Adds speed 8 to vp9 as reference

Adds a speed 8 to VP9 where only the nearestmv (0 mv) is searched.
This seems to be about the same speed as vp8 speed 5.

Adds a new speed feature to disable inter modes based on a mask for
each blocksize.

Adds code for having lower complexity motion search methods
in nonrd pick mode function, even though speed 7 still uses DIAMOND
search for now.

Also uses HEX search for speed 6 rather than FAST_HEX which improves
psnr by 0.56% without any noticeable speed drop (tested on gipsmotion).

Change-Id: Ic13176572dbd3aed5884a26786940a4b1bbd8a75

diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 46072a2..9930484 100644 (file)
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -2357,7 +2357,8 @@ static void nonrd_use_partition(VP9_COMP *cpi, const TileInfo *const tile,
       set_offsets(cpi, tile, row, col, bs);
 
       if (cm->frame_type != KEY_FRAME)
-        vp9_pick_inter_mode(cpi, x, tile, row, col, &brate, &bdist, bs);
+        vp9_pick_inter_mode(cpi, x, tile, row, col,
+                            &brate, &bdist, bs);
       else
         set_mode_info(&xd->mi_8x8[0]->mbmi, bs, mode, row, col);
 

diff --git a/vp9/encoder/vp9_onyx_if.c b/vp9/encoder/vp9_onyx_if.c
index 3921ea8..9f63591 100644 (file)
--- a/vp9/encoder/vp9_onyx_if.c
+++ b/vp9/encoder/vp9_onyx_if.c
@@ -734,6 +734,7 @@ static void set_good_speed_feature(VP9_COMMON *cm,
     sf->mode_skip_start = 6;
   }
 }
+
 static void set_rt_speed_feature(VP9_COMMON *cm,
                                  SPEED_FEATURES *sf,
                                  int speed) {
@@ -853,10 +854,17 @@ static void set_rt_speed_feature(VP9_COMMON *cm,
   }
   if (speed >= 6) {
     sf->partition_search_type = VAR_BASED_FIXED_PARTITION;
+    sf->search_method = HEX;
   }
   if (speed >= 7) {
     sf->partition_search_type = VAR_BASED_FIXED_PARTITION;
     sf->use_nonrd_pick_mode = 1;
+    sf->search_method = NSTEP;
+  }
+  if (speed >= 8) {
+    int i;
+    for (i = 0; i < BLOCK_SIZES; ++i)
+      sf->disable_inter_mode_mask[i] = 14;   // only search NEARESTMV (0)
   }
 }
 
@@ -918,6 +926,8 @@ void vp9_set_speed_features(VP9_COMP *cpi) {
   sf->mode_skip_start = MAX_MODES;  // Mode index at which mode skip mask set
   sf->use_nonrd_pick_mode = 0;
   sf->encode_breakout_thresh = 0;
+  for (i = 0; i < BLOCK_SIZES; ++i)
+    sf->disable_inter_mode_mask[i] = 0;
 
   switch (cpi->oxcf.mode) {
     case MODE_BESTQUALITY:

diff --git a/vp9/encoder/vp9_onyx_int.h b/vp9/encoder/vp9_onyx_int.h
index c4b018a..019cb13 100644 (file)
--- a/vp9/encoder/vp9_onyx_int.h
+++ b/vp9/encoder/vp9_onyx_int.h
@@ -416,6 +416,10 @@ typedef struct {
   // This variable sets the encode_breakout threshold. Currently, it is only
   // enabled in real time mode.
   int encode_breakout_thresh;
+
+  // A binary mask indicating if NEARESTMV, NEARMV, ZEROMV, NEWMV
+  // modes are disabled in order from LSB to MSB for each BLOCK_SIZE.
+  int disable_inter_mode_mask[BLOCK_SIZES];
 } SPEED_FEATURES;
 
 typedef struct {

diff --git a/vp9/encoder/vp9_pickmode.c b/vp9/encoder/vp9_pickmode.c
index 87f20fa..811f09c 100644 (file)
--- a/vp9/encoder/vp9_pickmode.c
+++ b/vp9/encoder/vp9_pickmode.c
@@ -99,14 +99,27 @@ static int full_pixel_motion_search(VP9_COMP *cpi, MACROBLOCK *x,
   mvp_full.row >>= 3;
 
