EXA patches

[Maarten Maathuis]

2009/8/3 Michel Dänzer <michel at daenzer.net>:
> On Mon, 2009-08-03 at 18:56 +0200, Maarten Maathuis wrote:
>> There is one thing that did change, driver pixmaps are not allocated
>> by default (only when exa deems acceleration doable). So for dri2
>> pixmaps you would need to create a scratch pixmap and attach a bo
>> afterwards.
>
> That doesn't sound like a very nice solution. Maybe
> exaGetPixmapDriverPrivate() could call the driver CreatePixmap(2) hook
> if there is no private yet, but I'm not sure offhand what should happen
> if that returns NULL then.

I have something in mind which would work nicely, except if it's
1bpp/4bpp (since they are classified as useless to accelerate) or if
the driver fails to return something useable, is that ok?

>
>
>> About the performance, i didn't notice that, but we (nouveau) place
>> pixmaps in vram from the moment it's created, [...]
>
> So does the radeon driver I'm using. So you're not seeing the slowdown
> with nouveau?

A minor slowdown, maybe 1% is observed. They all fall into the range
of 472K-477K, the first being the lowest value for mixed, the latter
the highest value for !mixed. Often it's 473K vs 475K.

>
>
> --
> Earthling Michel Dänzer           |                http://www.vmware.com
> Libre software enthusiast         |          Debian, X and DRI developer
>