Short circuit flat blocks when coding screen content at realtime speed.

In inter mode search skip all modes except NEARESTMV and DC_PRED.

10% less encode latency for large frames using the chromium remoting_perftests.

+0.313% BDRATE on the screencast set at speed -6.

Change-Id: Ib97a39dd8bcdeab545509e0e02d78ce7033f8c63

vp9/encoder/vp9_pickmode.c

@@ -1251,6 +1251,17 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
   mi->tx_size = VPXMIN(max_txsize_lookup[bsize],
                        tx_mode_to_biggest_tx_size[cm->tx_mode]);
 
+  if (sf->short_circuit_flat_blocks) {
+#if CONFIG_VP9_HIGHBITDEPTH
+    if (xd->cur_buf->flags & YV12_FLAG_HIGHBITDEPTH)
+      x->source_variance = vp9_high_get_sby_perpixel_variance(
+          cpi, &x->plane[0].src, bsize, xd->bd);
+    else
+#endif  // CONFIG_VP9_HIGHBITDEPTH
+      x->source_variance =
+          vp9_get_sby_perpixel_variance(cpi, &x->plane[0].src, bsize);
+  }
+
 #if CONFIG_VP9_TEMPORAL_DENOISING
   vp9_denoiser_reset_frame_stats(ctx);
 #endif
@@ -1289,6 +1300,11 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
     if (cpi->use_svc)
       this_mode = ref_mode_set_svc[idx].pred_mode;
 
+    if (sf->short_circuit_flat_blocks && x->source_variance == 0 &&
+        this_mode != NEARESTMV) {
+      continue;
+    }
+
     if (!(cpi->sf.inter_mode_mask[bsize] & (1 << this_mode)))
       continue;
 
@@ -1704,6 +1720,10 @@ void vp9_pick_inter_mode(VP9_COMP *cpi, MACROBLOCK *x,
       const PREDICTION_MODE this_mode = intra_mode_list[i];
       THR_MODES mode_index = mode_idx[INTRA_FRAME][mode_offset(this_mode)];
       int mode_rd_thresh = rd_threshes[mode_index];
+      if (sf->short_circuit_flat_blocks && x->source_variance == 0 &&
+          this_mode != DC_PRED) {
+        continue;
+      }
 
       if (!((1 << this_mode) & cpi->sf.intra_y_mode_bsize_mask[bsize]))
         continue;

vp9/encoder/vp9_rd.h

@@ -182,6 +182,15 @@ void vp9_setup_pred_block(const MACROBLOCKD *xd,
 int vp9_get_intra_cost_penalty(int qindex, int qdelta,
                                vpx_bit_depth_t bit_depth);
 
+unsigned int vp9_get_sby_perpixel_variance(struct VP9_COMP *cpi,
+                                           const struct buf_2d *ref,
+                                           BLOCK_SIZE bs);
+#if CONFIG_VP9_HIGHBITDEPTH
+unsigned int vp9_high_get_sby_perpixel_variance(struct VP9_COMP *cpi,
+                                                const struct buf_2d *ref,
+                                                BLOCK_SIZE bs, int bd);
+#endif
+
 #ifdef __cplusplus
 }  // extern "C"
 #endif

vp9/encoder/vp9_rdopt.h

@@ -29,15 +29,6 @@ void vp9_rd_pick_intra_mode_sb(struct VP9_COMP *cpi, struct macroblock *x,
                                struct RD_COST *rd_cost, BLOCK_SIZE bsize,
                                PICK_MODE_CONTEXT *ctx, int64_t best_rd);
 
-unsigned int vp9_get_sby_perpixel_variance(VP9_COMP *cpi,
-                                           const struct buf_2d *ref,
-                                           BLOCK_SIZE bs);
-#if CONFIG_VP9_HIGHBITDEPTH
-unsigned int vp9_high_get_sby_perpixel_variance(VP9_COMP *cpi,
-                                                const struct buf_2d *ref,
-                                                BLOCK_SIZE bs, int bd);
-#endif
-
 void vp9_rd_pick_inter_mode_sb(struct VP9_COMP *cpi,
                                struct TileDataEnc *tile_data,
                                struct macroblock *x,

vp9/encoder/vp9_speed_features.c

@@ -401,6 +401,9 @@ static void set_rt_speed_feature(VP9_COMP *cpi, SPEED_FEATURES *sf,
             sf->intra_y_mode_bsize_mask[i] = INTRA_DC_H_V;
       }
     }
+    if (content == VP9E_CONTENT_SCREEN) {
+      sf->short_circuit_flat_blocks = 1;
+    }
   }
 
   if (speed >= 6) {
@@ -534,6 +537,7 @@ void vp9_set_speed_features_framesize_independent(VP9_COMP *cpi) {
   sf->recode_tolerance = 25;
   sf->default_interp_filter = SWITCHABLE;
   sf->simple_model_rd_from_var = 0;
+  sf->short_circuit_flat_blocks = 0;
 
   // Some speed-up features even for best quality as minimal impact on quality.
   sf->adaptive_rd_thresh = 1;

vp9/encoder/vp9_speed_features.h

@@ -438,6 +438,10 @@ typedef struct SPEED_FEATURES {
 
   // Fast approximation of vp9_model_rd_from_var_lapndz
   int simple_model_rd_from_var;
+
+  // Skip a number of expensive mode evaluations for blocks with zero source
+  // variance.
+  int short_circuit_flat_blocks;
 } SPEED_FEATURES;
 
 struct VP9_COMP;