Bug 85126 - [BDW] Synmark2_OglGeomPoint performance reduced ~25%
Summary: [BDW] Synmark2_OglGeomPoint performance reduced ~25%
Status: RESOLVED WONTFIX
Alias: None
Product: Mesa
Classification: Unclassified
Component: Drivers/DRI/i965 (show other bugs)
Version: unspecified
Hardware: All Linux (All)
: low normal
Assignee: Ian Romanick
QA Contact: Intel 3D Bugs Mailing List
URL:
Whiteboard:
Keywords: bisected
Depends on:
Blocks:
 
Reported: 2014-10-17 07:11 UTC by lili
Modified: 2017-09-14 22:34 UTC (History)
2 users (show)

See Also:
i915 platform:
i915 features:


Attachments

Description lili 2014-10-17 07:11:30 UTC
Environment:
-----------------------------------
Platform:BDW
Libdrm:(master)libdrm-2.4.58-4-g00847fa48b83a85b0cb882594a12ed1511f780db
Mesa:(master)b86f814afd4a8cc1d5a17308fcdd05fbb937ae93
Xserver:(master)xorg-server-1.16.0-418-g61a292adf45405641de1c522a04c148e0a152acd
Xf86_video_intel:(master)2.99.916-95-g49376ba3bddd15e62df55928198160d9970dc1c6
Cairo:(master)ab2a7b61d59258b71e1c3256edf2f42ea6cc1eb2
Libva:(master)ccd93de5a707e92a629cccd595757c8d436fa3cc
Libva_intel_driver:(master)f11176415ec26eb5960ba6841d2d9c22f2cabc60
Kernel:(drm-intel-nightly)ead7d71357628dc1231535886ec09853e9164d6d

Bug detailed description:
---------------------------------------------
Synmark2_OglGeomPoint performance reduced ~25% on BDW.
This problem exist both on gnome-session and Raw X.
It's Mesa regression,bisect result first bad commit is:

commit b7844d12487dbac702ecd734faeffa6fe1a61a83
Author:     Kenneth Graunke <kenneth@whitecape.org>
AuthorDate: Tue Sep 30 01:15:56 2014 -0700
Commit:     Kenneth Graunke <kenneth@whitecape.org>
CommitDate: Thu Oct 9 15:03:32 2014 +0200

    i965: Use BDW_MOCS_PTE for renderbuffers.

Reproduce steps:
---------------------------------------------
1.	xinit&
2.	./synmark2 OglGeomPoint
Comment 1 Eero Tamminen 2014-10-23 14:54:50 UTC
The commit itself is fine, it improves performance in *many* tests significantly, to what their performance should have been.

We'll still want to know why only this test [1] was affected by caching mode setting, so bug can be kept open.  However, GL_POINTS is less important test and its performance is currently good enough, so setting bug importance to lower.

[1] Were any other tests affected negatively (>10%)?
Comment 2 lili 2014-10-24 02:30:35 UTC
(In reply to Eero Tamminen from comment #1)
> The commit itself is fine, it improves performance in *many* tests
> significantly, to what their performance should have been.

We'll still want
> to know why only this test [1] was affected by caching mode setting, so bug
> can be kept open.  However, GL_POINTS is less important test and its
> performance is currently good enough, so setting bug importance to lower.
> [1] Were any other tests affected negatively (>10%)?

According to our test results , only for this case its performance is down 25%. The others are positively affected.
Comment 3 shuo.wang 2014-11-26 02:55:42 UTC
This bug is also exist on Mesa10.4RC1


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.