100 Days of learning about graph compute – Day 13

Optimizing, optimizing, optimizing

Kernel profiling, tracing and detailed performance monitoring are the kinds of things which one does not need to master UNTIL the time comes, but it is necessary at the start to have a modicum of awareness of how these tools might help refine developments or how optimization might be achieved. As we moved to [possibly underpowered] edge platforms, this kind of competency becomes more valuable or likely essential.