Created attachment 71187 [details] intel_reg_dumper after resume Upon resume, I'm seeing irregular flickering (about once every 3 seconds, and sporadically repeating) where about 140 lines miss some pixels (regular black columnar pattern). Smells like FIFO underrun, WM?
Recovers after a subsequent modeset.
s&r definitely seems to be a key component; afterwards any modeset now has the ability to produce garbage on the DP. Tried the 'disable CPT phase w/a' just in case. Now trying the FDI chicken.
Worse. Poking the FDI chicken simply causes training failures (but at least we spewed an ERROR for it.)
Created attachment 71234 [details] [review] poke fdi chicken patch For reference ...
I've just tested on my mobile ilk with 13x7 lvds panel and 19x12 external DP screen. Works like a charm, survived 10 s/r cycles without hickups ...
Created attachment 71247 [details] [review] set fdi TX enhanced framing before enabling port Random bit of lore I've found while strolling bspec. Would fit somewhat with things getting out of sync.
Pushed another branch to the for-ickle branch in my personal repo.
(In reply to comment #6) > Created attachment 71247 [details] [review] [review] > set fdi TX enhanced framing before enabling port No improvement.
(In reply to comment #7) > Pushed another branch to the for-ickle branch in my personal repo. No improvement, render latency much worse than dinq+.
Unstable PLLs? http://cgit.freedesktop.org/~danvet/drm/commit/?h=pll-limits-mess&id=64ac8695df0df5e9937da1b55ddb0e3fa37e6258
The machine is dead, I am not going to be able to test any 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.