On 04.03.2015 07:07, Alex Deucher wrote: > On Tue, Mar 3, 2015 at 3:50 PM, Tormod Volden <[email protected]> wrote: >> On Tue, Mar 3, 2015 at 9:21 PM, Alex Deucher wrote: >>> On Tue, Mar 3, 2015 at 2:21 PM, Tormod Volden wrote: >>>> On Tue, Mar 3, 2015 at 3:18 AM, Michel Dänzer wrote: >>>>> About patch 1, since mgaDownloadFromScreen doesn't do anything fancier >>>>> than memcpy, you can just remove it altogether and leave >>>>> pExa->DownloadFromScreen NULL. EXA then does basically the same thing >>>>> mgaDownloadFromScreen does. >>>> >>>> Yes, Connor also suggested that. However, I would like in any case to >>>> correct the old code first. Does the "general" EXA also do the >>>> equivalent to QUIESCE_DMA()? >>>> >>> >>> It calls waitMarker before doing sw ops if that is what you are asking. >> >> Thanks. I took a look, they do different things, and QUIESCE_DMA() >> goes into mga_dri.c and places sprinkled with "FIXME what about EXA?". >> Anyway, I think I leave further changes to someone who can test :) > > If anything, it's likely to be more correct if you let EXA do it if > there are a bunch of fixme's in the QUIESCE_DMA code.
Yep. If that doesn't work, mgaWaitMarker is broken (which will affect other software rendering paths as well) and needs to be fixed. -- Earthling Michel Dänzer | http://www.amd.com Libre software enthusiast | Mesa and X developer _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
