Bug 104981

Summary:	[BSW] dEQP test run times > 30 seconds
Product:	Mesa	Reporter:	Clayton Craft <clayton.a.craft>
Component:	Drivers/DRI/i965	Assignee:	Intel 3D Bugs Mailing List <intel-3d-bugs>
Status:	RESOLVED MOVED	QA Contact:	Intel 3D Bugs Mailing List <intel-3d-bugs>
Severity:	normal
Priority:	medium	CC:	chadversary, lemody
Version:	git
Hardware:	Other
OS:	All
Whiteboard:
i915 platform:		i915 features:

Description Clayton Craft 2018-02-06 23:24:50 UTC

A number of deqp tests take greater than 30 seconds to execute on BSW. A number of these are just barely over 30 seconds (the ones at 30 sec below are a fraction of a second over), but some are quite a bit over (e.g. 54 sec)

dEQP-GLES31.functional.shaders.builtin_functions.precision.faceforward.highp_compute.vec4       30 sec
dEQP-GLES31.functional.shaders.builtin_functions.precision.faceforward.lowp_compute.vec4        38 sec
dEQP-GLES31.functional.shaders.builtin_functions.precision.refract.highp_compute.vec4   35 sec
dEQP-GLES31.functional.shaders.builtin_functions.precision.refract.mediump_compute.vec4 36 sec
dEQP-GLES31.functional.ssbo.layout.random.all_shared_buffer.5   46 sec
dEQP-GLES31.functional.shaders.builtin_functions.precision.faceforward.mediump_compute.vec4     31 sec
dEQP-GLES31.functional.shaders.builtin_functions.precision.refract.lowp_compute.vec4    44 sec
dEQP-GLES31.functional.ssbo.layout.random.all_shared_buffer.23  54 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.faceforward.lowp_vertex.vec4  37 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.mediump_vertex.vec4   35 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.highp_fragment.vec4   34 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.mediump_fragment.vec4 35 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.faceforward.mediump_vertex.vec4       30 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.highp_vertex.vec4     34 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.faceforward.mediump_fragment.vec4     30 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.lowp_vertex.vec4      43 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.faceforward.lowp_fragment.vec4        37 sec
dEQP-GLES3.functional.shaders.builtin_functions.precision.refract.lowp_fragment.vec4    44 sec

Comment 1 Andrii K 2019-08-07 10:42:19 UTC

If tests were compiled in release mode, difference in speed was equal to difference between architectures, so seems like problem is BSW is just slow CPU but  some tests (marked red in table) are much slower because they spend a lot of time in generating test cases.

Tests profiling results: https://docs.google.com/spreadsheets/d/1VLpkSNszQ2ELSROFfDULCvfiV7fVXIs2XphsbQtTzcM/edit?usp=sharing

Comment 2 Mark Janes 2019-08-07 21:26:54 UTC

My recollection is that this is a problem because Google's qualification for android and chromeos drivers limits the run time of tests.

Andrii is correct that BSW is just slow, and there's not a lot we can do about it.  We do need to identify patches that increase the run-time for tests substantially, especially over the 30s threshold.  It may be that a compiler optimization pass hurts the run-time, for example.

Intel's Mesa CI runs with a debug build, which has additional time consuming verification steps.

Comment 3 GitLab Migration User 2019-09-25 19:08:32 UTC

-- GitLab Migration Automatic Message --

This bug has been migrated to freedesktop.org's GitLab instance and has been closed from further activity.

You can subscribe and participate further through the new bug through this link to our GitLab instance: https://gitlab.freedesktop.org/mesa/mesa/issues/1687.

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.