Bug 60034 - rv790 etqw regression since r600g: add async for staging buffer upload v2
Summary: rv790 etqw regression since r600g: add async for staging buffer upload v2
Status: RESOLVED FIXED
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/Gallium/r600 (show other bugs)
Version: git
Hardware: x86 (IA32) Linux (All)
: medium normal
Assignee: Default DRI bug account
QA Contact:
URL:
Whiteboard:
Keywords:
Depends on:
Blocks:
 
Reported: 2013-01-29 19:57 UTC by Andy Furniss
Modified: 2013-02-07 15:47 UTC (History)
0 users

See Also:
i915 platform:
i915 features:


Attachments
worse case from cold boot (1.66 MB, image/png)
2013-01-29 19:59 UTC, Andy Furniss
Details
best case after repeated runs (1.54 MB, image/png)
2013-01-29 20:00 UTC, Andy Furniss
Details
compressed text output from etqw apitrace dump (700.70 KB, application/octet-stream)
2013-01-31 20:33 UTC, Andy Furniss
Details
work around (796 bytes, patch)
2013-02-04 22:48 UTC, Jerome Glisse
Details | Splinter Review
Fix (3.60 KB, patch)
2013-02-06 18:57 UTC, Jerome Glisse
Details | Splinter Review

Description Andy Furniss 2013-01-29 19:57:49 UTC
Card PCIE HD4890, drm fixes kernel.

Since -

commit 325422c49449acdd8df1eb2ca8ed81f7696c38cc
Author: Jerome Glisse <jglisse@redhat.com>
Date:   Mon Jan 7 17:45:59 2013 -0500

    r600g: add async for staging buffer upload v2
    
    v2: Add virtual address to dma src/dst offset for cayman


I can see some reflection rendering errors in etqw.

They are not visible on most maps but looking from high up on island where the sky is reflected in the sea they are obvious from a cold boot, but far more subtle after running repeatedly for bisect.

attached screenshots showing both cases.
Comment 1 Andy Furniss 2013-01-29 19:59:28 UTC
Created attachment 73868 [details]
worse case from cold boot
Comment 2 Andy Furniss 2013-01-29 20:00:35 UTC
Created attachment 73869 [details]
best case after repeated runs
Comment 3 Andy Furniss 2013-01-29 20:11:18 UTC
Comment on attachment 73868 [details]
worse case from cold boot

Ignore low fps on this compared to other screen - the card was on low.
Comment 4 Jerome Glisse 2013-01-30 17:24:52 UTC
Please update mesa again a bunch of fixes were push. Make sure you have up to :
e1598cb642334c809e6ec219d793e7bc85a213de
Comment 5 Andy Furniss 2013-01-30 18:10:45 UTC
(In reply to comment #4)
> Please update mesa again a bunch of fixes were push. Make sure you have up
> to :
> e1598cb642334c809e6ec219d793e7bc85a213de

I am still seeing it on current head.
Comment 6 Jerome Glisse 2013-01-31 16:34:43 UTC
Can you capture a trace using https://github.com/apitrace/apitrace your distribution might have package for it. Also try to make the trace as small as possible by jumping directly to the issue.
Comment 7 Andy Furniss 2013-01-31 20:20:23 UTC
(In reply to comment #6)
> Can you capture a trace using https://github.com/apitrace/apitrace your
> distribution might have package for it. Also try to make the trace as small
> as possible by jumping directly to the issue.

Well I got one but it's 186 meg compressed.

That's as small as I can get it - made a demo where you can see the issue and then loaded it directly from command line +exec a_script thus bypassing the 3D menu screen and made the game quit after rendering just 5 frames (1 isn't any better).

I guess the screen while the map is (slowly) loading is accounting for a lot of the data.

Do you have somewhere I can upload 186 meg?
Comment 8 Andy Furniss 2013-01-31 20:33:33 UTC
Created attachment 74016 [details]
compressed text output from etqw apitrace dump

If it's just the output of apitrace dump that's needed then this is a lot smaller.
Comment 9 Jerome Glisse 2013-01-31 21:06:01 UTC
No it's the trace that's needed. You can try on some of those free file hosting otherwise you can dcc send it to me on irc (freenode glisse) and i will put it somewhere
Comment 10 Andy Furniss 2013-02-01 16:06:03 UTC
(In reply to comment #9)
> No it's the trace that's needed. You can try on some of those free file
> hosting otherwise you can dcc send it to me on irc (freenode glisse) and i
> will put it somewhere

Here it is -

http://www.andyqos.ukfsn.org/etqw-trace.xz
Comment 11 Jerome Glisse 2013-02-04 22:48:16 UTC
Created attachment 74200 [details] [review]
work around

Can you check that this work around fix the rendering ?
Comment 12 Andy Furniss 2013-02-04 23:45:24 UTC
(In reply to comment #11)
> Created attachment 74200 [details] [review] [review]
> work around
> 
> Can you check that this work around fix the rendering ?

Yes, this fixes etqw for me.

I see someone else reported an issue with this commit and xonotic.

I haven't checked if xonotic was OK before, but I can see the issue and this patch does not help it.

https://bugs.freedesktop.org/show_bug.cgi?id=60236
Comment 13 Jerome Glisse 2013-02-06 18:57:05 UTC
Created attachment 74299 [details] [review]
Fix

Do this fix the etqw issue for you too ?
Comment 14 Andy Furniss 2013-02-06 19:39:06 UTC
(In reply to comment #13)
> Created attachment 74299 [details] [review] [review]
> Fix
> 
> Do this fix the etqw issue for you too ?

Yes it fixes etqw, but not xonotic.
Comment 15 Jerome Glisse 2013-02-06 20:04:35 UTC
I attached fix for xonotic to the xonotic bug
Comment 16 Jerome Glisse 2013-02-07 15:47:21 UTC
Fix pushed upstream closing this one


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.