kernels

Engine upgrades change defaults, kernel selection and scheduling at once, so attribution is the whole problem. The bisection that works on a version rather than a commit, the four categories of cause, and the pre-upgrade baseline without which none of this is possible.

The weights are portable and almost nothing else is. Four layers have to exist before the model runs at all, the quantization format is the one most likely to be missing, and the honest plan states what will not work in the first quarter rather than promising parity.

Two divisions decide it: the kernel's FLOPs per byte against the card's ridge point. The worked case, the profiler counters that give the inputs, and what to change once you know which wall you hit.