Bug 101232

Summary:	[BAT][SNB] igt@gem_exec_flush@basic-batch-kernel-default tests are 2-8x slower
Product:	DRI	Reporter:	Martin Peres <martin.peres>
Component:	DRM/Intel	Assignee:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Status:	CLOSED FIXED	QA Contact:	Intel GFX Bugs mailing list <intel-gfx-bugs>
Severity:	critical
Priority:	high	CC:	intel-gfx-bugs, tomi.p.sarvela
Version:	DRI git
Hardware:	Other
OS:	All
Whiteboard:	ReadyForDev
i915 platform:	SNB	i915 features:	GEM/Other

Description Martin Peres 2017-05-30 11:50:49 UTC

Starting with either CI_DRM_2668 or CI_DRM_2668, the execution time of the following tests on fi-snb-2520m increased drastically:

 - igt@gem_exec_flush@basic-batch-kernel-default-wb: 15.35-> 129.21s
 - igt@gem_exec_flush@basic-batch-kernel-default-uc: 16.71 -> 32.21s

Full logs:
 - before: https://intel-gfx-ci.01.org/CI/CI_DRM_2667/fi-snb-2520m/
 - after: https://intel-gfx-ci.01.org/CI/CI_DRM_2669/fi-snb-2520m/

Comment 1 Martin Peres 2017-05-30 12:05:08 UTC

With the latest run, the execution times went down significantly:

 - igt@gem_exec_flush@basic-batch-kernel-default-wb: 21.60s
 - igt@gem_exec_flush@basic-batch-kernel-default-uc: 19.84s

Comment 2 Chris Wilson 2017-05-30 12:12:42 UTC

intel_powerclamp.

Comment 3 Martin Peres 2017-05-30 17:28:38 UTC

(In reply to Chris Wilson from comment #2)
> intel_powerclamp.

Well spotted, but this does not explain the results of run CI_DRM_2670:

 21.60 igt@gem_exec_flush@basic-batch-kernel-default-wb
 19.84 igt@gem_exec_flush@basic-batch-kernel-default-uc

Yet, idle injection was active during the execution of these tests: https://intel-gfx-ci.01.org/CI/CI_DRM_2670/fi-snb-2520m/dmesg-during.log

But anyway, I guess we need to check if this machine is not overheating or otherwise consuming too much power (as reported by RAPL).

Comment 4 Martin Peres 2017-05-30 17:29:04 UTC

+ Tomi.

Comment 5 Martin Peres 2017-06-05 13:59:25 UTC

(In reply to Martin Peres from comment #4)
> + Tomi.

Tomi checked the temperature, and if the machine was dusty. It is not, and the temperature appears to be fine.

Any other idea?

Comment 6 Martin Peres 2017-06-05 14:06:27 UTC

Actually, the increased execution time started with CI_DRM_2665 and ended with CI_DRM_2673.

Outside of this commit range, the execution time is fine. I would thus say that whatever was the problem, it was a core problem which has been fixed already. I will thus close the bug as fixed.

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.