shithub: libvpx

Info  •  Files  •  Log  •  Branches

mcomp: Remove an obsolete comment.

This was copied over from VP8. VP9 doesn't seem to do this buffer copy.

Change-Id: I28a8bbf0503a7f99b2cb60620ab3674adde863bb

--- a/vp10/encoder/mcomp.c

+++ b/vp10/encoder/mcomp.c

@@ -145,16 +145,6 @@

   cfg->searches_per_step = 8;

 }

-/*

- * To avoid the penalty for crossing cache-line read, preload the reference

- * area in a small buffer, which is aligned to make sure there won't be crossing

- * cache-line read while reading from this buffer. This reduced the cpu

- * cycles spent on reading ref data in sub-pixel filter functions.

- * TODO: Currently, since sub-pixel search range here is -3 ~ 3, copy 22 rows x

- * 32 cols area that is enough for 16x16 macroblock. Later, for SPLITMV, we

- * could reduce the area.

- */

-

 /* estimated cost of a motion vector (r,c) */

 #define MVC(r, c)                                       \

     (mvcost ?                                           \

--- a/vp9/encoder/vp9_mcomp.c

+++ b/vp9/encoder/vp9_mcomp.c

@@ -144,16 +144,6 @@

   cfg->total_steps = ss_count / cfg->searches_per_step;

 }

-/*

- * To avoid the penalty for crossing cache-line read, preload the reference

- * area in a small buffer, which is aligned to make sure there won't be crossing

- * cache-line read while reading from this buffer. This reduced the cpu

- * cycles spent on reading ref data in sub-pixel filter functions.

- * TODO: Currently, since sub-pixel search range here is -3 ~ 3, copy 22 rows x

- * 32 cols area that is enough for 16x16 macroblock. Later, for SPLITMV, we

- * could reduce the area.

- */

-

 /* Estimated (square) error cost of a motion vector (r,c). The 14 scale comes

  * from the same math as in mv_err_cost(). */

 #define MVC(r, c)                                              \