[Intel-gfx] Accel question for i965

Lukas Hejtmanek xhejtman at ics.muni.cz
Fri Mar 6 23:39:23 CET 2009


On Fri, Mar 06, 2009 at 01:11:56PM -0800, Eric Anholt wrote:
> It falls back to a two-pass system that is accelerated.  Until hardware
> has programmable blending, it's the best known way of doing accelerated
> component alpha.

and i965 does not support programmable blending?

I'm curious if we can see 1-2M glyphs/sec one day on i965 HW.

> > And is there any change to support get_image operation accelerated?
> 
> Sort of -- we could blit the contents to a new BO, and map the BO
> cached.  It may or may not be a win.

I doubt that there will be a win if cache flushes are involved.

-- 
Lukáš Hejtmánek



More information about the Intel-gfx mailing list