Mesa (lp-binning): llvmpipe: pass mask as a linear encoding of the 4x4 block

[Keith Whitwell]

Module: Mesa
Branch: lp-binning
Commit: 7b116e13a2aa28a699e30c907c1b1ae5e04cab28
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=7b116e13a2aa28a699e30c907c1b1ae5e04cab28

Author: Keith Whitwell <keithw at vmware.com>
Date:   Tue Oct 20 03:17:17 2009 +0100

llvmpipe: pass mask as a linear encoding of the 4x4 block

---

 src/gallium/drivers/llvmpipe/lp_rast.c      |   41 +++++++++++++++++++++++----
 src/gallium/drivers/llvmpipe/lp_setup_tri.c |   15 +++------
 2 files changed, 40 insertions(+), 16 deletions(-)

diff --git a/src/gallium/drivers/llvmpipe/lp_rast.c b/src/gallium/drivers/llvmpipe/lp_rast.c
index 39fb8cd..6fd6acc 100644
--- a/src/gallium/drivers/llvmpipe/lp_rast.c
+++ b/src/gallium/drivers/llvmpipe/lp_rast.c
@@ -216,16 +216,45 @@ void lp_rast_shade_quads( struct lp_rasterizer *rast,
    struct lp_rast_tile *tile = &rast->tile;
    void *color;
    void *depth;
-   uint32_t ALIGN16_ATTRIB masks[16];
-   unsigned ix, iy, i;
+   uint32_t ALIGN16_ATTRIB masks[2][2][2][2];
+   unsigned ix, iy;
 
    /* Sanity checks */
    assert(x % TILE_VECTOR_WIDTH == 0);
    assert(y % TILE_VECTOR_HEIGHT == 0);
 
-   /* mask */
-   for (i = 0; i < 16; ++i)
-      masks[i] = mask & (1 << i) ? ~0 : 0;
+   /* mask: the rasterizer wants to treat pixels in 4x4 blocks, but
+    * the pixel shader wants to swizzle them into 4 2x2 quads.
+    * 
+    * Additionally, the pixel shader wants masks as full dword ~0,
+    * while the rasterizer wants to pack per-pixel bits tightly.
+    */
+#if 0
+   unsigned qx, qy;
+   for (qy = 0; qy < 2; ++qy)
+      for (qx = 0; qx < 2; ++qx)
+	 for (iy = 0; iy < 2; ++iy)
+	    for (ix = 0; ix < 2; ++ix)
+	       masks[qy][qx][iy][ix] = mask & (1 << (qy*8+iy*4+qx*2+ix)) ? ~0 : 0;
+#else
+   masks[0][0][0][0] = mask & (1 << (0*8+0*4+0*2+0)) ? ~0 : 0;
+   masks[0][0][0][1] = mask & (1 << (0*8+0*4+0*2+1)) ? ~0 : 0;
+   masks[0][0][1][0] = mask & (1 << (0*8+1*4+0*2+0)) ? ~0 : 0;
+   masks[0][0][1][1] = mask & (1 << (0*8+1*4+0*2+1)) ? ~0 : 0;
+   masks[0][1][0][0] = mask & (1 << (0*8+0*4+1*2+0)) ? ~0 : 0;
+   masks[0][1][0][1] = mask & (1 << (0*8+0*4+1*2+1)) ? ~0 : 0;
+   masks[0][1][1][0] = mask & (1 << (0*8+1*4+1*2+0)) ? ~0 : 0;
+   masks[0][1][1][1] = mask & (1 << (0*8+1*4+1*2+1)) ? ~0 : 0;
+
+   masks[1][0][0][0] = mask & (1 << (1*8+0*4+0*2+0)) ? ~0 : 0;
+   masks[1][0][0][1] = mask & (1 << (1*8+0*4+0*2+1)) ? ~0 : 0;
+   masks[1][0][1][0] = mask & (1 << (1*8+1*4+0*2+0)) ? ~0 : 0;
+   masks[1][0][1][1] = mask & (1 << (1*8+1*4+0*2+1)) ? ~0 : 0;
+   masks[1][1][0][0] = mask & (1 << (1*8+0*4+1*2+0)) ? ~0 : 0;
+   masks[1][1][0][1] = mask & (1 << (1*8+0*4+1*2+1)) ? ~0 : 0;
+   masks[1][1][1][0] = mask & (1 << (1*8+1*4+1*2+0)) ? ~0 : 0;
+   masks[1][1][1][1] = mask & (1 << (1*8+1*4+1*2+1)) ? ~0 : 0;
+#endif
 
    ix = x % TILE_SIZE;
    iy = y % TILE_SIZE;
@@ -251,7 +280,7 @@ void lp_rast_shade_quads( struct lp_rasterizer *rast,
                         inputs->a0,
                         inputs->dadx,
                         inputs->dady,
-                        &masks[0],
+                        &masks[0][0][0][0],
                         color,
                         depth);
 #else
diff --git a/src/gallium/drivers/llvmpipe/lp_setup_tri.c b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
index a5a0407..cf8643f 100644
--- a/src/gallium/drivers/llvmpipe/lp_setup_tri.c
+++ b/src/gallium/drivers/llvmpipe/lp_setup_tri.c
@@ -369,18 +369,13 @@ do_triangle_ccw(struct setup_context *setup,
       int ystep3 = tri->dx31;
       
       int ix, iy;
-      int qx, qy;
       int i = 0;
       
-      for (qy = 0; qy < 4; qy += 2) {
-	 for (qx = 0; qx < 4; qx += 2) {
-	    for (iy = 0; iy < 2; iy++) {
-	       for (ix = 0; ix < 2; ix++, i++) {
-		  tri->step[0][i] = (xstep1 * (qx+ix)) + (ystep1 * (qy+iy));
-		  tri->step[1][i] = (xstep2 * (qx+ix)) + (ystep2 * (qy+iy));
-		  tri->step[2][i] = (xstep3 * (qx+ix)) + (ystep3 * (qy+iy));
-	       }
-	    }
+      for (iy = 0; iy < 4; iy++) {
+	 for (ix = 0; ix < 4; ix++, i++) {
+	    tri->step[0][i] = xstep1 * ix + ystep1 * iy;
+	    tri->step[1][i] = xstep2 * ix + ystep2 * iy;
+	    tri->step[2][i] = xstep3 * ix + ystep3 * iy;
 	 }
       }
    }