[Bug 57842] New: r200: Culling is broken when rendering to an FBO

[Bug 57842] New: r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 1781 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

          Priority: medium
            Bug ID: 57842
          Assignee: dri-devel@lists.freedesktop.org
           Summary: r200: Culling is broken when rendering to an FBO
          Severity: normal
    Classification: Unclassified
                OS: All
          Reporter: stefandoesinger@gmx.at
          Hardware: Other
            Status: NEW
           Version: git
         Component: Drivers/DRI/r200
           Product: Mesa

Created attachment 70972
  --> https://bugs.freedesktop.org/attachment.cgi?id=70972&action=edit
Test case

When rendering to an FBO with culling enabled, the r200 driver discards all
geometry. As soon as GL_CULL_FACE is enabled, nothing is drawn. The glCullFace,
glFrontFace and primitive orientation don't seem to make a difference. When
rendering to a glx front or back buffer, culling behaves correctly.

The attached test case shows the issue. It creates an FBO, clears it with a
blue color, and then draws a front- and back facing triangle in red and green
respectively, and culls front-facing primitives. Afterwards the FBO texture is
drawn to the window. The expected result is that the green triangle shows up.
The actual result is a solid blue window.

I've had a quick look at the driver source code. r200FrontFace swaps front- and
back facing geometry because the driver renders everything upside down on FBOs.
This code doesn't seem to have any effect on this bug. What seems to trigger it
is not inverting the viewport in r200UpdateWindow(), but I don't understand
why.

The test case can be compiled with gcc fbo.c -o fbo -lGL -lglut , and started
by simply running the binary without any parameters.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 3194 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 945 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #1 from Roland Scheidegger <sroland@vmware.com> ---
Hmm this is odd I can't see anything wrong with the culling setup wrt fbos.
The viewport, window handling though seems somewhat bogus.
First, r200UpdateViewportOffset() has no callers (it is used in
radeon->vtbl.update_viewport_offset but that one seems to have no callers
neither). This function would not take the y flip into account but since it
isn't used... I'd just get rid of it but it's the only thing handling stippling
that probably needs to move somewhere else.
Also, both r200UpdateViewportOffset() and r200UpdateWindow() use the
driDrawable height for the yoffset in the viewport. I don't think that will
work for fbos, should just use the ctx->DrawBuffer height instead?
I can't see though why that would just cause failures with culling.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1727 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 517 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #2 from Roland Scheidegger <sroland@vmware.com> ---
Created attachment 71014
  --> https://bugs.freedesktop.org/attachment.cgi?id=71014&action=edit
potential fixes for viewport trouble

I had something like this in mind. Untested though and all that viewport
updating stuff seems really weird. I hope I didn't create circular calling
chains...
Might not fix this bug at all...

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1540 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 600 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #3 from Stefan Dösinger <stefandoesinger@gmx.at> ---
Created attachment 71022
  --> https://bugs.freedesktop.org/attachment.cgi?id=71022&action=edit
Fix for attachment 71014

The patch needed this small fix to compile. It seems that my gcc doesn't like
logical operations on floats, which kinda makes sense.

Otherwise I see no difference, but I haven't really tested polygon stipple or
the scissor rect on fbos yet. Clipping is still broken.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1938 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 308 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #4 from Stefan Dösinger <stefandoesinger@gmx.at> ---
(In reply to comment #3)
> Clipping is still broken.
Sorry, this should read "culling is still broken".

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1186 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 815 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #5 from smoki <smoki00790@gmail.com> ---
Created attachment 72047
  --> https://bugs.freedesktop.org/attachment.cgi?id=72047&action=edit
Myr200FrontFace

 Seems like i managed to fix it, but with dirty hack... see function in
attachment. tcl somehow is not aware of wind inverting, and i just inverted
commands, but it works that way :).

 This fbo test case works, fbotexture and tri-cull from mesa demos - with both
swtcl and tcl. What i tested, yeah Aquaria game is no more black with fbo usage
and also Supertuxkart now show icons in minimap, etc.

 Anyhow, maybe someone know how to make that better to be acceptible for mesa
and pushed in git - i don't know better :).

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1701 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 368 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #6 from Stefan Dösinger <stefandoesinger@gmx.at> ---
I'll give the patch a try after the holidays. My I'm currently 100 miles away
from my r200-equipped laptop and won't have access to it until Wednesday or
Thursday.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1166 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 669 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #7 from Rafał Mużyło <galtgendo@o2.pl> ---
Perhaps the message in this commit
(http://cgit.freedesktop.org/mesa/mesa/commit/src/mesa/drivers/dri/r200/r200_state.c?id=60e60bb3026a269fefe1cfd3312fdf3a7e4c595f)
is of note:

Tested with r300, radeon and r200 compile tested only.


