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

[bugzilla-daemon at freedesktop.org]

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

--- Comment #8 from Eero Tamminen <eero.t.tamminen at intel.com> ---
(In reply to Timothy Arceri from comment #7)
> I've attached dumps from unrolling the loops with the different passes. It
> doesn't look like there is any easy solution, both have the same instruction
> count it looks like the result comes down to scheduling.

NIR version schedules the second last "send" worse, and it has more register
bank conflicts with "mad" instructions.

Is it possible that the patch would affect anything else besides shaders with
(non-dead) loops?

-- 
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/20161230/34cb0a9f/attachment.html>