| Summary: | [HSW, regression] transform feedback fails on Linux 4.8 | ||
|---|---|---|---|
| Product: | Mesa | Reporter: | Mark Janes <mark.a.janes> |
| Component: | Drivers/DRI/i965 | Assignee: | Kenneth Graunke <kenneth> |
| Status: | RESOLVED FIXED | QA Contact: | Intel 3D Bugs Mailing List <intel-3d-bugs> |
| Severity: | normal | ||
| Priority: | medium | CC: | intel-gfx-bugs, mark.a.janes |
| Version: | git | ||
| Hardware: | x86-64 (AMD64) | ||
| OS: | Linux (All) | ||
| Whiteboard: | |||
| i915 platform: | HSW | i915 features: | |
In my case it was not intermittent. I was also able to reproduce with piglit's arb_transform_feedback2-draw-auto test. arb_transform_feedback2-draw-auto bisects to 6761d0a184efe5a735dfcd9c9f7ac08b2d44d657 is the first bad commit commit 6761d0a184efe5a735dfcd9c9f7ac08b2d44d657 Author: Kenneth Graunke <kenneth@whitecape.org> Date: Fri May 6 08:50:14 2016 +0100 drm/i915: Allow MI_LOAD_REGISTER_REG between whitelisted registers. Allowing register copies where the source and destination are both whitelisted should be safe, and is useful. For example, Mesa uses this to load the command streamer math registers with data from the pipeline statistics counters. Just setting cmd_parser.version=6 is enough to make the test pass, suggesting that this is a very different bug to ES3-CTS.functional.transform_feedback.array_element.interleaved.lines.lowp_vec2 Missing flush before saving SOL_OFFSET in hsw_pause_transform_feedback(). Thanks Chris! That was indeed the problem. Patch on the list: https://lists.freedesktop.org/archives/mesa-dev/2017-January/139754.html Fixed by: commit 2138347a45fa6dad1934b1c58a9e7d7f53194828 Author: Kenneth Graunke <kenneth@whitecape.org> Date: Fri Jan 6 00:09:53 2017 -0800 i965: Properly flush in hsw_pause_transform_feedback(). Fixes a number of transform feedback tests when run with Linux 4.8, which allows us to use the MI_LOAD_REGISTER_REG command, at which point we started using this new broken path. ES3-CTS.functional.transform_feedback.array_element.interleaved.lines.* and Piglit's arb_transform_feedback2/draw-auto are both fixed by this patch, for example. Thanks to Chris Wilson for catching this mistake! Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=99030 Signed-off-by: Kenneth Graunke <kenneth@whitecape.org> Reviewed-by: Anuj Phogat <anuj.phogat@gmail.com> |
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.
dEQP / GLES CTS transform feedback tests fail intermittently on HSW with upstream Linux 4.8: ES3-CTS.functional.transform_feedback.array_element.interleaved.lines.lowp_vec2 Standard Output Transform feedback varyings: { v_varA[1], v_varB[0], v_varB[3] } Transform feedback varyings reported by compiler: GL_TRANSFORM_FEEDBACK_VARYINGS = 3 v_varA[1]: vec2[1] v_varB[0]: vec2[1] v_varB[3]: vec2[1] Tests pass reliably on Linux 4.7 Bisection was proposed to be done on EZBench in Intel's bug tracker: https://jira01.devtools.intel.com/browse/VIZ-8798 However, since that hasn't been done, perhaps someone else should find and fix the issue.