[FFmpeg-cvslog] avcodec/motion_est: Avoid branches for put(_no_rnd) selection

[Andreas Rheinhardt]

ffmpeg | branch: master | Andreas Rheinhardt <andreas.rheinhardt at outlook.com> | Sun Jun 30 11:31:04 2024 +0200| [d4fd475005a2febbfcd0b76c8ab3584a87430972] | committer: Andreas Rheinhardt

avcodec/motion_est: Avoid branches for put(_no_rnd) selection

MotionEstContext contains pointers (to function pointers) that
have been set on a per-frame basis depending upon no_rounding
in ff_me_init_pic() to avoid branches like these. Also makes
MotionEstContext more independent of MpegEncContext.

Reviewed-by: Ramiro Polla <ramiro.polla at gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at outlook.com>

> http://git.videolan.org/gitweb.cgi/ffmpeg.git/?a=commit;h=d4fd475005a2febbfcd0b76c8ab3584a87430972
---

 libavcodec/motion_est.c | 25 +++++--------------------
 1 file changed, 5 insertions(+), 20 deletions(-)

diff --git a/libavcodec/motion_est.c b/libavcodec/motion_est.c
index e4f17fb2d8..46c4ca2dd9 100644
--- a/libavcodec/motion_est.c
+++ b/libavcodec/motion_est.c
@@ -661,18 +661,12 @@ static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
                 const uint8_t *ref = c->ref[block][0] + (mx4>>2) + (my4>>2)*stride;
                 dxy = ((my4 & 3) << 2) | (mx4 & 3);
 
-                if(s->no_rounding)
-                    s->qdsp.put_no_rnd_qpel_pixels_tab[1][dxy](dest_y, ref, stride);
-                else
-                    s->qdsp.put_qpel_pixels_tab[1][dxy](dest_y, ref, stride);
+                c->qpel_put[1][dxy](dest_y, ref, stride);
             }else{
                 const uint8_t *ref = c->ref[block][0] + (mx4>>1) + (my4>>1)*stride;
                 dxy = ((my4 & 1) << 1) | (mx4 & 1);
 
-                if(s->no_rounding)
-                    s->hdsp.put_no_rnd_pixels_tab[1][dxy](dest_y    , ref    , stride, h);
-                else
-                    s->hdsp.put_pixels_tab       [1][dxy](dest_y    , ref    , stride, h);
+                c->hpel_put[1][dxy](dest_y, ref, stride, h);
             }
             dmin_sum+= (mv_penalty[mx4-pred_x4] + mv_penalty[my4-pred_y4])*c->mb_penalty_factor;
         }else
@@ -713,13 +707,8 @@ static inline int h263_mv4_search(MpegEncContext *s, int mx, int my, int shift)
 
         offset= (s->mb_x*8 + (mx>>1)) + (s->mb_y*8 + (my>>1))*s->uvlinesize;
 
-        if(s->no_rounding){
-            s->hdsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad    , s->last_pic.data[1] + offset, s->uvlinesize, 8);
-            s->hdsp.put_no_rnd_pixels_tab[1][dxy](c->scratchpad + 8, s->last_pic.data[2] + offset, s->uvlinesize, 8);
-        }else{
-            s->hdsp.put_pixels_tab       [1][dxy](c->scratchpad    , s->last_pic.data[1] + offset, s->uvlinesize, 8);
-            s->hdsp.put_pixels_tab       [1][dxy](c->scratchpad + 8, s->last_pic.data[2] + offset, s->uvlinesize, 8);
-        }
+        c->hpel_put[1][dxy](c->scratchpad    , s->last_pic.data[1] + offset, s->uvlinesize, 8);
+        c->hpel_put[1][dxy](c->scratchpad + 8, s->last_pic.data[2] + offset, s->uvlinesize, 8);
 
         dmin_sum += c->mb_cmp[1](s, s->new_pic->data[1] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad,     s->uvlinesize, 8);
         dmin_sum += c->mb_cmp[1](s, s->new_pic->data[2] + s->mb_x * 8 + s->mb_y * 8 * s->uvlinesize, c->scratchpad + 8, s->uvlinesize, 8);
@@ -825,11 +814,7 @@ static int interlaced_search(MpegEncContext *s, int ref_index,
                 const uint8_t *ref = c->ref[field_select+ref_index][0] + (mx_i>>1) + (my_i>>1)*stride;
                 dxy = ((my_i & 1) << 1) | (mx_i & 1);
 
-                if(s->no_rounding){
-                    s->hdsp.put_no_rnd_pixels_tab[size][dxy](c->scratchpad, ref    , stride, h);
-                }else{
-                    s->hdsp.put_pixels_tab       [size][dxy](c->scratchpad, ref    , stride, h);
-                }
+                c->hpel_put[size][dxy](c->scratchpad, ref, stride, h);
                 dmin = c->mb_cmp[size](s, c->src[block][0], c->scratchpad, stride, h);
                 dmin+= (mv_penalty[mx_i-c->pred_x] + mv_penalty[my_i-c->pred_y] + 1)*c->mb_penalty_factor;
             }else