Automatic Parallelisation of Sturctured Mesh Computations with SYCL
Gábor Dániel Balogh, István Z. Reguly · 2021
Structured meshes are widely used for scientific computations such as Computational Fluid Dynamics (CFD) applications or finance. Modern applications often have grid points in the millions. To perform such computations parallelisation is crucial. However it is unfeasible to port each application every time a new architecture arrives, hence in recent years the demand for automatic parallelisation and optimisation for the used hardware is increasing. The OPS (Oxford Parallel library for Structured mesh solvers) has shown good performance and scaling on a wide range of HPC architectures. This research aims to extend the OPS framework with a SYCL backend to extend the range of architectures that OPS can support and further increase Performance Portability of OPS applications. The performance of the Intel OneAPI is struggling with reductions due to high synchronisation cost, but shows promising performance gain on builtin reduction constructs on an Intel® Xeon® Gold 6226R. We compare the performance of hipSYCL on NVidia V100 GPU to the CUDA implementations.