Fix vmx_combine_vmx_atop_ca()

It didn't compute the mask correct before.

diff --git a/pixman/pixman-vmx.c b/pixman/pixman-vmx.c
index 0038752..fbea24e 100644 (file)
--- a/pixman/pixman-vmx.c
+++ b/pixman/pixman-vmx.c
@@ -1215,7 +1215,7 @@ vmx_combine_over_reverse_ca (pixman_implementation_t *imp,
     vector unsigned int vdest, vsrc, vmask;
     vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
        dest_mask, mask_mask, src_mask, store_mask;
-    
+
     COMPUTE_SHIFT_MASKC (dest, src, mask);
 
     /* printf("%s\n",__PRETTY_FUNCTION__); */
@@ -1421,30 +1421,32 @@ vmx_combine_atop_ca (pixman_implementation_t *imp,
                      int                      width)
 {
     int i;
-    vector unsigned int vdest, vsrc, vmask;
+    vector unsigned int vdest, vsrc, vmask, vsrca;
     vector unsigned char tmp1, tmp2, tmp3, tmp4, edges,
        dest_mask, mask_mask, src_mask, store_mask;
-    
+
     COMPUTE_SHIFT_MASKC (dest, src, mask);
-    
+
     /* printf ("%s\n",__PRETTY_FUNCTION__); */
     for (i = width / 4; i > 0; i--)
     {
-       
        LOAD_VECTORSC (dest, src, mask);
-       
-       vdest = pix_add_mul (pix_multiply (vsrc, vmask), splat_alpha (vdest),
-                            vdest,
-                            negate (pix_multiply (vmask,
-                                                  splat_alpha (vmask))));
-       
+
+       vsrca = splat_alpha (vsrc);
+
+       vsrc = pix_multiply (vsrc, vmask);
+       vmask = pix_multiply (vmask, vsrca);
+
+       vdest = pix_add_mul (vsrc, splat_alpha (negate (vdest)),
+                            negate (vmask), vdest);
+
        STORE_VECTOR (dest);
-       
+
        src += 4;
        dest += 4;
        mask += 4;
     }
-    
+
     for (i = width % 4; --i >= 0;)
     {
        uint32_t a = mask[i];
@@ -1452,7 +1454,7 @@ vmx_combine_atop_ca (pixman_implementation_t *imp,
        uint32_t d = dest[i];
        uint32_t sa = ALPHA_8 (s);
        uint32_t da = ALPHA_8 (d);
-       
+
        UN8x4_MUL_UN8x4 (s, a);
        UN8x4_MUL_UN8 (a, sa);
        UN8x4_MUL_UN8x4_ADD_UN8x4_MUL_UN8 (d, ~a, s, da);