Row based multi-threading of encoding stage

(Yunqing Wang)
This patch implements the row-based multi-threading within tiles in
the encoding pass, and substantially speeds up the multi-threaded
encoder in VP9.

Speed tests at speed 1 on STDHD(using 4 tiles) set show that the
average speedups of the encoding pass(second pass in the 2-pass
encoding) is 7% while using 2 threads, 16% while using 4 threads,
85% while using 8 threads, and 116% while using 16 threads.

Change-Id: I12e41dbc171951958af9e6d098efd6e2c82827de

1 files changed, 18 insertions, 0 deletions

diff --git a/vp9/encoder/vp9_speed_features.c b/vp9/encoder/vp9_speed_features.c
index 9ce756efe..09e91fc17 100644
--- a/vp9/encoder/vp9_speed_features.c
+++ b/vp9/encoder/vp9_speed_features.c
@@ -585,6 +585,15 @@ void vp9_set_speed_features_framesize_dependent(VP9_COMP *cpi) {
       rd->thresh_mult_sub8x8[i] = INT_MAX;
     }
   }
+
+  // With row based multi-threading, the following speed features
+  // have to be disabled to guarantee that bitstreams encoded with single thread
+  // and multiple threads match
+  if (cpi->oxcf.ethread_bit_match) {
+    sf->adaptive_rd_thresh = 0;
+    sf->allow_exhaustive_searches = 0;
+    sf->adaptive_pred_interp_filter = 0;
+  }
 }
 
 void vp9_set_speed_features_framesize_independent(VP9_COMP *cpi) {
@@ -747,4 +756,13 @@ void vp9_set_speed_features_framesize_independent(VP9_COMP *cpi) {
   if (!cpi->oxcf.frame_periodic_boost) {
     sf->max_delta_qindex = 0;
   }
+
+  // With row based multi-threading, the following speed features
+  // have to be disabled to guarantee that bitstreams encoded with single thread
+  // and multiple threads match
+  if (cpi->oxcf.ethread_bit_match) {
+    sf->adaptive_rd_thresh = 0;
+    sf->allow_exhaustive_searches = 0;
+    sf->adaptive_pred_interp_filter = 0;
+  }
 }