DLR's TAU Code on Massively Parallel Architectures - Hybrid MPI/Shared-Memory Parallelization
Jens Jägersküpper · elib (German Aerospace Center) · 2010
Outline - Intro to DLR's TAU code for CFD - The HICFD project, short intro - SIMD - single instruction multiple data - DLR’s TAU code on many cores - Hybrid parallelization: shared memory (threads) + MPI - Domain decomposition (for MPI) - Multiple threads for a single (MPI) domain - Summary and Outlook