|Summary:||[SKL] [regression] piglit compressed textures tests fail with kernel upgrade|
|Product:||Mesa||Reporter:||Mark Janes <mark.a.janes>|
|Component:||Drivers/DRI/i965||Assignee:||Nanley Chery <nanleychery>|
|Status:||CLOSED FIXED||QA Contact:||Intel 3D Bugs Mailing List <intel-3d-bugs>|
|i915 platform:||i915 features:|
Description Mark Janes 2015-09-08 23:31:36 UTC
When the SKL kernels were upgraded, the following piglit tests regressed: piglit.spec.ext_texture_compression_s3tc.getteximage-targets cube_array s3tc piglit.spec.arb_get_texture_sub_image.arb_get_texture_sub_image-getcompressed piglit.spec.arb_texture_cube_map_array.fbo-generatemipmap-cubemap array s3tc_dxt1 Failing kernel version: 38090ed drm-intel-nightly: 2015y-09m-02d-15h-24m-23s UTC integration manifest The passing kernel version was a drm nightly 4.1.0 kernel pulled around June 7.
Comment 1 Ben Widawsky 2015-10-23 15:07:01 UTC
Mark, why did you assign this to Ian if the kernel update made the failure happen?
Comment 2 Mark Janes 2015-10-23 16:34:04 UTC
Ian gets bugs by default. If I know who to assign a bug to, then I assign it directly. Who are the candidates to get this bug?
Comment 3 Ben Widawsky 2015-10-23 16:36:35 UTC
On Fri, Oct 23, 2015 at 04:34:04PM +0000, email@example.com wrote: > https://bugs.freedesktop.org/show_bug.cgi?id=91927 > > --- Comment #2 from Mark Janes <firstname.lastname@example.org> --- > Ian gets bugs by default. If I know who to assign a bug to, then I assign it > directly. Who are the candidates to get this bug? > Kernel bugs should be assigned to Intel GFX Bugs mailing list by default.
Comment 4 Nanley Chery 2015-10-28 00:30:49 UTC
(In reply to Mark Janes from comment #0) > When the SKL kernels were upgraded, the following piglit tests regressed: > > piglit.spec.ext_texture_compression_s3tc.getteximage-targets cube_array s3tc > piglit.spec.arb_get_texture_sub_image.arb_get_texture_sub_image- > getcompressed > piglit.spec.arb_texture_cube_map_array.fbo-generatemipmap-cubemap array > s3tc_dxt1 > > Failing kernel version: > 38090ed drm-intel-nightly: 2015y-09m-02d-15h-24m-23s UTC integration manifest > > The passing kernel version was a drm nightly 4.1.0 kernel pulled around June > 7. Was mesa updated as well? Neil has pointed out that some of the ASTC patches may have caused these regressions.
Comment 5 Mark Janes 2015-10-28 16:23:30 UTC
When I wrote this bug up, I was upgrading SKL kernels and attributed the failure to that change. However, now that I am testing SKL on a regular basis, I can see that these tests regressed in Mesa since 11.0. I bisected the test, and found that it regressed when ASTC was enabled for SKL: Author: Nanley Chery <email@example.com> AuthorDate: Tue May 19 12:28:20 2015 -0700 Commit: Nanley Chery <firstname.lastname@example.org> CommitDate: Mon Aug 31 17:29:36 2015 -0700 i965: advertise ASTC support for Skylake v2: remove OES ASTC extension reference. Reviewed-by: Anuj Phogat <email@example.com> Signed-off-by: Nanley Chery <firstname.lastname@example.org>
Comment 6 Nanley Chery 2015-10-28 16:51:05 UTC
I've submitted a patch to the mailing list which fixes this issue: http://lists.freedesktop.org/archives/mesa-dev/2015-October/098524.html
Comment 7 Neil Roberts 2015-12-17 18:35:03 UTC
I assume this bug is now fixed because these tests are passing and Nanley's patch has been merged: http://cgit.freedesktop.org/mesa/mesa/commit/?id=d1212abf505a468c9947a66dbf