Optimization of a Three-Dimensional Diagnostic Flow Solver.

摘要

In this study, we investigate the performance of a three-dimensional diagnostic flow solver. The numerical solver is profiled to find the 'hot spots' of the code. These hot spots are then optimized for performance using techniques such as loop reordering and cache blocking. OpenMP is used to parallelize the code execution, further enhancing its performance. Finally, preliminary efforts to use NVIDIA Inc. Compute Unified Device Architecture (CUDA) programming model are discussed.