Enable skipping reference frame check in rd loop

This commit allows encoder to compare the SAD cost associated with
the best motion vector predictor, per frame. If one reference frame
has this cost more than 4 times of the best SAD cost given by other
reference frames, skip NEARESTMV, NEARMV, ZEROMV mode check of this
reference frame.

This setting is turned on in speed 2 and above. Compression quality
change in speed 2:
derf  -0.014%
yt    -0.097%
hd    -0.023%
stdhd  0.046%

It reduces the speed 2 runtime of test sequences:
pedestrian_area_1080p 4000 kbps 310763 ms -> 303595 ms
bluesky_1080p 6000 kbps         259852 ms -> 251920 ms

Change-Id: I7f59cf79503d51836d61d56d50dc5bdf0e502e22

diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index a4d5fbe8460aeff792300f35526c7689f2ade843..43f803842660aa4b14f1ce8bdee8a5965363e16d 100644 (file)
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -1925,9 +1925,6 @@ static void encode_sb_row(VP9_COMP *cpi, const TileInfo *const tile,
 
     vp9_zero(cpi->mb.pred_mv);
 
-    if (cpi->sf.reference_masking)
-      rd_pick_reference_frame(cpi, tile, mi_row, mi_col);
-
     if (cpi->sf.use_lastframe_partitioning ||
         cpi->sf.use_one_partition_size_always ) {
       const int idx_str = cm->mode_info_stride * mi_row + mi_col;

diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 2cdf45d3fb3a25ff500776aa3dcbb54bb14d26a7..cab3ef100f66b262572ab08ac76509dfeb02e8a8 100644 (file)
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -878,6 +878,7 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
         sf->use_rd_breakout = 1;
         sf->adaptive_motion_search = 1;
         sf->adaptive_pred_filter_type = 2;
+        sf->reference_masking = 1;
         sf->auto_mv_step_size = 1;
 
         sf->disable_filter_search_var_thresh = 50;
@@ -914,6 +915,7 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
         sf->use_rd_breakout = 1;
         sf->adaptive_motion_search = 1;
         sf->adaptive_pred_filter_type = 2;
+        sf->reference_masking = 1;
         sf->auto_mv_step_size = 1;
 
         sf->disable_filter_search_var_thresh = 100;
@@ -948,6 +950,7 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
         sf->use_rd_breakout = 1;
         sf->adaptive_motion_search = 1;
         sf->adaptive_pred_filter_type = 2;
+        sf->reference_masking = 1;
         sf->auto_mv_step_size = 1;
 
         sf->disable_filter_search_var_thresh = 200;

diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 97dc1e0ffadc9ebebbf0a37bd9e0e031e2824f77..d940db94517bdb93e07825101cf0a4212cef3de1 100644 (file)
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -3174,7 +3174,7 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
 
   *returnrate = INT_MAX;
 
-  for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ref_frame++) {
+  for (ref_frame = LAST_FRAME; ref_frame <= ALTREF_FRAME; ++ref_frame) {
     x->pred_mv_sad[ref_frame] = INT_MAX;
     if (cpi->ref_frame_flags & flag_list[ref_frame]) {
       setup_buffer_inter(cpi, x, tile, get_ref_frame_idx(cpi, ref_frame),
@@ -3185,6 +3185,18 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
     frame_mv[ZEROMV][ref_frame].as_int = 0;
   }
 
+  cpi->ref_frame_mask = 0;
+  for (ref_frame = LAST_FRAME;
+       ref_frame <= ALTREF_FRAME && cpi->sf.reference_masking; ++ref_frame) {
+    int i;
+    for (i = LAST_FRAME; i <= ALTREF_FRAME; ++i) {
+      if ((x->pred_mv_sad[ref_frame] >> 2) > x->pred_mv_sad[i]) {
+        cpi->ref_frame_mask |= (1 << ref_frame);
+        break;
+      }
+    }
+  }
+
   for (mode_index = 0; mode_index < MAX_MODES; ++mode_index) {
     int mode_excluded = 0;
     int64_t this_rd = INT64_MAX;
@@ -3234,8 +3246,7 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
     }
 
     // Skip if the current reference frame has been masked off
-    if (cpi->sf.reference_masking && !cpi->set_ref_frame_mask &&
-        (cpi->ref_frame_mask & (1 << ref_frame)))
+    if (cpi->ref_frame_mask & (1 << ref_frame) && this_mode != NEWMV)
       continue;
 
     // Test best rd so far against threshold for trying this mode.
@@ -3640,11 +3651,6 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
     }
   }
 
-  // If we are using reference masking and the set mask flag is set then
-  // create the reference frame mask.
-  if (cpi->sf.reference_masking && cpi->set_ref_frame_mask)
-    cpi->ref_frame_mask = ~(1 << vp9_mode_order[best_mode_index].ref_frame[0]);
-
   // Flag all modes that have a distortion thats > 2x the best we found at
   // this level.
   for (mode_index = 0; mode_index < MB_MODE_COUNT; ++mode_index) {
@@ -3805,6 +3811,18 @@ int64_t vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
     frame_mv[ZEROMV][ref_frame].as_int = 0;
   }
 
+  cpi->ref_frame_mask = 0;
+  for (ref_frame = LAST_FRAME;
+       ref_frame <= ALTREF_FRAME && cpi->sf.reference_masking; ++ref_frame) {
+    int i;
+    for (i = LAST_FRAME; i <= ALTREF_FRAME; ++i) {
+      if ((x->pred_mv_sad[ref_frame] >> 1) > x->pred_mv_sad[i]) {
+        cpi->ref_frame_mask |= (1 << ref_frame);
+        break;
+      }
+    }
+  }
+
   for (mode_index = 0; mode_index < MAX_REFS; ++mode_index) {
     int mode_excluded = 0;
     int64_t this_rd = INT64_MAX;
@@ -3852,11 +3870,6 @@ int64_t vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
         continue;
     }
 
-    // Skip if the current reference frame has been masked off
-    if (cpi->sf.reference_masking && !cpi->set_ref_frame_mask &&
-        (cpi->ref_frame_mask & (1 << ref_frame)))
-      continue;
-
     // Test best rd so far against threshold for trying this mode.
     if ((best_rd <
          ((int64_t)cpi->rd_thresh_sub8x8[segment_id][bsize][mode_index] *
@@ -4366,11 +4379,6 @@ int64_t vp9_rd_pick_inter_mode_sub8x8(VP9_COMP *cpi, MACROBLOCK *x,
     }
   }
 
-  // If we are using reference masking and the set mask flag is set then
-  // create the reference frame mask.
-  if (cpi->sf.reference_masking && cpi->set_ref_frame_mask)
-    cpi->ref_frame_mask = ~(1 << vp9_ref_order[best_mode_index].ref_frame[0]);
-
   if (best_rd == INT64_MAX && bsize < BLOCK_8X8) {
     *returnrate = INT_MAX;
     *returndistortion = INT_MAX;