Bug 17762

Summary: [rendercheck] very low performance when run Rendercheck test(composite&cacomposit)
Product: xorg Reporter: liuhaien <haien.liu>
Component: Driver/intelAssignee: Chris Wilson <chris>
Status: CLOSED WONTFIX QA Contact: Xorg Project Team <xorg-team>
Severity: enhancement    
Priority: lowest CC: shuang.he
Version: git   
Hardware: Other   
OS: Linux (All)   
Whiteboard:
i915 platform: i915 features:
Attachments:
Description Flags
xorg.0.log none

Description liuhaien 2008-09-24 20:40:08 UTC
Created attachment 19166 [details]
xorg.0.log

System Environment:
--------------------------

--Platform: q965
--Architecture(32-bit,64-bit,compatiblity): 32-bit
--2D driver: xf86-video-intel-2.5-branch
                 8408995ffbf705aa0bc09ab72c58c2e31a4b70c3
--3D driver: intel-2008-q3 branch e636f5b76bbcfef95092d21646c844c0dfe770e0
--DRM:shipped with kernel 2.6.27-rc6
--libdrm: master 2db8e0c8ef8c7a66460fceda129533b364f6418c
--Xserver: 1.5.1
--Kernel: 2.6.27-rc6

Bug detailed description:
--------------------------
start X ,then run Rendercheck test .due to very low performance,it will take more than 5 hours now  but just need 1 hour before.

Reproduce steps:
----------------
1. xinit &
2. ./rendercheck -t composite  -o src,over,overreverse,xor -f a8r8g8b8,a8,x8r8g8b8,r8g8b8,b8g8r8,r5g5b5,b5g5r5,x1r5g5b5,x1b5g5r5,r5g6b5,b5g6r5,x8b8g8r8
Comment 1 Eric Anholt 2008-09-25 14:07:28 UTC
In exchange for better cpu usage behavior when doing real work, we may have hurt runtime of this regression test.

Very low priority.
Comment 2 liuhaien 2008-10-28 00:02:56 UTC
still exists with :
Mesa:           (master)2a877411dbe35abdd8c15fb4821d9232619d89cc
Xserver:                (master)102c4dac7c521941f52652152b1660cd7f559d56
Xf86_video_intel:              
(master)87ea531c5dc5b39809395b277c330854aaaaf019
Kernel_gem:            
(drm-intel-next)43f7ccc48dd7587983cfe16de578042a7ad3a9b3
Comment 3 liuhaien 2009-03-22 22:14:30 UTC
still exists against:
Libdrm:         (master)00d8e960ca665b7f0528438331f4d0ae77fbb4cc
Mesa:(mesa_7_4_branch)b009a32bf428192fef2dc4787d25f022a472854f          
Xserver:        (server-1.6-branch)60c161545af80eb78eb790a05bde79409dfdf16e
Xf86_video_intel:       (2.7)e2465249a90b9aefe6d7a96eb56a51fde54698a0
Kernel:       (for-airlied)a2e785c32b886dd7f0289d1cf15fc14e9c81bc01
Comment 4 liuhaien 2009-05-05 20:18:09 UTC
still exists with the latest driver:
Libdrm:		(master)68103b2758029b3c1fbfcf995baa758bfd2676de
Mesa:		(mesa_7_5_branch)dd4c142e90a0cba5b445990bb522ce9199d7f565
Xserver:		(master)1b1b20d6e3e696e4437a9ef56128cde70a485f66
Xf86_video_intel:		(master)a8a771a853478e5f45f71d0eff3c4d55bf24d0ad
Kernel:       (for-linus)091438dd5668396328a3419abcbc6591159eb8d1
Comment 5 zhao jian 2010-02-24 06:31:28 UTC
Now with the newest code on master branch, on both 945GM and G45, rendercheck/composite passes but it needs more than 7 hours to finish. 
Comment 6 Chris Wilson 2010-12-01 10:34:33 UTC
As I found myself waiting upon rendercheck, I fixed one of the TODOs and reduced the number of round-trips required for result processing. So rendercheck as a whole is significantly faster. However, we still incur considerable slow-down due to fallback ping-pong and there is very little incentive to fix that (accelerating more ops is likely to cause rendercheck to slow down even more...).

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.