   if (cpi->sf.search_method == FAST_HEX) {
-    vp9_fast_hex_search(x, &mvp_full, step_param, sadpb, &cpi->fn_ptr[bsize],
-                        1, &ref_mv.as_mv, &tmp_mv->as_mv);
+    bestsme = vp9_fast_hex_search(x, &mvp_full, step_param, sadpb,
+                                  &cpi->fn_ptr[bsize], 1,
+                                  &ref_mv.as_mv, &tmp_mv->as_mv);
+  } else if (cpi->sf.search_method == HEX) {
+    bestsme = vp9_hex_search(x, &mvp_full, step_param, sadpb, 1,
+                             &cpi->fn_ptr[bsize], 1,
+                             &ref_mv.as_mv, &tmp_mv->as_mv);
+  } else if (cpi->sf.search_method == SQUARE) {
+    bestsme = vp9_square_search(x, &mvp_full, step_param, sadpb, 1,
+                                &cpi->fn_ptr[bsize], 1,
+                                &ref_mv.as_mv, &tmp_mv->as_mv);
+  } else if (cpi->sf.search_method == BIGDIA) {
+    bestsme = vp9_bigdia_search(x, &mvp_full, step_param, sadpb, 1,
+                                &cpi->fn_ptr[bsize], 1,
+                                &ref_mv.as_mv, &tmp_mv->as_mv);
   } else {
-    vp9_full_pixel_diamond(cpi, x, &mvp_full, step_param, sadpb, further_steps,
-                           1, &cpi->fn_ptr[bsize], &ref_mv.as_mv,
-                           &tmp_mv->as_mv);
+    bestsme = vp9_full_pixel_diamond(cpi, x, &mvp_full, step_param,
+                                     sadpb, further_steps, 1,
+                                     &cpi->fn_ptr[bsize],
+                                     &ref_mv.as_mv, &tmp_mv->as_mv);
   }
-
   x->mv_col_min = tmp_col_min;
   x->mv_col_max = tmp_col_max;
   x->mv_row_min = tmp_row_min;
@@ -200,8 +213,8 @@ int64_t vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
   static const int flag_list[4] = { 0, VP9_LAST_FLAG, VP9_GOLD_FLAG,
                                     VP9_ALT_FLAG };
   int64_t best_rd = INT64_MAX;
-  int64_t this_rd;
-  static const int cost[4]= { 0, 50, 75, 100 };
+  int64_t this_rd = INT64_MAX;
+  static const int cost[4]= { 0, 2, 4, 6 };
 
   const int64_t inter_mode_thresh = 300;
   const int64_t intra_mode_cost = 50;
@@ -239,7 +252,6 @@ int64_t vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
 
   for (ref_frame = LAST_FRAME; ref_frame <= LAST_FRAME ; ++ref_frame) {
     int rate_mv = 0;
-
     if (!(cpi->ref_frame_flags & flag_list[ref_frame]))
       continue;
 
@@ -252,11 +264,15 @@ int64_t vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
     mbmi->ref_frame[0] = ref_frame;
 
     for (this_mode = NEARESTMV; this_mode <= NEWMV; ++this_mode) {
-      int rate = cost[INTER_OFFSET(this_mode)];
+      int rate = cost[INTER_OFFSET(this_mode)]
+          << (num_pels_log2_lookup[bsize] - 4);
       int64_t dist;
+      if (cpi->sf.disable_inter_mode_mask[bsize] &
+          (1 << INTER_OFFSET(this_mode)))
+        continue;
 
       if (this_mode == NEWMV) {
-        if (this_rd < 500)
+        if (this_rd < (1 << num_pels_log2_lookup[bsize]))
           continue;
 
         x->mode_sad[ref_frame][INTER_OFFSET(NEWMV)] =
@@ -314,6 +330,5 @@ int64_t vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
       }
     }
   }
-
   return INT64_MAX;
 }

diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index b57b948..261dfa8 100644 (file)
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1725,6 +1725,8 @@ static void rd_check_segment_txsize(VP9_COMP *cpi, MACROBLOCK *x,
 
         mode_idx = INTER_OFFSET(this_mode);
         bsi->rdstat[i][mode_idx].brdcost = INT64_MAX;
+        if (cpi->sf.disable_inter_mode_mask[bsize] & (1 << mode_idx))
+          continue;
 
         // if we're near/nearest and mv == 0,0, compare to zeromv
         if ((this_mode == NEARMV || this_mode == NEARESTMV ||
@@ -3316,6 +3318,9 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
 
     this_mode = vp9_mode_order[mode_index].mode;
     ref_frame = vp9_mode_order[mode_index].ref_frame[0];
+    if (ref_frame != INTRA_FRAME &&
+        cpi->sf.disable_inter_mode_mask[bsize] & (1 << INTER_OFFSET(this_mode)))
+      continue;
     second_ref_frame = vp9_mode_order[mode_index].ref_frame[1];
 
     comp_pred = second_ref_frame > INTRA_FRAME;