Challenges Faced When Porting Ginkgo to the SYCL Ecosystem
We present our experience of a platform-portable sparse linear algebra ibrary – Gingko. Previously, we already ported essential parts of our library like SpMV or Krylov solvers to oneAPI based on sycl-1.2.1. We present challenges and our solutions when porting functionality and in stayin up to date due to the changed SYCL 2020, issues with per-kernel code splitting, and supporting varying hardware. We generate kernels suitable for different devices and give our approach of pre-compiled kernels. To wrap up, we show some experimental results of preconditioners or Algebraic Multigrad.
Resources:
https://github.com/ginkgo-project/ginkgo/
Speaker(s)
Yu-Hsiang Mike Tsai
Yu-Hsiang Mike Tsai is a Ph.D. Student at the Karlsruhe Institute of Technology. He joined the Ginkgo Team in 2019. Currently, he focues on designing mixed precision algorithms, algebraic multigrid methods, and accelerating GPU kernels for AMD, Intel, and NVIDIA.
