[Bug 99221] >2% perf drop in GfxBench T-Rex with "i965: disable loop unrolling in GLSL IR"

bugzilla-daemon at freedesktop.org bugzilla-daemon at freedesktop.org
Thu Jan 5 09:33:47 UTC 2017


https://bugs.freedesktop.org/show_bug.cgi?id=99221

--- Comment #17 from Eero Tamminen <eero.t.tamminen at intel.com> ---
The commit fixes PSPom perf on all platforms where I have data (on GEN6 & GEN7
there was no clear drop on it, nor now increase, changes affected just GEN8 &
GEN9).

What you propose should be done about T-Rex & PSPhong:
* Close this bug now, as the clear regression was fixed, or
* Keep this bug open until Curro's scheduling rework lands, and/or somebody
adds register bank conflict handling to Mesa
?

Of latter, register bank conflict handling should be clear win (especially on
older platforms), but scheduling is all about compromises, so some tests can go
down while on average things improve.

-- 
You are receiving this mail because:
You are the assignee for the bug.
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/20170105/2f3cb8c8/attachment.html>


More information about the intel-3d-bugs mailing list