Fast computation path for forward transform and quantization

This commit enables a fast path computational flow for forward
transformation. It checks the sse and variance of prediction
residuals and decides if the quantized coefficients are all
zero, dc only, or more. It then selects the corresponding coding
path in the forward transformation and quantization stage.

It is currently enabled in rtc coding mode. Will do it for rd
coding mode next.

In speed -6, the runtime for pedestrian_area 1080p at 1000 kbps
goes down from 14234 ms to 13704 ms, i.e., about 4% speed-up.
Overall coding performance for rtc set is changed by -0.18%.

Change-Id: I0452da1786d59bc8bcbe0a35fdae9f623d1d44e1

1 files changed, 12 insertions, 0 deletions

diff --git a/vp9/encoder/vp9_encodeframe.c b/vp9/encoder/vp9_encodeframe.c
index 87e0c08d2..75416f2aa 100644
--- a/vp9/encoder/vp9_encodeframe.c
+++ b/vp9/encoder/vp9_encodeframe.c
@@ -1338,6 +1338,7 @@ static void update_state_rt(VP9_COMP *cpi, PICK_MODE_CONTEXT *ctx,
   }
 
   x->skip = ctx->skip;
+  x->skip_txfm = mbmi->segment_id ? 0 : ctx->skip_txfm;
 }
 
 static void encode_b_rt(VP9_COMP *cpi, const TileInfo *const tile,
@@ -2577,6 +2578,7 @@ static void nonrd_pick_partition(VP9_COMP *cpi, const TileInfo *const tile,
     nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col,
                         &this_rate, &this_dist, bsize);
     ctx->mic.mbmi = xd->mi[0]->mbmi;
+    ctx->skip_txfm = x->skip_txfm;
 
     if (this_rate != INT_MAX) {
       int pl = partition_plane_context(xd, mi_row, mi_col, bsize);
@@ -2663,6 +2665,7 @@ static void nonrd_pick_partition(VP9_COMP *cpi, const TileInfo *const tile,
                         &this_rate, &this_dist, subsize);
 
     pc_tree->horizontal[0].mic.mbmi = xd->mi[0]->mbmi;
+    pc_tree->horizontal[0].skip_txfm = x->skip_txfm;
 
     sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
 
@@ -2672,6 +2675,7 @@ static void nonrd_pick_partition(VP9_COMP *cpi, const TileInfo *const tile,
                           &this_rate, &this_dist, subsize);
 
       pc_tree->horizontal[1].mic.mbmi = xd->mi[0]->mbmi;
+      pc_tree->horizontal[1].skip_txfm = x->skip_txfm;
 
       if (this_rate == INT_MAX) {
         sum_rd = INT64_MAX;
@@ -2701,12 +2705,14 @@ static void nonrd_pick_partition(VP9_COMP *cpi, const TileInfo *const tile,
     nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col,
                         &this_rate, &this_dist, subsize);
     pc_tree->vertical[0].mic.mbmi = xd->mi[0]->mbmi;
+    pc_tree->vertical[0].skip_txfm = x->skip_txfm;
     sum_rd = RDCOST(x->rdmult, x->rddiv, sum_rate, sum_dist);
     if (sum_rd < best_rd && mi_col + ms < cm->mi_cols) {
       load_pred_mv(x, ctx);
       nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col + ms,
                           &this_rate, &this_dist, subsize);
       pc_tree->vertical[1].mic.mbmi = xd->mi[0]->mbmi;
+      pc_tree->vertical[1].skip_txfm = x->skip_txfm;
       if (this_rate == INT_MAX) {
         sum_rd = INT64_MAX;
       } else {
@@ -2795,14 +2801,17 @@ static void nonrd_use_partition(VP9_COMP *cpi,
     case PARTITION_NONE:
       nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col, totrate, totdist, subsize);
       pc_tree->none.mic.mbmi = xd->mi[0]->mbmi;
+      pc_tree->none.skip_txfm = x->skip_txfm;
       break;
     case PARTITION_VERT:
       nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col, totrate, totdist, subsize);
       pc_tree->vertical[0].mic.mbmi = xd->mi[0]->mbmi;
+      pc_tree->vertical[0].skip_txfm = x->skip_txfm;
       if (mi_col + hbs < cm->mi_cols) {
         nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col + hbs,
                             &rate, &dist, subsize);
         pc_tree->vertical[1].mic.mbmi = xd->mi[0]->mbmi;
+        pc_tree->vertical[1].skip_txfm = x->skip_txfm;
         if (rate != INT_MAX && dist != INT64_MAX &&
             *totrate != INT_MAX && *totdist != INT64_MAX) {
           *totrate += rate;
@@ -2813,10 +2822,12 @@ static void nonrd_use_partition(VP9_COMP *cpi,
     case PARTITION_HORZ:
       nonrd_pick_sb_modes(cpi, tile, mi_row, mi_col, totrate, totdist, subsize);
       pc_tree->horizontal[0].mic.mbmi = xd->mi[0]->mbmi;
+      pc_tree->horizontal[0].skip_txfm = x->skip_txfm;
       if (mi_row + hbs < cm->mi_rows) {
         nonrd_pick_sb_modes(cpi, tile, mi_row + hbs, mi_col,
                             &rate, &dist, subsize);
         pc_tree->horizontal[1].mic.mbmi = xd->mi[0]->mbmi;
+        pc_tree->horizontal[1].skip_txfm = x->skip_txfm;
         if (rate != INT_MAX && dist != INT64_MAX &&
             *totrate != INT_MAX && *totdist != INT64_MAX) {
           *totrate += rate;
@@ -3019,6 +3030,7 @@ static void encode_frame_internal(VP9_COMP *cpi) {
   init_encode_frame_mb_context(cpi);
   set_prev_mi(cm);
 
+  x->skip_txfm = 0;
   if (sf->use_nonrd_pick_mode) {
     // Initialize internal buffer pointers for rtc coding, where non-RD
     // mode decision is used and hence no buffer pointer swap needed.