Bug 28033

Summary: [r300, r300g] World of Warcraft login screen is "blocky" with RV350
Product: Mesa Reporter: Chris Rankin <rankincj>
Component: Drivers/DRI/r300Assignee: Default DRI bug account <dri-devel>
Status: RESOLVED FIXED QA Contact:
Severity: normal    
Priority: medium    
Version: git   
Hardware: x86 (IA32)   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments: Screenshot of the World of Warcraft login page
AGP patch

Description Chris Rankin 2010-05-08 17:15:08 UTC
Created attachment 35513 [details]
Screenshot of the World of Warcraft login page

The login screen for World of Warcraft contains a "snow scape". However, some of the snow looks as if it's been misrendered as blocks.

This bug happens both with "classic" Mesa and gallium. The card is an AGP Radeon 9550 (RV350); Linux 2.6.33.3, KMS, xf86-xorg-drv from git.
Comment 1 Miklós Máté 2010-06-19 14:12:57 UTC
This looks like a duplicate of #8056. Both r300c and r300g have problems setting up tiling on r3xx cards.
Comment 2 Marek Olšák 2010-06-19 14:28:31 UTC
To my knowledge, tiling in r300g just works.

Chris, can you try the game again with current mesa master?
Comment 3 Chris Rankin 2010-06-19 15:47:06 UTC
(In reply to comment #2)
> Chris, can you try the game again with current mesa master?

It's still broken with mesa master and a 2.6.33.5 kernel. Do I need a 2.6.34.x kernel instead?
Comment 4 Marek Olšák 2010-06-19 16:11:41 UTC
No, you don't.

I think I know that's wrong. I'll try to fix it soon.
Comment 5 Marek Olšák 2010-06-19 16:12:44 UTC
s/that/what/
Comment 6 Marek Olšák 2010-06-19 21:18:49 UTC
I've pushed the fix to master. Can you test now?
Comment 7 Chris Rankin 2010-06-20 05:09:16 UTC
(In reply to comment #6)
> I've pushed the fix to master. Can you test now?

Yes, the login screen looks *much* better under r300g now, thanks :-). But the exact same bug is still present with r300c. Any chance that this fix can be applied to r300c too, please?
Comment 8 Marek Olšák 2010-06-20 05:15:25 UTC
r300c is no longer being developed, though it should be quite easy to backport the fix. If there is any other issue with r300g, please file another bug. Closing...
Comment 9 Chris Rankin 2010-06-20 05:35:19 UTC
(In reply to comment #8)
> r300c is no longer being developed, though it should be quite easy to backport
> the fix.

Thanks; backporting the fix would be a good idea because Gallium is currently very, VERY slow with AGP cards... ;-).
Comment 10 Marek Olšák 2010-06-20 05:56:51 UTC
Created attachment 36376 [details] [review]
AGP patch

Does the attached patch improve the situation on your AGP chipset?
Comment 11 Chris Rankin 2010-06-20 06:26:34 UTC
(In reply to comment #10)
> Does the attached patch improve the situation on your AGP chipset?

No, I'm afraid not. My test case is to run celestia, and then try to rotate the Earth by right-dragging it with the mouse. This is speedy with r300c, but drags horribly with r300g, even with your patch applied.
Comment 12 Chris Rankin 2010-06-27 13:05:54 UTC
(In reply to comment #8)
> r300c is no longer being developed, though it should be quite easy to backport
> the fix. 

I've had a look at backporting this fix to r300c myself because I suspect that it's one of the very last bugs keeping WoW "second rate" on Radeon hardware with classic Mesa. (Gallium would currently be unplayably slow, even if it didn't crash first...) However, while the fix seems conceptually described as "calculate a texture cache region and set these bits in the texture format, I cannot even begin to identify which are the corresponding parts of the r300c code that would need to be changed in order to backport the fix correctly.

So I'm asking again: can someone please backport this fix into classic Mesa? It would seem like an obvious candidate, because it's a genuinely identified problem that users in the Real World are actually hitting, and where using Gallium instead is simply not possible.

Thank you.

Use of freedesktop.org services, including Bugzilla, is subject to our Code of Conduct. How we collect and use information is described in our Privacy Policy.