commit cdae732fdec302539649269132248abd506132a7
author Jerome Jiang <jianj@google.com> Mon Jan 06 15:23:32 2020 -0800
committer Jerome Jiang <jianj@google.com> Tue Jan 07 17:46:51 2020 +0000
tree 116181cb52640ea0c5440b27d612c161b878b7f5
parent 140bc83bcecd5273a06a34d5b648b73201a060c3

Clean up TODO for prune_single_motion_modes_by_simple_trans

The speed/quality trade-off looks good to me.

almost 2x speed up with 2.9% quality loss on rtc.

This is only used for RT mode in RD path.

Change-Id: I55d97b745e77f96e0ff6c2fe856c1e495cc0384f

av1/encoder/speed_features.c

diff --git a/av1/encoder/speed_features.c b/av1/encoder/speed_features.c
index afc7174..a40f53d 100644
--- a/av1/encoder/speed_features.c
+++ b/av1/encoder/speed_features.c
@@ -611,9 +611,6 @@
     sf->inter_sf.disable_wedge_search_edge_thresh = 0;
     sf->inter_sf.prune_comp_type_by_comp_avg = 1;
     sf->inter_sf.prune_motion_mode_level = 2;
-    // TODO(jianj): Following speed feature will be further explored to
-    // identify the appropriate tradeoff between encoder performance and its
-    // speed.
     sf->inter_sf.prune_single_motion_modes_by_simple_trans = 1;
 
     sf->interp_sf.cb_pred_filter_search = 1;
@@ -920,7 +917,6 @@
   inter_sf->disable_sb_level_mv_cost_upd = 0;
   inter_sf->prune_comp_search_by_single_result = 0;
   inter_sf->skip_repeated_newmv = 0;
-  // TODO(any) Cleanup this speed feature
   inter_sf->prune_single_motion_modes_by_simple_trans = 0;
   inter_sf->inter_mode_rd_model_estimation = 0;
   inter_sf->prune_compound_using_single_ref = 0;