P04: Unstructured-Grid CFD Algorithms on Many-Core Architectures
Abstract: In the field of computational fluid dynamics (CFD), the Navier-Stokes equations are often solved using an unstructured-grid approach to accommodate geometric complexity. Furthermore, turbulent flows encountered in aerospace applications generally require highly anisotropic meshes, driving the need for implicit solution methodologies to efficiently solve the discrete equations. These approaches require frequent construction and solution of large, tightly-coupled systems of block-sparse linear equations.
We explore the transition of two representative CFD kernels from a coarse-grained MPI-based model originally developed for multi-core systems to a shared-memory model suitable for many-core platforms. Results for the Intel Xeon Phi Knights Landing, NVIDIA Pascal P100, and NVIDIA Volta V100 architectures are compared with the aforementioned MPI-based implementation for the multi-core Intel Xeon Broadwell (BWL) processor. We observe substantial speedups over BWL as well as higher performance per dollar MSRP and performance per watt for the many-core architectures.
Award: Best Poster Finalist (BP): no
Two-page extended abstract: pdf