[Bug 105363] [CI][SHARDS] igt at kms_flip@flip-vs-expired-vblank - Failed assertion: gettime_us() - start < 500

[bugzilla-daemon at freedesktop.org]

https://bugs.freedesktop.org/show_bug.cgi?id=105363

--- Comment #9 from Arek Hiler <arkadiusz.hiler at intel.com> ---
Few notes:

1. The message could be improved. It seems to suggest that this is interval
between vblanks instead of the time it takes doing a wait on expired one

2. Going through recent logs the time is usually ~25us with occasional spikes
to ~150us. And there are the outliers which we fail on:

(kms_flip:1816) DEBUG: Vblank took 1456us
(kms_flip:5124) DEBUG: Vblank took 1088us
(kms_flip:1373) DEBUG: Vblank took 501us
(kms_flip:2150) DEBUG: Vblank took 842us
(kms_flip:5198) DEBUG: Vblank took 3301us

Still no idea what may cause this. It seems to be too much of a lag even for
very unfavourable scheduling. The only locking that happens on the call path is
in drm_vblank_get() which takes on a few spinlocks. I believe it's worth
investigating.

3ms of delay for 16ms frames (60fps) is a lot.

-- 
You are receiving this mail because:
You are the QA Contact for the bug.
You are the assignee for the bug.
You are on the CC list for the bug.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/intel-gfx-bugs/attachments/20190719/f343774d/attachment-0001.html>