This seems to only affect HD600 (2x6), and not HD605 (3x6). List of failing tests: piglit.spec.ext_framebuffer_multisample.accuracy 6 depth_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 8 stencil_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 6 stencil_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 8 depth_resolve small piglit.spec.ext_framebuffer_multisample.accuracy 6 stencil_resolve small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 4 stencil_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 16 stencil_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 16 depth_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 2 stencil_resolve small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 6 stencil_draw small piglit.spec.ext_framebuffer_multisample.accuracy all_samples stencil_draw small piglit.spec.ext_framebuffer_multisample.accuracy 2 depth_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy all_samples depth_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy all_samples stencil_resolve small piglit.spec.ext_framebuffer_multisample.accuracy 4 depth_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy all_samples depth_draw small piglit.spec.ext_framebuffer_multisample.accuracy all_samples stencil_draw small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 4 stencil_resolve small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy 4 depth_resolve small depthstencil piglit.spec.ext_framebuffer_multisample.accuracy all_samples stencil_resolve small depthstencil Output from one of the failing tests: /tmp/build_root/m64/lib/piglit/bin/ext_framebuffer_multisample-accuracy 6 depth_draw small depthstencil -auto -fbo Pixels that should be unlit count = 126470 Perfect output Pixels that should be totally lit count = 131408 RMS error = 0.141088 The error threshold for unlit and totally lit pixels test is 0.010323 Pixels that should be partially lit count = 4266 RMS error = 0.077997 The error threshold for partially lit pixels is 0.206460
Hi Clayton, could you please clarify GPU models, marked as "HD600 (2x6), and not HD605 (3x6)? I wanna dig into this, but not sure whether I have needed configurations.
HD605: https://www.newegg.com/Product/Product.aspx?Item=N82E16856102204 HD600: https://www.newegg.com/Product/Product.aspx?Item=N82E16856102203
A common theme among these failing tests is that they all have the "small" option enabled.
Running the tests with INTEL_DEBUG=sync causes them to pass.
INTEL_DEBUG=reemit is similarly useful.(In reply to Nanley Chery from comment #4) > Running the tests with INTEL_DEBUG=sync causes them to pass. INTEL_DEBUG=reemit is similarly useful.
Hacking to driver to disable push constants also makes the tests pass.
You don't even need to re-emit everything, just re-emitting 3DSTATE_PUSH_CONSTANT_ALLOC_* fixes it.
Nanley, Kenneth, any news on this bug ? Notice that I've extenced the Release Candidates period for 18.2 in order to get room for fixing blockers from bug 107457. If you don't think this can be solved in a reasonable period of time we should target for 18.3 and warn in the release notes about this known bug. Please, let me know as soon as possible.
Additionally, which is the reason for blocking mesa-18.2 ? If this is a regression, do we have a bisect of it?
(In reply to Andrés Gómez García from comment #8) > Nanley, Kenneth, any news on this bug ? > > Notice that I've extenced the Release Candidates period for 18.2 in order to > get room for fixing blockers from bug 107457. > > If you don't think this can be solved in a reasonable period of time we > should target for 18.3 and warn in the release notes about this known bug. > > Please, let me know as soon as possible. Hi Andrés, Thank you for the extra time. We have a fix for this bug going through CI and I will send it out to the list shortly.
The fix for this bug (and mitigation for the issue) is on the list: https://patchwork.freedesktop.org/patch/245950/
v2: https://patchwork.freedesktop.org/patch/246461/
Thank you for the report. This should be fixed by: commit 4a57b997092851a2c5e170bb1b58787524006704 Author: Nanley Chery <nanley.g.chery@intel.com> Date: Wed Aug 22 10:43:32 2018 -0700 i965/gen7_urb: Re-emit PUSH_CONSTANT_ALLOC on some gen9
Looks like I actually failed to push the fix. It's upstream now and the commit is 904c2a617d86944fbdc2c955f327aacd0b3df318 .
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.