System Environment: ---------------------------------------------- Platform: IVB desktop i7-3770K CPU @ 3.50GHz Libdrm: (master)libdrm-2.4.41-3-g303ca37e722e68900cb7eb43ddbef8069b0c711b Mesa: (master)85c2e99039394292474b1a84e3dcb2fee30a0836 Xserver:(master)xorg-server-1.13.99.901-36 Kernel: (drm-intel-nightly) 418ccc855c65e0e90f81012bbc34de20b9f45cbd Bug detailed description: ---------------------------------------------- 3DMMES2.0 Taiji/Hoverjet/Uimark performance reduce by 10% with gnome-session. It's Mesa regression. By bisected, show that 75b7e1df is the first bad commit. commit 75b7e1df139676f2456fea4d3a57cf0044d8409e Author: Ian Romanick <ian.d.romanick@intel.com> Date: Sun Jan 20 20:39:54 2013 -0500 intel: Don't expose XRGB8888 visuals any more Performance: ----------------------------------------------- 9bdf5be is 75b7e1d parents commit. 9bdf5be 75b7e1d 9bdf5be vs. 75b7e1d gnome-session 117 103 13% Reproduce steps: --------------------------------------------- 1, xinit& 2, gnome-session& 3, sed s/WIDTH/1920/g taiji.template > tmp1 sed s/HEIGHT/1080/g tmp1 > script.lua vblank_mode=0 LD_LIBRARY_PATH=lib:$LD_LIBRARY_PATH ../src/common/fm_oes_player/fm_oesplayer
The other two subcases Hoverjet and uimark also drop ~8%.
Is this with or without compositing? If these numbers are with compositing, is there a performance change without?
(In reply to comment #2) > Is this with or without compositing? If these numbers are with compositing, > is there a performance change without? This is with compositing. Taiji's FPS value as below: taiji 9bdf5be dd59918(latest) with compositing 117 103 without compositing 170 165 xinit 161 160
EGL is choosing X's ARGB visual to match the GL ARGB visual, when it is supposed to be choosing the XRGB X visual, so the compositor is doing a bunch of extra work to blend the result.
oh, this bug need to keep opening?
Created attachment 75395 [details] [review] egl: Allow 24-bit visuals for 32-bit RGBA8888 configs
Can you guys retry with this patch?
okay, this issue was fixed with above patch. please put this patch into mesa branch.
It's already on mesa master branch git-68a147e, so verify.
*** Bug 62144 has been marked as a duplicate of this bug. ***
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.