On Tue, Dec 30, 2014 at 5:54 PM, Eric Anholt <[email protected]> wrote: > I've been looking into X performance on VC4 recently. The first > obvious thing happening was that we're hitting some fallbacks in the > driver for things like GL_QUADS, so I thought "what if I use the GLES2 > paths instead?" Turns out the GLES2 paths are thoroughly broken now. > I wrote two quick patches that clean up a bunch of errors, but it's > not working yet.
What about simply not using GL_QUADS for the normal GL paths? Is using quads, vs tri's and a few extra vertices really going to be a win on some other hw? If not, avoiding quads would be a big help for freedreno too.. > The big performance win, though, is fixing copyarea to give the GL > information about the area that might be damaged by the operation, > using the scissor. Given that it's negative lines of code and not > significant on i965, I think this is a pretty good idea. I do wonder > if we don't want to just always leave scissoring on, like I did to > logic op. I'd implemented a similar thing in XA once upon a time... pretty much must-have for tilers.. BR, -R > I'm not sure if we want to slip these in to the current release, > though -- #1 is touchy enough that it might regress the normal case, > and it's not like GLES2 actually ends up working. I'd be fine with > delaying until the merge window reopens. > > _______________________________________________ > [email protected]: X.Org development > Archives: http://lists.x.org/archives/xorg-devel > Info: http://lists.x.org/mailman/listinfo/xorg-devel _______________________________________________ [email protected]: X.Org development Archives: http://lists.x.org/archives/xorg-devel Info: http://lists.x.org/mailman/listinfo/xorg-devel
