On 01.11.2014 03:10, Alex Deucher wrote: > On Thu, Oct 30, 2014 at 4:36 AM, Michel Dänzer <michel at daenzer.net> wrote: >> From: Michel Dänzer <michel.daenzer at amd.com> >> >> In the words of Daniel Vetter: >> >> «I think SEARCH_BEST is pretty much always a bad idea - it rips apart >> allocations from the same execbuf, and usually those get recycled around >> the same time. Which means you'll just fragment your mm even more if you >> try to find the best hole instead of just picking one and the stuffing >> the entire execbuf into it. So imo we might as well just kill it.» >> >> Signed-off-by: Michel Dänzer <michel.daenzer at amd.com> > > Applied this series to my -next tree.
Thanks. Unfortunately, the removal of DRM_MM_SEARCH_BEST caused performance regressions, presumably due to excessive eviction from VRAM, see https://bugs.freedesktop.org/show_bug.cgi?id=84662#c49 . I can reproduce a significant bad effect with the UE4 Elemental demo on my development machine. Please drop this patch. The original patches 4 & 5 have the same effect, so they shouldn't be applied either. Maybe we can try this again once TTM is using the MM scan roaster for eviction. -- Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer
