Fig. 5From: Accelerating CFD simulation with high order finite difference method on curvilinear coordinates for modern GPU clustersDomain decomposition strategies of HiResX. The domain is divided into several blocks according to performance of CPU and GPU in order to balance workloads of each process. For the blocks computed by CPU, the whole block is computed by a CPU process. For the blocks computed by GPU, each grid cell is computed by one CUDA threadBack to article page