Nouveau - Sep 2015 - [Bug 91895] New: [nve7] Shadow Warrior: black gun & messed up colors

https://bugs.freedesktop.org/show_bug.cgi?id=91895

            Bug ID: 91895
           Summary: [nve7] Shadow Warrior: black gun & messed up colors
           Product: Mesa
           Version: git
          Hardware: x86-64 (AMD64)
                OS: Linux (All)
            Status: NEW
          Severity: normal
          Priority: medium
         Component: Drivers/DRI/nouveau
          Assignee: nouveau at lists.freedesktop.org
          Reporter: marcin.slusarz at gmail.com
        QA Contact: nouveau at lists.freedesktop.org

Created attachment 118102
  --> https://bugs.freedesktop.org/attachment.cgi?id=118102&action=edit
Shadow Warrior on Nouveau

Screenshots comparing intel and nouveau attached.
Apitrace:
http://people.freedesktop.org/~mslusarz/ShadowWarrior.bin.x86.trace.xz (174MB)

This is on mesa-git (3c6c4d4f298ec81fe57992790a68aaab2e573519) with the patch
from bug#91890.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/1e241b20/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

Marcin Slusarz <marcin.slusarz at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
 Attachment #118102|text/plain                  |image/png
          mime type|                            |

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/cac37be0/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #1 from Marcin Slusarz <marcin.slusarz at gmail.com> ---
Created attachment 118103
  --> https://bugs.freedesktop.org/attachment.cgi?id=118103&action=edit
Shadow Warrior on Intel

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/b43120f3/attachment-0001.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #2 from Ilia Mirkin <imirkin at alum.mit.edu> ---
GlennK reproduced the same issue on r600, so it's most likely a st/mesa bug.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/e60e42b1/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #3 from Ilia Mirkin <imirkin at alum.mit.edu> ---
(In reply to Ilia Mirkin from comment #2)
> GlennK reproduced the same issue on r600, so it's most likely a st/mesa
bug.
Errr... two issues. One is the lighting issue, which also happens on r600.
Another is the blacked out bits, which is nouveau-specific.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/b7cf9130/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #4 from Ilia Mirkin <imirkin at alum.mit.edu> ---
Probably relevant to the nouveau-specific issue:

ERROR: no viable spill candidates left

I bet the end result is black, which explains those spots. Haven't done more
debugging on that yet.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/c685833e/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #5 from Ilia Mirkin <imirkin at alum.mit.edu> ---
Created attachment 118107
  --> https://bugs.freedesktop.org/attachment.cgi?id=118107&action=edit
vertex shader that fails to compile

Don't know if this is *the* issue, but it's certainly *an* issue. A
little
surprising that such a simple shader dies -- no control flow, no nothing. Fails
to find spill candidates.

Perhaps the function calls somehow mess it up.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/eb120165/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #6 from Marcin Slusarz <marcin.slusarz at gmail.com> ---
Hmm, I'm not sure anymore which lighting is the correct one...

One more trace with better view:
http://people.freedesktop.org/~mslusarz/ShadowWarrior.bin.x86.2.trace.xz (181
MB)

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/48ab0210/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #7 from Ilia Mirkin <imirkin at alum.mit.edu> ---
OK, I think I see what's happening... the sequence goes something like this:

  3: mov u32 %r438 0x3f800000 (0)

 16: mov u32 $r0 %r445 (0)
 17: mov u32 $r1 %r452 (0)
 18: call abs BUILTIN:1 (0)
 19: mov u32 %r455 $r1 (0)
 20: nop - { $r0 $r2d } (0)
 21: nop - $p0q (0)

337: mov u32 %r1227 %r1012 (0)
338: mov u32 %r1228 %r1014 (0)
339: mov u32 %r1229 %r954 (0)
340: merge b128 %r1222q %r1227 %r1228 %r1229 %r438 (0)
341: export b128 # o[0x10] %r1222q (0)

Now the resulting "code" looks like:

  1: mov u32 $r2 0x3f800000 (8)
  7: mov u32 $r10 0x00000004 (8)

263: mov u32 $r1073741823 $r10 (8)
264: mov u32 $r0 $r11 (8)
265: mov u32 $r1 $r16 (8)
266: export b128 # o[0x10] %r1222q (8)

So it must be trying to stick this stuff into r0..r3. However there's a
clobber
of $r2 between the initial move and the desire to use it, without any
constraint move being inserted. So the RA fails because it doesn't
understand
the situation properly. I *think*.

