On Sun, Oct 16, 2011 at 9:25 PM, Michael <[email protected]> wrote:
> Hello,
>
> On Sun, 16 Oct 2011 21:08:36 -0400
> Alex Deucher <[email protected]> wrote:
>
>> On Sun, Oct 16, 2011 at 9:06 PM, Michael <[email protected]> wrote:
>> > Hello,
>> >
>> > On Sun, 16 Oct 2011 11:37:16 -0400
>> > Matt Turner <[email protected]> wrote:
>> >
>> >> On Sat, Oct 15, 2011 at 6:20 PM, Mark Kettenis <[email protected]> 
>> >> wrote:
>> >> >> Date: Thu, 13 Oct 2011 20:26:18 -0700
>> >> >> From: Matt Kraai <[email protected]>
>> >> >>
>> >> >> > Also, that code is pretty much going to be made obsolete by Jeremy
>> >> >> > Huddleston's pci bus cleanup diff.  A better approach would be to
>> >> >> > convert the relevant drivers for loongson to use the new pci_io_xxx
>> >> >> > interfaces in libpciaccess instead of the raw in/out calls.
>> >> >>
>> >> >> My system uses the siliconmotion driver.  Would I need to start with
>> >> >> the instructions at
>> >> >>
>> >> >>  http://www.x.org/wiki/PciReworkHowto
>> >> >
>> >> > The work described there has already been done some time ago.
>> >> >
>> >> >> and then replace the raw in/out calls?
>> >> >
>> >> > Yes.  You'd want something like the (untested) diff below.  Should
>> >> > work if you have a recent enough libpciaccess.
>> >> >
>> >> > This raises the question how far one should go in providing backwards
>> >> > compatibility in the drivers for older libpciaccess libraries.  Do we
>> >> > need to worry about that at all?
>> >>
>> >> This particular hardware is really only XAA-capable to my knowledge.
>> >
>> > SM502 at least supports separate offsets and strides for source and 
>> > destination in blitter ops, so supporting EXA wouldn't be hard at all. 
>> > There seems to be some (limited) alpha blending support as well.
>>
>> There is working EXA support already in the smi driver, but IIRC, the
>> 3D hardware only has 1 texture unit, so it's not real useful for
>> composite.
>
> Nice :)
> You're probably talking about SM7xx since SM502 has no 3D capabilities at all.
> ( I only messed with Linux on my Gdium which is quite outdated. NetBSD is 
> almost there though, I have Xorg 1.10 working but without hw acceleration so 
> far. )
> It would already help if we could only accelerate anti-aliased font 
> rendering, doing that in software on a machine like that is quite annoyingly 
> slow ( Just watch CPU usage when a lot of junk scrolls through a konsole 
> window )
>

That generally requires a mask which needs 2 texture units.

Alex
_______________________________________________
[email protected]: X.Org development
Archives: http://lists.x.org/archives/xorg-devel
Info: http://lists.x.org/mailman/listinfo/xorg-devel

Reply via email to