Kokkos III
摘要
In this final chapter on Kokkos, we look deeper into extracting more parallelism from our Kokkos code. So far, we have seen how to use Kokkos for high-level parallelism at the level of loops. However, when loops are nested within other loops, for example, we can potentially extract even more parallelism if the hardware on which we are running our code has the capability to execute greater degrees of parallelism and obtain more performance. Accordingly, we look into hierarchical parallelism and the methods that Kokkos provides for extracting additional parallelism from our code.