Merge "Fix incorrect 32x64, 16x32 subpel variance" into experimental

vp9/encoder/vp9_variance_c.c

diff --git a/vp9/encoder/vp9_variance_c.c b/vp9/encoder/vp9_variance_c.c
index f7916b4..c4c70df 100644
--- a/vp9/encoder/vp9_variance_c.c
+++ b/vp9/encoder/vp9_variance_c.c
@@ -87,9 +87,9 @@
 
   var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
                                     1, 65, 32, hfilter);
-  var_filter_block2d_bil_second_pass(fdata3, temp2, 64, 64, 64, 32, vfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 64, 32, vfilter);
 
-  return vp9_variance32x64_c(temp2, 64, dst_ptr, dst_pixels_per_line, sse);
+  return vp9_variance32x64_c(temp2, 32, dst_ptr, dst_pixels_per_line, sse);
 }
 
 unsigned int vp9_variance32x16_c(const uint8_t *src_ptr,
@@ -155,9 +155,9 @@
 
   var_filter_block2d_bil_first_pass(src_ptr, fdata3, src_pixels_per_line,
                                     1, 33, 16, hfilter);
-  var_filter_block2d_bil_second_pass(fdata3, temp2, 32, 32, 32, 16, vfilter);
+  var_filter_block2d_bil_second_pass(fdata3, temp2, 16, 16, 32, 16, vfilter);
 
-  return vp9_variance16x32_c(temp2, 32, dst_ptr, dst_pixels_per_line, sse);
+  return vp9_variance16x32_c(temp2, 16, dst_ptr, dst_pixels_per_line, sse);
 }
 
 unsigned int vp9_variance64x64_c(const uint8_t *src_ptr,