|Summary:||[HSW,IVB] 11.0 stable branch regressions|
|Product:||Mesa||Reporter:||Mark Janes <mark.a.janes>|
|Component:||Drivers/DRI/i965||Assignee:||Emil Velikov <emil.l.velikov>|
|Status:||RESOLVED FIXED||QA Contact:||Intel 3D Bugs Mailing List <intel-3d-bugs>|
|Priority:||medium||CC:||idr, jason, mark.a.janes|
|i915 platform:||i915 features:|
Description Mark Janes 2015-12-18 22:47:29 UTC
The series pushed to 11.0 in preparation for 11.0.8 ending in b8bac2f99b5807f24b0ed4cdaad8ad09af149ff1 had significant regressions in the conformance test suite, piglit, and dEQP. GL Conform: 13 failures Piglit: 112 failures dEQP: 26 failures The failures are hitting the following assertion: deqp-gles2: brw_fs_generator.cpp:110: brw_reg brw_reg_from_fs_reg(fs_inst*, fs_reg*): Assertion `reg->type == reg->fixed_hw_reg.type' failed.
Comment 1 Mark Janes 2015-12-18 23:57:11 UTC
bisected to 0ae22b3ebde780c2c88b5bfceaf172e311bd4742 Author: Jason Ekstrand <email@example.com> AuthorDate: Fri Dec 18 12:25:53 2015 +0000 Commit: Emil Velikov <firstname.lastname@example.org> CommitDate: Fri Dec 18 12:26:22 2015 +0000 i965/fs: Use a stride of 1 and byte offsets for UBOs Cc: "11.0" <email@example.com> Reviewed-by: Kenneth Graunke <firstname.lastname@example.org> (cherry picked from commit 13ad8d03f201a4d09bf7ab9078b00807d61dfada) [Emil Velikov] - s/const_offset_reg.ud/const_offset_reg.fixed_hw_reg.dw1.ud/ - s/brw_imm_ud/fs_reg/ Signed-off-by: Emil Velikov <email@example.com> Conflicts: src/mesa/drivers/dri/i965/brw_fs.cpp src/mesa/drivers/dri/i965/brw_fs_nir.cpp
Comment 2 Jason Ekstrand 2015-12-19 00:09:21 UTC
I think the answer here is to just revert those patches. They could be made to work, but it would take more reworking than just the trivial conflicts.
Comment 3 Kenneth Graunke 2015-12-20 02:25:21 UTC
Er, I think those were mistagged. I had hoped they would be in 11.1.0, but never intended them to hit 11.0.x. It's too invasive of a change for a stable branch...
Comment 4 Mark Janes 2015-12-24 02:06:04 UTC
Emil fixed up the branch.