Created attachment 115426 [details] dmesg file ==System Environment== -------------------------- Kernel Regression: YES Non-working platforms: BDW/BSW/SKL Used GFX SW Stack: Libdrm: (master)libdrm-2.4.60-34-g0d78b37b1cac304ce5e84d1207f0a43abd29c000 Mesa: (master)d5a15a89f081bdd23237b8ba5391a2ebea60468b Xserver: (master)xorg-server-1.17.0-76-gb1029716e41e252f149b82124a149da180607c96 Xf86_video_intel: (master)2.99.917-287-g09d8e5e1b3215c0e58c2ce44048458152c523270 Cairo: (master)2cf2d8e340a325adb205baf8e4bd64e1d1858008 ==kernel== Bisect from drm-intel-next-queued branch, the first bad commit show as below: commit 9258811c961879450502f767773227f5c0d79120 71b7e54f71b899db9f8def67a0e976969384e699 Author: Daniel Vetter <daniel.vetter@ffwll.ch> AuthorDate: Tue Apr 14 17:35:19 2015 +0200 Commit: Daniel Vetter <daniel.vetter@ffwll.ch> CommitDate: Thu Apr 23 21:06:02 2015 +0200 drm/i915: Don't use atomics for pg_dirty_rings It's already protected by the bkl^Wdev->struct_mutex. While at it realign some related code. Reviewed-by: Mika Kuoppala <mika.kuoppala@intel.com> Signed-off-by: Daniel Vetter <daniel.vetter@intel.com> ==Bug detailed description== BDW Platform, synmark2_6 OglBatch6 performance dropped by 12%, OglBatch7 performance dropped by 22% SKL platform, synmark2_6 OglBatch6 performance dropped by 21%, OglBatch7 performance dropped by 24% ==Reproduce steps== 1. cd /usr/local/games/OpenGL/SynMark2 2. ./synmark2 OglBatch6 Dmesg file has been attached.
Drop on this day is visible also in other (partly) CPU bound SynMark tests (Multithread, GSCloth), potentially also in GfxBench driver test. Average performance didn't improve in any of the tests, so it's just a perf regression.
Created attachment 115451 [details] [review] drm/i915/gtt: Allocate va range only if vma is not bound
(In reply to Mika Kuoppala from comment #2) > Created attachment 115451 [details] [review] [review] > drm/i915/gtt: Allocate va range only if vma is not bound Patch pass on latest nightly kernel.
Fixed by commit 75d04a3773ecee617847de963ae4195d6aa74c28 Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> Date: Tue Apr 28 17:56:17 2015 +0300 drm/i915/gtt: Allocate va range only if vma is not bound in drm-intel-next-queued. If the problem persists with that, please reopen.
(In reply to Jani Nikula from comment #4) > Fixed by > > commit 75d04a3773ecee617847de963ae4195d6aa74c28 > Author: Mika Kuoppala <mika.kuoppala@linux.intel.com> > Date: Tue Apr 28 17:56:17 2015 +0300 > > drm/i915/gtt: Allocate va range only if vma is not bound > > in drm-intel-next-queued. If the problem persists with that, please reopen. change this bug status to verified.
Hello, this patch regresses on my G45, bug logged here: https://bugs.freedesktop.org/show_bug.cgi?id=91133
Hi! I am currently out of office 6.7-3.8. I have very limited access to mail during that time. In please contact : Tomi Sarvela in Performance Benchmark System related question, Martin Peres and Eero Tamminen with performance related questions or my manager Jani Saarinen.
Closing old verified.
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.