From: Eric A. <et...@lc...> - 2004-10-10 19:56:21
|
On Sun, 2004-10-10 at 10:13, Andreas Stenglein wrote: > Am 2004.10.10 11:14:11 +0200 schrieb(en) Eric Anholt: > > > > http://pdx.freedesktop.org/~anholt/dri/r200-projtex-6.diff > > > > #4 had broken nontcl quite significantly. I'm thinking I just stomped > > some of my changes with another editor window at some point. Things > > seem decent again with the link above, but tcl doom is still in bad > > shape. > > > > Eric, > > could you have a look at > https://freedesktop.org/bugzilla/show_bug.cgi?id=1461 > > there is some code for swtcl to detect "transition" from > 3 component coord STR <-> STQ which might happen in the rare case > if TMUx was used for a cubemap and then for projective texturing > and the coord components of the other TMUs didnt change, too. > > At the moment projective texturing in non-tcl mode can't work because > the q-coord doesn't get emitted ever, its always the r-coord. > > > (Maybe the "global" projtex variable from r200_context.h > should get inized somewhere) My patch should be basically equivalent to that one, I think, except that I'm still emitting 4 texcoords instead of 3. While 3 would be a small savings, I wonder if that small savings matters compared to having to handle more state switching (switching RE_CNTL for moving from TCL to non-) to do it that way. However, I haven't tested that one, primarily because it was attached inline rather than as a downloadable attachment. The question would be if it actually made projtex work correctly (so that there were no extra corners), versus just getting the scale right. -- Eric Anholt et...@lc... http://people.freebsd.org/~anholt/ anholt@FreeBSD.org |