CVS and fglrx -- success!

Vladimir Dergachev volodya at mindspring.com
Mon Mar 7 20:42:45 PST 2005



On Sun, 6 Mar 2005, Adam Jackson wrote:

> On Sunday 06 March 2005 20:14, Ethan Glasser-Camp wrote:
>> With this setup I was able to run Crystal Space well, though jerkily,
>> which was the whole point the whole time. But I'm probably going to
>> forget about this whole mess for a while because it was a real pain to
>> get to, I have to shut down my old XFree86 server to use it (or get
>> terrible video corruption that will persist until a reboot), and I have
>> to use proprietary drivers. So I figured I'd ask -- are there plans to
>> support ATI's Mobility Radeon 9600 with OpenGL support and DRI in an
>> open source driver any time soon? How about over long-term? Are there
>> ways to watch this progress?
>
> http://r300.sf.net/
>
> It's being worked on, without the benefit of docs from ATI so it's slow
> progress.  It's getting to the point of being good enough to run quake3, but
> is still _far_ from stable.

Let me try to qualify that a bit.

The current state of Open Source r300 driver is that it is under active 
development and many features important for a 3d driver have been 
understood.

Adam is very correct in that the code base is changing rapidly and thus is 
not stable. Also not all rendering functions are working properly.
Lastly, the driver locks up on the some hardware.

The good news is that the driver is stable on some hardware from the 
point of view that it does not lockup.

For example, I have spent considerable time running Quake, kpovmodeller, 
nerverputt, neverball, ppracer, screensavers and have not had a lockup.
(This is with Radeon Mobility M10 (RV350) and  Pentium M using 
"gliding_penguin" driver snapshot).

On the other hand we know that R350 cards appear to lock up relatively 
soon after starting an application and this is somehow correlates to 
frame rate.

Newer Radeons have proven to be quite touchy and there are hardware 
differences between R300, RV350 and R420.

Thus it is likely that either you will experience a lockup right away or 
it will work for you.

Also here is a short (and likely incomplete) list of broken features:

      * stencil is broken
      * gourad is broken
      * there is some texture breakage, albeit "plain" textures work fine.
      * multitexturing is broken, but left enabled for experiments
        (so you'll have to disable it in application if you want "clean"
         rendering)
      * vertex buffer extension appears to be broken

You can find current list of working features at 
http://r300.sf.net/r300_dri.php

                     best

                        Vladimir Dergachev

>
> Most of the r300 discussions happen on the dri-devel@ mailing list.
>
> - ajax
>



More information about the xorg mailing list