Summary: | rendering artifacts with WebGL zombies | ||
---|---|---|---|
Product: | Mesa | Reporter: | Tapani Pälli <lemody> |
Component: | Drivers/DRI/i965 | Assignee: | Intel 3D Bugs Mailing List <intel-3d-bugs> |
Status: | RESOLVED NOTOURBUG | QA Contact: | Intel 3D Bugs Mailing List <intel-3d-bugs> |
Severity: | normal | ||
Priority: | medium | CC: | ciprian.tomoiaga, eero.t.tamminen, mattst88 |
Version: | git | ||
Hardware: | Other | ||
OS: | All | ||
Whiteboard: | |||
i915 platform: | i915 features: | ||
Attachments: |
screencapture
sandybridge screenshot screenshot with Mesa 13.1.0-devel (git-8015746) attachment-13021-0.html |
Description
Tapani Pälli
2014-11-28 10:21:45 UTC
Created attachment 110160 [details]
screencapture
With intel it looks the same for me, with radeonsi there is no blue flickering: https://www.youtube.com/watch?v=KwV7KFndRoM Looks like a i965 fs backend optimization bug. Running with INTEL_DEBUG=no16 removes the blue flickering for me, and disabling i965/fs optimizations fixes other misrendering. Still has the weird pink triangle though. Created attachment 112337 [details]
sandybridge screenshot
on SNB this is more visible, similar pattern though
With NIR this is *completely* broken, INTEL_DEBUG=no16 again fixes the situation. Seems fixed. Tested on HSW. Created attachment 127747 [details]
screenshot with Mesa 13.1.0-devel (git-8015746)
for me it still looks like in the attached screenshot, there's noise and the weird triangle on bottom
Created attachment 127748 [details]
attachment-13021-0.html
On what hardware?
(In reply to Matt Turner from comment #8) > Created attachment 127748 [details] > attachment-13021-0.html > > On what hardware? this is on Mesa DRI Intel(R) Haswell Desktop (0x412) I tried it on SKL GT2. With Mesa 11.2 there are small white triangles at bottom left corner. With Mesa git from few minutes ago, there's a large noisy triangle, and few smaller ones on top of each other (smaller triangles have larger noisy "pixels"). Same both with Firefox & Chromium. I wasn't able to get working apitrace trace from Firefox, but I got one from Chromium. When replaying the trace, Apitrace outputs several times following: * message: api issue 4: FBO incomplete: Unsupported HW texture/renderbuffer format attached: MESA_FORMAT_RGB_FLOAT32 * message: api issue 5: FBO incomplete: driver marked FBO as incomplete [-1] * warning: 0:1(12): warning: extension `GL_ARB_gpu_shader5' unsupported in vertex shader (In reply to Eero Tamminen from comment #10) > * warning: 0:1(12): warning: extension `GL_ARB_gpu_shader5' unsupported in > vertex shader This warning is both for vertex & fragment shaders. This extension is GL 4.0 one, the GLES one is GL_OES_gpu_shader5, and that's only in GLES 3.2 (at least according to Mesa features.txt). Neither is mentioned in WebGL extensions list: https://www.khronos.org/registry/webgl/extensions/ Hi, On Firefox 50 only, I am seeing this "GL_ARB_gpu_shader5" warning in an app that I am developing with THREEjs. However, I grep-ed for "*shader5*" and it's not used anywhere in the code (neither mine nor THREEjs). So where could this be coming from? For reference, the `zombies` rendering shows like this: http://i.imgur.com/fdhPve9.png (along with its errors) If relevant, I'm on Ubuntu 14.04.3 with : OpenGL core profile version string: 3.1 (Core Profile) Mesa 10.1.3 OpenGL core profile shading language version string: 1.40 OpenGL version string: 3.0 Mesa 10.1.3 OpenGL shading language version string: 1.30 I have a Sandy Bridge i7-2670QM We are working on this issue at StreamComputing. Variable add is uninitialized before operation add += color = org = texture2D( map, uv ); (line 266 in webgl_zombies.html). Initializing add or changing that line to add = color = org = texture2D( map, uv ); fixes the problem. (Setting glsl_zero_init=true before starting browser also fixes the problem.) Kind regards, Stream HPC Team https://streamhpc.com/ per comment #14 I'm resolving this as NOTOURBUG, thanks Jakub! |
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.