Merge remote branch 'internal/upstream' into HEAD

1 files changed, 55 insertions, 30 deletions

diff --git a/vp8/encoder/x86/variance_ssse3.c b/vp8/encoder/x86/variance_ssse3.c
index 750ae8b86..eb5d486bf 100644
--- a/vp8/encoder/x86/variance_ssse3.c
+++ b/vp8/encoder/x86/variance_ssse3.c
@@ -76,8 +76,8 @@ unsigned int vp8_sub_pixel_variance16x16_ssse3
     unsigned int *sse
 )
 {
-    int xsum0, xsum1;
-    unsigned int xxsum0, xxsum1;
+    int xsum0;
+    unsigned int xxsum0;
 
     // note we could avoid these if statements if the calling function
     // just called the appropriate functions inside.
@@ -87,14 +87,6 @@ unsigned int vp8_sub_pixel_variance16x16_ssse3
             src_ptr, src_pixels_per_line,
             dst_ptr, dst_pixels_per_line, 16,
             &xsum0, &xxsum0);
-
-        vp8_half_horiz_variance16x_h_sse2(
-            src_ptr + 8, src_pixels_per_line,
-            dst_ptr + 8, dst_pixels_per_line, 16,
-            &xsum1, &xxsum1);
-
-        xsum0 += xsum1;
-        xxsum0 += xxsum1;
     }
     else if (xoffset == 0 && yoffset == 4)
     {
@@ -102,14 +94,6 @@ unsigned int vp8_sub_pixel_variance16x16_ssse3
             src_ptr, src_pixels_per_line,
             dst_ptr, dst_pixels_per_line, 16,
             &xsum0, &xxsum0);
-
-        vp8_half_vert_variance16x_h_sse2(
-            src_ptr + 8, src_pixels_per_line,
-            dst_ptr + 8, dst_pixels_per_line, 16,
-            &xsum1, &xxsum1);
-
-        xsum0 += xsum1;
-        xxsum0 += xxsum1;
     }
     else if (xoffset == 4 && yoffset == 4)
     {
@@ -117,24 +101,65 @@ unsigned int vp8_sub_pixel_variance16x16_ssse3
             src_ptr, src_pixels_per_line,
             dst_ptr, dst_pixels_per_line, 16,
             &xsum0, &xxsum0);
+    }
+    else
+    {
+        vp8_filter_block2d_bil_var_ssse3(
+            src_ptr, src_pixels_per_line,
+            dst_ptr, dst_pixels_per_line, 16,
+            xoffset, yoffset,
+            &xsum0, &xxsum0);
+    }
 
-        vp8_half_horiz_vert_variance16x_h_sse2(
-            src_ptr + 8, src_pixels_per_line,
-            dst_ptr + 8, dst_pixels_per_line, 16,
-            &xsum1, &xxsum1);
+    *sse = xxsum0;
+    return (xxsum0 - ((xsum0 * xsum0) >> 8));
+}
 
-        xsum0 += xsum1;
-        xxsum0 += xxsum1;
+unsigned int vp8_sub_pixel_variance16x8_ssse3
+(
+    const unsigned char  *src_ptr,
+    int  src_pixels_per_line,
+    int  xoffset,
+    int  yoffset,
+    const unsigned char *dst_ptr,
+    int dst_pixels_per_line,
+    unsigned int *sse
+
+)
+{
+    int xsum0;
+    unsigned int xxsum0;
+
+    if (xoffset == 4 && yoffset == 0)
+    {
+        vp8_half_horiz_variance16x_h_sse2(
+            src_ptr, src_pixels_per_line,
+            dst_ptr, dst_pixels_per_line, 8,
+            &xsum0, &xxsum0);
+    }
+    else if (xoffset == 0 && yoffset == 4)
+    {
+        vp8_half_vert_variance16x_h_sse2(
+            src_ptr, src_pixels_per_line,
+            dst_ptr, dst_pixels_per_line, 8,
+            &xsum0, &xxsum0);
+    }
+    else if (xoffset == 4 && yoffset == 4)
+    {
+        vp8_half_horiz_vert_variance16x_h_sse2(
+            src_ptr, src_pixels_per_line,
+            dst_ptr, dst_pixels_per_line, 8,
+            &xsum0, &xxsum0);
     }
     else
     {
-      vp8_filter_block2d_bil_var_ssse3(
-          src_ptr, src_pixels_per_line,
-          dst_ptr, dst_pixels_per_line, 16,
-          xoffset, yoffset,
-          &xsum0, &xxsum0);
+        vp8_filter_block2d_bil_var_ssse3(
+            src_ptr, src_pixels_per_line,
+            dst_ptr, dst_pixels_per_line, 8,
+            xoffset, yoffset,
+            &xsum0, &xxsum0);
     }
 
     *sse = xxsum0;
-    return (xxsum0 - ((xsum0 * xsum0) >> 8));
+    return (xxsum0 - ((xsum0 * xsum0) >> 7));
 }