Do mesh editors ever use GPUs for mesh operations? For example, some of the global operations seem like they could be accelerated effectively. If a mesh editor were to use a GPU, are there mesh data structures designed to work more efficiently for these processors? It seems like a lot of the pointer chasing operations we do to operate on the mesh might be inefficient since they look hard to coalesce and might diverge across threads.

It seems like GOUs would only be useful in a subsect of operations (like global ops for example), most local operations probably don't need or use GPUs.

Agreed. I imagine some of the other features of the mesh editor might also work well on a GPU (e.g., animation)