Mesa (master): llvmpipe: slightly simplify build_mask

[Keith Whitwell]

Module: Mesa
Branch: master
Commit: e38d2f716381385e2aad219a3d125065ec0a01bd
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=e38d2f716381385e2aad219a3d125065ec0a01bd

Author: Keith Whitwell <keithw at vmware.com>
Date:   Tue Aug 24 23:05:57 2010 +0100

llvmpipe: slightly simplify build_mask

---

 src/gallium/drivers/llvmpipe/lp_rast_tri.c |    8 ++------
 1 files changed, 2 insertions(+), 6 deletions(-)

diff --git a/src/gallium/drivers/llvmpipe/lp_rast_tri.c b/src/gallium/drivers/llvmpipe/lp_rast_tri.c
index bdb8d13..dbaa8e0 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast_tri.c
@@ -229,13 +229,9 @@ build_mask(int c, int dcdx, int dcdy)
     */
    __m128i cstep0 = _mm_add_epi32(c0, step);
 
-   /* Scale up step for moving between quads.  This should probably
-    * be an arithmetic shift left, but there doesn't seem to be
-    * such a thing in SSE.  It's unlikely that the step value is
-    * going to be large enough to overflow across 4 pixels, though
-    * if it is that big, rendering will be incorrect anyway.
+   /* Scale up step for moving between quads.
     */
-   __m128i step4 = _mm_slli_epi32(step, 1);
+   __m128i step4 = _mm_add_epi32(step, step);
 
    /* Get values for the remaining quads:
     */