Re: [Intel-gfx] [PATCH] drm/i915: Don't stall too much for cursor vs. pageflip

2016-06-22 Thread Chris Wilson
On Wed, Jun 22, 2016 at 03:47:15PM +0200, Daniel Vetter wrote: > - We don't need to wait for the previous commit to have fully > completed, the waiting for hw_done in swap_state is sufficient. > > - We need to make sure that the pure page_flip signals hw_done early > enough. This is done throu

[Intel-gfx] [PATCH] drm/i915: Don't stall too much for cursor vs. pageflip

2016-06-22 Thread Daniel Vetter
- We don't need to wait for the previous commit to have fully completed, the waiting for hw_done in swap_state is sufficient. - We need to make sure that the pure page_flip signals hw_done early enough. This is done through a gross hack by recomputing stuff. I think it'd be better to fix thi