shithub: libvpx

Download patch

ref: 8dbf0dcc67caaed0042ca97988a1fd61afb2e724
parent: db1222bf3cc3fab9e404c0abaf6e3b2dc79344b1
author: Jingning Han <jingning@google.com>
date: Thu Jul 12 04:50:31 EDT 2018

Use regular filter type for tpl model motion compensation

This slightly improves the compression performance by 0.05%.

Change-Id: Ice0b1f5e1f24a77008b093f7830e51fcd6cbfa8e

--- a/vp9/encoder/vp9_encoder.c
+++ b/vp9/encoder/vp9_encoder.c
@@ -5717,7 +5717,7 @@
   MACROBLOCK *x = &td->mb;
   MACROBLOCKD *xd = &x->e_mbd;
   int mi_row, mi_col;
-  const InterpKernel *const kernel = vp9_filter_kernels[EIGHTTAP_SHARP];
+  const InterpKernel *const kernel = vp9_filter_kernels[EIGHTTAP];
 
 #if CONFIG_VP9_HIGHBITDEPTH
   DECLARE_ALIGNED(16, uint16_t, predictor16[32 * 32 * 3]);