On Fri, Apr 19, 2013 at 11:24:32AM +0200, Daniel Vetter wrote: > Hi all, > > This fixes all the bugs I've found in my various systems when using non-24bpp > modes as the first part of the series. > > And with working non-standard bpp support I've figured we can go fancy and > implemented auto-dithering if we hit an fdi bw limit. Which means that you can > now use 3-pipe pch configurations on ivb on pretty much everywhere. The only > restriction is that you need to fire up pipe C first, since without atomic > modeset pipe B will otherwise too much bw. > > One big thing here is that this will break Paulo's hsw eDP machine, specifically > the patch called "drm/i915: force bpp for eDP panels". But apparently without > that my machine here is broken ... Ideas highly welcome about how we could quirk > ourselves out of this mess. Just tested (danvet/fdi-dither 93b459d0a80024041daea5195ff6f8a576fa89bc) using depths 24 and 30 with Ironlake DP, still works. Still has the wait_for_pipe_off WARN though. -Chris -- Chris Wilson, Intel Open Source Technology Centre