Rendering *very slow* with Metro game (or likely also other GL app) with Mesa master, following message seen on console: [drm:i915_hangcheck_elapsed] *ERROR* stuck on render ring [drm:i915_set_reset_status] *ERROR* render ring hung inside bo (0x20d86000 ctx 2) at 0x20d87398 bisected to --- 8< --- commit d92f593d8776ec157ad0e7fa2ee8c9a17fd744ce Author: Eric Anholt <eric@anholt.net> Date: Thu Feb 13 21:37:50 2014 -0800 i965/fs: Use conditional sends to do FB writes on HSW+. This drops the MOVs for header setup, which are totally mis-scheduled. --- 8< --- commits before this change work fine.
(this on Haswell ULT)
Created attachment 94365 [details] error state information This is completely reproducible, also e.g. with Trine2. Both in mine and Tapani's case kernel is default Ubuntu 13.10 one. Attached is error state snapshot from this: [ 5296.924257] [drm:i915_hangcheck_elapsed] *ERROR* stuck on render ring [ 5296.924260] [drm] capturing error event; look for more information in /sys/kernel/debug/dri/0/i915_error_state [ 5302.910090] [drm:i915_hangcheck_elapsed] *ERROR* stuck on render ring
Reproduced using trine 2, spent a while trying to make a trace to look at and failed.
Reverting the guilty commit from Mesa fixes things also in latest Mesa.
commit 171ec9585f214c2dff3cfc97acebf5286bb5001d Author: Eric Anholt <eric@anholt.net> Date: Tue Mar 4 15:12:40 2014 -0800 i965: Fix predicated-send-based discards with MRT.
(BTW this was tested also with gt2 and using unigine heaven)
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.