That sounds plausible. I’m sure having the lowest possible latency was their goal. There are multiple popular rust libraries aiming to provide zero cost abstractions over a common subset of the metal, vulcan and dx12 APIs, but I’ve never actually used one.