Mesa (master): i965: Implement guardband clipping on Ivybridge.

[Kenneth Graunke]

Module: Mesa
Branch: master
Commit: 610910a66d1a97ca51ad8a003bdeadfe9563d721
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=610910a66d1a97ca51ad8a003bdeadfe9563d721

Author: Kenneth Graunke <kenneth at whitecape.org>
Date:   Fri May  4 17:06:39 2012 -0700

i965: Implement guardband clipping on Ivybridge.

Improves performance in Citybench:
- 320x240: 9.19589% +/- 0.557621%
- 1280x480: 3.90797% +/- 0.774429%

No apparent difference in OpenArena.

Signed-off-by: Kenneth Graunke <kenneth at whitecape.org>
Reviewed-by: Eric Anholt <eric at anholt.net>

---

 src/mesa/drivers/dri/i965/gen7_clip_state.c     |    1 +
 src/mesa/drivers/dri/i965/gen7_viewport_state.c |   19 ++++++++++++++-----
 2 files changed, 15 insertions(+), 5 deletions(-)

diff --git a/src/mesa/drivers/dri/i965/gen7_clip_state.c b/src/mesa/drivers/dri/i965/gen7_clip_state.c
index 5fede36..8ff31a8 100644
--- a/src/mesa/drivers/dri/i965/gen7_clip_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_clip_state.c
@@ -101,6 +101,7 @@ upload_clip_state(struct brw_context *brw)
 	     GEN6_CLIP_MODE_NORMAL |
              nonperspective_barycentric_enable_flag |
 	     GEN6_CLIP_XY_TEST |
+	     GEN6_CLIP_GB_TEST |
 	     userclip << GEN6_USER_CLIP_CLIP_DISTANCES_SHIFT |
 	     depth_clamp |
 	     provoking);
diff --git a/src/mesa/drivers/dri/i965/gen7_viewport_state.c b/src/mesa/drivers/dri/i965/gen7_viewport_state.c
index 2bcf338..7036ef6 100644
--- a/src/mesa/drivers/dri/i965/gen7_viewport_state.c
+++ b/src/mesa/drivers/dri/i965/gen7_viewport_state.c
@@ -43,11 +43,20 @@ gen7_upload_sf_clip_viewport(struct brw_context *brw)
    /* Also assign to clip.vp_offset in case something uses it. */
    brw->clip.vp_offset = brw->sf.vp_offset;
 
-   /* Disable guardband clipping (see gen6_viewport_state.c for rationale). */
-   vp->guardband.xmin = -1.0;
-   vp->guardband.xmax = 1.0;
-   vp->guardband.ymin = -1.0;
-   vp->guardband.ymax = 1.0;
+   /* According to the "Vertex X,Y Clamping and Quantization" section of the
+    * Strips and Fans documentation, Ivybridge and later don't have a maximum
+    * post-clamp delta.  However, the guardband extent must fit in [-32K, 32K)
+    * which gives us a maximum size of 64K.  Use 65000 rather than 65536 to be
+    * somewhat cautious---make the guardband slightly smaller than the maximum.
+    */
+   const float maximum_guardband_extent = 65000;
+   float gbx = maximum_guardband_extent / (float) ctx->Viewport.Width;
+   float gby = maximum_guardband_extent / (float) ctx->Viewport.Height;
+
+   vp->guardband.xmin = -gbx;
+   vp->guardband.xmax = gbx;
+   vp->guardband.ymin = -gby;
+   vp->guardband.ymax = gby;
 
    /* _NEW_BUFFERS */
    if (render_to_fbo) {