System Environment: -------------------------- Libdrm: (master)2.4.26-12-gb317c96361f88a0a4ccb2faeff09b0476d142c68 Mesa: (master)6235846cb779bce470bc903f8cfc701bc0343d73 Xserver: (master)xorg-server-1.11.0-193- gfb84be47db7cdaff406792c08e34670e8e0cbda9 Xf86_video_intel:(master)2.16.0-156-gf1bb4ebfd8991f2f9eb9c38b9259792c11e7c86a Cairo: (master)26c9994393f590c43714ba8d799093b84dd94dc6 Libva: (master)5c7b9abb93717a2d379e5338b595951fa0c2abbe Kernel: (drm-intel-next) 64a742fac3a22f57303d8f1b7e347350a1c48254 Bug detailed description: ------------------------- There is a regression on 3D performance cases ,such as urbanterror , padman, openarena and so on, which caused by mesa . I have already bisected the first bad mesa commit:abaebcee787eeb8a89bf7a82ed4d1532fcde5e39. Reproduce steps: ------------------------- 1.xinit& 2.gnome-session 3.run wop-engine.x86_64
This seems to be: commit abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 Author: Eric Anholt <eric@anholt.net> Date: Wed Sep 21 15:20:20 2011 -0700 intel: Drop the immediate validation of the texture object in TFP. but it does not makes much sense for me why it could cause regressions. Li, what kind of regressions are you seeing?
the fps value of 3D performace cases have about 6.04% ~ 9.38% regressions on average. (In reply to comment #1) > This seems to be: > > commit abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 > Author: Eric Anholt <eric@anholt.net> > Date: Wed Sep 21 15:20:20 2011 -0700 > > intel: Drop the immediate validation of the texture object in TFP. > > but it does not makes much sense for me why it could cause regressions. > > Li, what kind of regressions are you seeing?
openarena -8.96 % urbanterror -6.04 % lightsmark -6.25 % padman -9.38 % Egypt -7.72 % PRO -8.88 % (In reply to comment #1) > This seems to be: > > commit abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 > Author: Eric Anholt <eric@anholt.net> > Date: Wed Sep 21 15:20:20 2011 -0700 > > intel: Drop the immediate validation of the texture object in TFP. > > but it does not makes much sense for me why it could cause regressions. > > Li, what kind of regressions are you seeing?
The bisect looks bad to me. In my testing of openarena performance between abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 and 4fc9a98a0e01195202cd5f3f5a78cf28c5e8843c, I see a 2.9% performance hit on Sandybridge at commit 7ec2b0d0d6b6a0f760e55ffdee0bdb385a3e900a (which was known, and a planned decision since it is neutral/helps on other chipsets), and a 1% hit at 3cc0a7be23ab603ed40d602595f673a44e079885. Did you manually test before and after the commit you mentioned before you reported these results?
(In reply to comment #4) > The bisect looks bad to me. In my testing of openarena performance between > abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 and > 4fc9a98a0e01195202cd5f3f5a78cf28c5e8843c, I see a 2.9% performance hit on > Sandybridge at commit 7ec2b0d0d6b6a0f760e55ffdee0bdb385a3e900a (which was > known, and a planned decision since it is neutral/helps on other chipsets), and > a 1% hit at 3cc0a7be23ab603ed40d602595f673a44e079885. > Did you manually test before and after the commit you mentioned before you > reported these results? Hi Eric, commit abaebcee787eeb8a89bf7a82ed4d1532fcde5e39 is the first bad commit, and now it still bad. So I think you should use the commit before it(like d430e81c3287eba4ee84ca1639a23f92bbe22c8e) as the good commit to compare with the newest commit. Can you have a try?
That was just a typo, I meant to say abaebcee787eeb8a89bf7a82ed4d1532fcde5e39~1 and didn't type ~1.
Oh, are you using a GL-based compositing manager (gnome-shell) there? If so, that would explain this.
commit 15b58d8c2233f0e67257cb907c7d90fa23c723a5 Author: Eric Anholt <eric@anholt.net> Date: Tue Nov 1 09:24:47 2011 -0700 Revert "intel: Drop the immediate validation of the texture object in TFP." This reverts commit abaebcee787eeb8a89bf7a82ed4d1532fcde5e39. The assertion I made was that "the zero-copy code in validation" would zero copy. Of course, I deleted that check back in January because the two sites that would trigger it (glTexImage() and this one) both immediately bound their mt to the object, making the other check pointless. Removes two extra blits in glx-tfp. Also fixed the Android home screen, which wasn't rendering because the extra copy broke the relationship between the texture and the eglimage. Bugzilla: https://bugs.freedesktop.org/show_bug.cgi?id=42152 Tested-by: Chad Versace <chad@chad-versace.us> Tested-by: Eugeni Dodonov <eugeni.dodonov@intel.com>
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.