Though looking at the changes of attachment 72047, perhaps the other cards
should be rechecked too - these changes look more like the changes made in that
commit to r300, so it might not be as much of a hack as the author thought.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1793 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 461 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #8 from Rafał Mużyło <galtgendo@o2.pl> ---
...minor correction: that r300 was for the non-galium driver (that was removed
awhile ago), but it seems that the logic in that commit was incorrect, so the
"hack" from that attachement seems to be a logic fix and most likely *is* valid
for radeon/radeon_state.c.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1261 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 885 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #9 from smoki <smoki00790@gmail.com> ---

 That Dänzer's commit is good, it inverts winding properly and it works good
but for swtcl only. My tries to make it work but for both swtcl and tcl, and
for Stefan's test case also.

 And it works for me both swtcl and tcl for those tests, but i am not very good
at programming ;).

 "Fixes fgl_glxgears and progs/demos/fbotexture after pressing 'c'."

 Yeah that both works too "pressing 'c'" now works. Pbuffered fgl_glxgears
works, but with -fbo switch - that never showing a texture on r200 :).

 So i've tested: fgl_glxgears, fbotexture, supertuxkart, fbo test from this
bug...
and also tri-cull to be sure no-fbo case work. And that all works for swtcl and
tcl on rv280.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1690 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 439 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #10 from smoki <smoki00790@gmail.com> ---
Created attachment 72050
  --> https://bugs.freedesktop.org/attachment.cgi?id=72050&action=edit
STK minimap icon


 He, he, i'am not programmer at all ;) but i managed to fix fogcoords, lighting
and culling on rv280 with tcl this year.

 Happy holidays :).

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1328 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 408 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

smoki <smoki00790@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
  Attachment #72050|text/plain                  |image/png
          mime type|                            |

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1052 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 629 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #11 from Rafał Mużyło <galtgendo@o2.pl> ---
Created attachment 72051
  --> https://bugs.freedesktop.org/attachment.cgi?id=72051&action=edit
attachment 72047 as patch for r200/radeon

@comment 9: your own attachment says it was not.
Anyway, attachment 72047 translated into a patch.

'cull_face = (mode == GL_CW) ? R200_FFACE_CULL_CW : R200_FFACE_CULL_CCW'
is bit of a mouthful, but atm. I've got not idea how to turn that into proper
obfuscation style of mesa sources.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 2103 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 300 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #12 from smoki <smoki00790@gmail.com> ---

 Yeah i tried your patch and it works, thanks Raphal. Someone could pushed it
in git if it's in good looking now ;).

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1066 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 1305 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #13 from Roland Scheidegger <sroland@vmware.com> ---
(In reply to comment #11)
> Created attachment 72051 [details] [review]
> attachment 72047 [details] as patch for r200/radeon
> 
> @comment 9: your own attachment says it was not.
> Anyway, attachment 72047 [details] translated into a patch.
> 
> 'cull_face = (mode == GL_CW) ? R200_FFACE_CULL_CW : R200_FFACE_CULL_CCW'
> is bit of a mouthful, but atm. I've got not idea how to turn that into
> proper obfuscation style of mesa sources.

patch looks correct to me. TCL culling is part of the tcl engine and hence
ignored if there's no TCL. But viewport transform (which is where the fbo
inversion happens) is done later, hence TCL culling is not subject to that
inversion.
FWIW it would also be possible to simply disable TCL culling, setup culling is
enough. The docs are actually saying it TCL culling is a performance
optimization with questionable quality (as the coords aren't snapped to
rasterizer subpixels). I guess it would also be possible to disable setup
culling instead (of course only with tcl) but I'm not sure if that would make
any performance compared to having enabled culling at both places.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 2902 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 397 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #14 from Stefan Dösinger <stefandoesinger@gmx.at> ---
The patch fixes the culling issues in the dx8 d3d samples as well. Smoki, can
you  mail the patch to the mesa-dev list? I could do it as well, but usually
it's better if the author does this.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1196 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 294 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #15 from smoki <smoki00790@gmail.com> ---

 Feel free to send it Stefan. Shame on me :( i don't know how to make proper
mesa patch nor how to send it :(.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1060 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 338 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

--- Comment #16 from Stefan Dösinger <stefandoesinger@gmx.at> ---
Commit it to Git and use git send-email to send it to
mesa-dev@lists.freedesktop.org . Google finds many tutorials on how to do this.

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 1197 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel

[Bug 57842] r200: Culling is broken when rendering to an FBO

[bugzilla-daemon]

[-- Attachment #1.1: Type: text/plain, Size: 578 bytes --]

https://bugs.freedesktop.org/show_bug.cgi?id=57842

Alex Deucher <agd5f@yahoo.com> changed:

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

--- Comment #17 from Alex Deucher <agd5f@yahoo.com> ---
pushed:
http://cgit.freedesktop.org/mesa/mesa/commit/?id=5bf357db89836d0f4e4b8a4cb559755d4734b81b

thanks!

-- 
You are receiving this mail because:
You are the assignee for the bug.

[-- Attachment #1.2: Type: text/html, Size: 2133 bytes --]

[-- Attachment #2: Type: text/plain, Size: 159 bytes --]

_______________________________________________
dri-devel mailing list
dri-devel@lists.freedesktop.org
http://lists.freedesktop.org/mailman/listinfo/dri-devel