GPU scheduling

nobond

Newcomer
DX10.1 looks like are talking something like scheduling.

I can understand that generally from a general processor point of view.
But general processor hw pipeline is short, so the schedule is mainly os's task.

GPU pipeline is very long. may suffer longer latency. Any thoughts or ideas?
 
Where are you getting this from?

The scheduling is part of WDDM which sits below the D3D10 API. The aim, last I saw of it, was to get WDDM to allow GPU based scheduling that could interrupt at the fragment (or primitive?) level.

Jack
 
Back
Top