Parallel Algorithms for Astrophysics Problems

Abstract

The algorithm and the mathematical modeling package for 3-D gravitational gasdynamics problems with ultra-high resolution meshes are described. The modeling results of filamentary formations, i.e. concentrated areas with high gas density in molecular clouds (MC), and the nonisothermic compression calculation data are discussed. The spatial mesh resolution required for satisfying Jeans conditions in modeling is substantiated. The programming code developed uses dynamic gridding called local adaptive mesh refinement (AMR) at several (up to 10) resolution levels. To provide adequate resolution themeshes are added automatically and dynamically as well as destroyed as needed. The computation paralleling algorithm withOpenMPand CUDA is given. The programming language chosen to compute the problems of gravitational gas dynamics efficiently is justified and substantiated. The practice of applying algorithms for modeling the MC fragmentation after collisions, the filament and protostellar clouds formation, the star formation stages is analyzed.