[Bug 111788] [i915] i915/vcs1-sema exceeds i915/vcs1-busy value
bugzilla-daemon at freedesktop.org
bugzilla-daemon at freedesktop.org
Tue Sep 24 08:28:29 UTC 2019
https://bugs.freedesktop.org/show_bug.cgi?id=111788
--- Comment #2 from Tvrtko Ursulin <tvrtko.ursulin at linux.intel.com> ---
85% is real, Andrey run that test with 1000Hz sampling. Note that this is with
GPU fixed at 700MHz and with default settings utilization is lower (~72% cap
but I think more variance than then fixed scenario). We probably shouldn't read
too much from this since we don't know how the other engines are looking. One
of them could easily be the limiting factor.
But it is interesting that the sampling error is largest in the 700Mhz case.
Must be some unlucky latching of batch periods vs 200Hz sampling.
I suggested to clamp sema to busy in the custom monitoring tool to start with.
Question is whether we can afford going to 1000Hz in upstream. It would require
testing for CPU usage, or maybe even going to like 500Hz would help.
--
You are receiving this mail because:
You are the assignee for the bug.
You are the QA Contact 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/20190924/ddf9efea/attachment.html>
More information about the intel-gfx-bugs
mailing list