XAA vs. EXA in old hardware support

[Adam Jackson]

On Tuesday 20 June 2006 17:15, Michael Lorenz wrote:
> Hello,
>
> > Of course, trying again with current hardware would make sense; things
> > are different now.
>
> Hmm, I did some measurements as well ( previous post... ) and repeating
> the same tests on XFree86 4.5 showed a consistent, significant
> performance advantage for XFree86. About 10% in low level tests like
> line- and font drawing, about 20% in -map.

I've seen scattered reports of similar.  I'd love to see someone profile it 
and find out exactly where we're spending our time.  The obvious first guess 
is the extra GC wrap for the Damage extension; try turning it off and 
measuring again.

Regarding the rest of the thread though, while it's completely plausible for 
exa to decompose the rest of core X down to calls to the driver's Solid 
routine - and in fact mi's span decomposition will do most of the work - it's 
also still the case that the driver is just as capable of wrapping the GC and 
accelerating rendering directly as it ever was.  exa has pretty decent 
fallback debugging these days.  Build it, turn it on, and see what you 
actually hit.

- ajax
-------------- next part --------------
A non-text attachment was scrubbed...
Name: not available
Type: application/pgp-signature
Size: 189 bytes
Desc: not available
URL: <http://lists.x.org/archives/xorg/attachments/20060620/8b8b1429/attachment.pgp>