On Fri, Aug 7, 2015 at 12:03 PM, Mathias Fröhlich <mathias.froehl...@gmx.net > wrote:
> > > Ilia, > > > > On Friday, August 07, 2015 11:20:42 Ilia Mirkin wrote: > > > Yeah, I tend to agree with you -- it seems like _tnl_wakeup is meant for > > > "oh hey, I shut you down and haven't been sending you updates, but I want > > > to use you again, so please update your stuff to match reality". Does it > > > even need to be called at all? nouveau_state.c:nouveau_update_state > > > unconditionally calls _tnl_InvalidateState. i965, on the other hand, the > > > only other user of tnl_wakeup, calls it at draw time, for RenderMode != > > > GL_RENDER, and never calls _tnl_InvalidateState, so the tnl module is > > > completely out-of-date there. > > > > > > So I'm actually in favor of removing it entirely. Thoughts? > > You mean remove _tnl_wakeup/_tnl_InvalidateState entirely from the > > nouveau initialzation? That may work - I have the same > > impression, but I cannot exactly foresee what we might break. > None of the other drivers appear to do it... should be safe. I'll def test it out before pushing, of course... I've been meaning to plug a nv1x in so I can play with a couple of minor items. Ideally it'd switch to the i965 method, and only call tnl_InvalidateState when in swtnl mode (as well as calling tnl_wakeup on hwtnl -> !hwtnl transitions) but... meh. Actually it looks like SWTNL is largely unimplemented and it falls straight back to SWRAST? That's a bit unfortunate for nv04/nv05 :( -ilia
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
