[Bug 110432] dEQP-VK tests running very long time

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=110432

--- Comment #5 from Tapani Pälli <lemody at gmail.com> ---
(In reply to Ian Romanick from comment #4)
> I wasn't able to reproduce 38:43, but ~2 seconds -> ~2 minutes is a pretty
> big jump.  What result do you get at tip of 18.3?

I made some new measurements, this time also used NIR_VALIDATE=0 and running
the test with 'time' on KBL laptop. Here are results:

18.3.6: real 4m9.892s
19.0.0: real 5m15.113s
19.0.0 with "intel/compiler: More peephole select" patch: real 0m1.842s
HEAD: I gave up at: real 30m27.735s
HEAD with "opt_simplify_bcsel_of_phi" commented out: real 0m1.859s

I also checked Mesa CI time usage, and it seems CI spends a lot of time in
particular with these tests, one can observer the differences between test
groups here:

https://mesa-ci.01.org/tpalli/builds/747/group/e706cc2d56fffd6fe11e94c430476ef0

There is something very specific happening with
'8bit_storage.8struct_to_32struct' versus other tests in
'dEQP-VK.spirv_assembly.instruction.graphics' group. I'll continue digging!

BTW I tested 19.0.0 tag because that is what Android QA are testing against.
They reported also that bringing back 'intel/compiler: More peephole select'
(be it buggy or not..) 'fixes' following tests:

--- 8< ---
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.8struct_to_32struct#storage_buffer_sint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.8struct_to_32struct#storage_buffer_uint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.32struct_to_8struct#storage_buffer_sint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.32struct_to_8struct#storage_buffer_uint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.32struct_to_8struct#uniform_sint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.8bit_storage.32struct_to_8struct#uniform_uint_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.uniform_32struct_to_16struct#uniform_buffer_block_geom
       
dEQP-VK.spirv_assembly.instruction.graphics.16bit_storage.uniform_32struct_to_16struct#uniform_geom

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-3d-bugs/attachments/20190417/a6c0ca34/attachment.html>