Not sure why the first 3 args of the merge have a mov but the 4th one
doesn't.
Perhaps something eliminates it for some reason? Also worth noting that this
works fine on GK110, presumably due to slightly different RA constraints.
However we don't get anywhere close to using up the 64 registers here
either.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/d8e31505/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #8 from Ilia Mirkin <imirkin at alum.mit.edu> ---
This patch should get it going, but I don't think I want to commit it as-is.
This gets me the same overly-lit rendering as r600, so any remaining issues are
in the st. Should probably investigate what the "real" rendering looks
like,
neither lighting matches any screenshots of the game I've found online.


diff --git a/src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp
b/src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp
index 0cd21cf..268ea0b 100644
--- a/src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp
+++ b/src/gallium/drivers/nouveau/codegen/nv50_ir_ra.cpp
@@ -2121,8 +2121,10 @@ RegAlloc::InsertConstraintsPass::insertConstraintMoves()

             Instruction *defi =
cst->getSrc(s)->defs.front()->getInsn();
             // catch some cases where don't really need MOVs
+/*
             if (cst->getSrc(s)->refCount() == 1 &&
!defi->constrainedDefs())
                continue;
+*/

             LValue *lval = new_LValue(func, cst->src(s).getFile());
             lval->reg.size = size;

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/defa03c8/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #9 from Marcin Slusarz <marcin.slusarz at gmail.com> ---
Yes, this fixes gun rendering.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150906/5bb6ede2/attachment-0001.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #10 from Ilia Mirkin <imirkin at alum.mit.edu> ---
This trace can't be used on i965:

62714 @0 glLinkProgram(program = 4038)
62714: warning: error: Uniform block MaterialCB too big (32768/16384)

Their UBO size is too small I guess.

GL_MAX_UNIFORM_BLOCK_SIZE = 16384

vs

GL_MAX_UNIFORM_BLOCK_SIZE = 65536

Although I bet that's a mistake...

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150908/e8167134/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #11 from Marcin Slusarz <marcin.slusarz at gmail.com> ---
The game behaves exactly the same WRT lighting under Intel driver. I can upload
a trace if you want.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150909/cf3245fa/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #12 from Ilia Mirkin <imirkin at alum.mit.edu> ---
(In reply to Marcin Slusarz from comment #11)
> The game behaves exactly the same WRT lighting under Intel driver. I can
> upload a trace if you want.
No, that's fine. The game probably doesn't check for UBO sizes and just
assumes
it'll work. I hacked it to expose a 64K buffer and it was fine. Ken sent a
patch to list which does this for real.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150909/e0ab9a6f/attachment-0001.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #13 from Marcin Slusarz <marcin.slusarz at gmail.com> ---
Fun, with bumped MAX_UNIFORM_BLOCK_SIZE lighting looks exactly the same on
Intel and Nouveau.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150909/a58054fc/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

Marcin Slusarz <marcin.slusarz at gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
            Summary|[nve7] Shadow Warrior:      |[nve7] Shadow Warrior:
                   |black gun & messed up       |black gun
                   |colors                      |

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20150909/ab26cba7/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

--- Comment #14 from Ilia Mirkin <imirkin at alum.mit.edu> ---
This patch fixes the trace: http://patchwork.freedesktop.org/patch/69589/

However it needs to get a lot more testing before I'll be ready to push
it...
all this RA stuff is a little over my head.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.freedesktop.org/archives/nouveau/attachments/20160107/d460e64f/attachment.html>

https://bugs.freedesktop.org/show_bug.cgi?id=91895

GitLab Migration User <gitlab-migration at fdo.invalid> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
         Resolution|---                         |MOVED
             Status|NEW                         |RESOLVED

--- Comment #15 from GitLab Migration User <gitlab-migration at
fdo.invalid> ---
-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been
closed from further activity.

You can subscribe and participate further through the new bug through this link
to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1082.

-- 
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<https://lists.freedesktop.org/archives/nouveau/attachments/20190918/28caf34c/attachment.html>