Mesa (master): softpipe: fix Z interpolation invariance bug

[Brian Paul]

Module: Mesa
Branch: master
Commit: a5f2010735b6af469f57c6c254af9603bb9db18d
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=a5f2010735b6af469f57c6c254af9603bb9db18d

Author: Brian Paul <brianp at vmware.com>
Date:   Tue Oct  4 15:01:05 2011 -0600

softpipe: fix Z interpolation invariance bug

We want quad/pixel Z values to be interpolated exactly the same for
multi-pass algorithms.  Because of how the optimized Z-test code is
written, we can't cull the first quad in a run even if it's totally
killed.  See the comment for more info.

NOTE: This is a candidate for the 7.11 branch.

---

 src/gallium/drivers/softpipe/sp_quad_fs.c |   11 ++++++++++-
 1 files changed, 10 insertions(+), 1 deletions(-)

diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index d74d6d4..7b08cd0 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -118,7 +118,16 @@ shade_quads(struct quad_stage *qs,
    machine->InterpCoefs = quads[0]->coef;
 
    for (i = 0; i < nr; i++) {
-      if (!shade_quad(qs, quads[i]))
+      /* Only omit this quad from the output list if all the fragments
+       * are killed _AND_ it's not the first quad in the list.
+       * The first quad is special in the (optimized) depth-testing code:
+       * the quads' Z coordinates are step-wise interpolated with respect
+       * to the first quad in the list.
+       * For multi-pass algorithms we need to produce exactly the same
+       * Z values in each pass.  If interpolation starts with different quads
+       * we can get different Z values for the same (x,y).
+       */
+      if (!shade_quad(qs, quads[i]) && i > 0)
          continue; /* quad totally culled/killed */
 
       if (/*do_coverage*/ 0)