Ginkgo’s SYCL backend and its integration into nekRS (EMEA)
The increasing diversity of hardware architectures makes it harder for simulation software to get everything working on different hardware. Simulation software stacks must rely on portability layers or software stacks that support different hardware architectures. We present Ginkgo, a math library that considers platform portability and can be used for linear solvers in nekRS CFD code. We have presented nekRS with Ginkgo on AMD and NVIDIA GPUs. With the next release of nekRS supporting Intel GPUs, we also extend the Ginkgo integration to keep the portability on Intel GPUs. This talk will show the performance results of the nekRS simulation with the Ginkgo solver on Intel GPUs.
