Nouveau Companion 35

Malc0 and Stillunknown are preparing to make Randr1.2 the default code path for Nouveau. In order to do that, they created a "Meta-Bug" which tracks all Randr1.2 related bugs, which prevent us from switching (bug report).

Some movement on the PPC front too. First thing was that moondrake found a workaround for the hard lockups on PPC. He needed to synchronize the FIFOs more often when uploading large amounts of data to the card. Kelnos checked moondrake's quick patch and found his lockups were gone too.

However this patch was found to be only avoid the real bug, which was memory corruption. This in turn was caused by AGP not working for write backs on his Powerbook.
Benh and Marcheu joined to help and patches were flying around but a final solution wasn't found yet. The latest patch incarnation from moondrake is here. While investigating the PPC problems, marcheu did notice though that the DDX code for AGP PPC was totally brain dead.

During testing the idea came up to compare certain register settings on both Linux and OSX to see where they might differ. kelnos offered to shrink his Linux partition in order to install OSX. Enter sbriglie who already had OSX running on the same hardware. He offered to test Airlied's radeontool for OSX, but he couldn't get it to work.

But moondrake didn't give up; he tried different modes, disabled SBA (Side Band Addressing) and finally prevent uninorth-agp from loading (a driver for the Apple northbridge used). That did the trick, his lockups and corruption problems went away.

Additionally, Malc0 fixed the "NV34 does work only every second try" problem when he encountered the same problem on x86 too (git commit).

Ok, now for the usual assortment of short topics:

- jb17some is still working on the XvMC acceleration. He summarized his current status on the Wiki
- Airlied did a presentation at LCA where he talked about the current status of graphics drivers. Rumor has it that he showcased our Gallium3D driver, too. But look here (Look for kittens).
- pq is working on including MMioTrace into the kernel. Although the kernel hackers like Ingo Molnar are very helpful, but the task is much bigger than originally estimated. So PQ is aiming for 2.6.26 now.
- Darktama is trying to find out a bit more about why 3D renderings fail completely on NV5x.
- And Darktama is thinking about creating a "voodoo generator" which would allow creating the voodoo for context switches programmatically.

Help Needed

As always:
- Look on our TestersWanted page.
- Check for regressions in the randr1.2 code

Please check the "use after blob" fix from Stillunknown on NV4x. And if you have hard lockups on NV4x (or even better NV40), please do contact stillunknown.

And yes, PPC testers are welcome (developers too :)) as are patches for BSD, as no one came forward when we asked and offered support for those users in one of our last issues.

Discuss this article in our forums, IRC channel, or email the author. You can also follow our content via RSS and on social networks like Facebook, Identi.ca, and Twitter (@Phoronix and @MichaelLarabel). Subscribe to Phoronix Premium to view our content without advertisements, view entire articles on a single page, and experience other benefits.