ConceptSystems & CS
Kernel-level overlap
An execution approach where GEMM computations and reduction-scatter communications are launched as separate kernels with device synchronizations in between.
Know first
Named as a prerequisite by the lectures that define this. The number is how many say so. A grey one has no page yet.
CUDA streams1GPU kernels1
Opens up
Concepts that name this one as a prerequisite.
Lecturers
1Connections
2The lectures that define it or assume it, and the concepts next to it.