the FLASH code

• DOI: 10.1002/cpe.2821
• Publication Date: 2012-02-14T03:32:29Z
• Authors: Christopher Daley; Marcos Vanella; Anshu Dubey; Klaus Weide; Elias Balaras
• Keywords: Multigrid method; Solver; Adaptive Mesh Refinement; Multiphysics; Benchmark (surveying); Code (set theory); Software portability

SUMMARY FLASH is a multiphysics multiscale adaptive mesh refinement (AMR) code originally designed for simulation of reactive flows often found in Astrophysics. With its wide user base and flexible applications configuration capability, has dual task maintaining scalability portability all solvers. The fully explicit solvers the tied very closely to that underlying mesh. Others such as Poisson solver based on multigrid method have more complex scaling behavior. Multigrid methods suffer from processor starvation dominating communication costs at coarser grids with increase number processors. In this paper, we propose combination uniform grid AMR mesh, merger two different sets overcome limitation FLASH. principal challenge proposed efficiency algorithm map back forth between AMR. We present parallel mapping algorithms also discuss results performance studies implementations. Copyright © 2012 John Wiley & Sons, Ltd.