Under the Hood of SYCL – An Initial Performance Analysis with An Unstructured-Mesh CFD Application

As the computing hardware landscape gets more diverse, and complexity of grows, need for a general purpose parallel programming model capable developing (performance) portable codes have become highly attractive. Intel’s OneAPI suite, which is based on SYCL standard aims to fill this gap using modern C++ API. In paper, we use parallelize MG-CFD, an unstructured-mesh computational fluid dynamics (CFD) code, explore current performance SYCL. The code benchmarked several processor systems from Intel (including CPUs latest Xe LP GPU), AMD, ARM Nvidia, making variety compilers, with particular focus how it maps CPU GPU architectures. We compare other parallelizations available in OP2, including SIMD, OpenMP, MPI CUDA. results are mixed; class applications, when parallelized SYCL, depends target architecture compiler, but many cases comes close currently prevalent models. However, still requires different parallelization strategies or code-paths be written obtain best performance.