From a142d6fc935cba1dc6ceb4d798b3aac68b34684b Mon Sep 17 00:00:00 2001
From: Jingning Han <jingning@google.com>
Date: Tue, 16 Jul 2013 12:04:07 -0700
Subject: [PATCH] Skip redundant motion search in 4x4 level rd loop
This commit makes the encoder to perform motion search only once
per reference frame type for each 4x4/4x8/8x4 block. For bus_cif
at 2000 kbps, the runtime goes from 253812ms -> 217817ms
(14% speed-up) for speed 0.
Change-Id: I5f17599ccc8cfaf93ccb4f98fcb6008af6d79e92
---
vp9/encoder/vp9_rdopt.c | 50 ++++++++++++++++++++++-------------------
1 file changed, 27 insertions(+), 23 deletions(-)
diff --git a/vp9/encoder/vp9_rdopt.c b/vp9/encoder/vp9_rdopt.c
index 91f606a1c4..58eeed2152 100644
--- a/vp9/encoder/vp9_rdopt.c
+++ b/vp9/encoder/vp9_rdopt.c
@@ -1842,7 +1842,8 @@ static void rd_check_segment_txsize(VP9_COMP *cpi, MACROBLOCK *x,
vpx_memcpy(t_left_s, t_left, sizeof(t_left_s));
// motion search for newmv (single predictor case only)
- if (mbmi->ref_frame[1] <= 0 && this_mode == NEWMV) {
+ if (mbmi->ref_frame[1] <= 0 && this_mode == NEWMV &&
+ seg_mvs[i][mbmi->ref_frame[0]].as_int == INVALID_MV) {
int step_param = 0;
int further_steps;
int thissme, bestsme = INT_MAX;
@@ -1914,7 +1915,10 @@ static void rd_check_segment_txsize(VP9_COMP *cpi, MACROBLOCK *x,
// restore src pointers
mi_buf_restore(x, orig_src, orig_pre);
- } else if (mbmi->ref_frame[1] > 0 && this_mode == NEWMV) {
+ }
+
+ if (mbmi->ref_frame[1] > 0 && this_mode == NEWMV &&
+ mbmi->interp_filter == vp9_switchable_interp[0]) {
if (seg_mvs[i][mbmi->ref_frame[1]].as_int == INVALID_MV ||
seg_mvs[i][mbmi->ref_frame[0]].as_int == INVALID_MV)
continue;
@@ -3415,28 +3419,28 @@ int64_t vp9_rd_pick_inter_mode_sb(VP9_COMP *cpi, MACROBLOCK *x,
if ((newbest && cm->mcomp_filter_type == SWITCHABLE) ||
(mbmi->interp_filter == cm->mcomp_filter_type &&
cm->mcomp_filter_type != SWITCHABLE)) {
- tmp_best_rdu = tmp_rd;
- tmp_best_rate = rate;
- tmp_best_ratey = rate_y;
- tmp_best_distortion = distortion;
- tmp_best_skippable = skippable;
- tmp_best_mbmode = *mbmi;
- tmp_best_partition = *x->partition_info;
- for (i = 0; i < 4; i++)
- tmp_best_bmodes[i] = xd->mode_info_context->bmi[i];
- pred_exists = 1;
- if (switchable_filter_index == 0 &&
- cpi->sf.use_rd_breakout &&
- best_rd < INT64_MAX) {
- if (tmp_best_rdu / 2 > best_rd) {
- // skip searching the other filters if the first is
- // already substantially larger than the best so far
- tmp_best_filter = mbmi->interp_filter;
- tmp_best_rdu = INT64_MAX;
- break;
- }
- }
+ tmp_best_rdu = tmp_rd;
+ tmp_best_rate = rate;
+ tmp_best_ratey = rate_y;
+ tmp_best_distortion = distortion;
+ tmp_best_skippable = skippable;
+ tmp_best_mbmode = *mbmi;
+ tmp_best_partition = *x->partition_info;
+ for (i = 0; i < 4; i++)
+ tmp_best_bmodes[i] = xd->mode_info_context->bmi[i];
+ pred_exists = 1;
+ if (switchable_filter_index == 0 &&
+ cpi->sf.use_rd_breakout &&
+ best_rd < INT64_MAX) {
+ if (tmp_best_rdu / 2 > best_rd) {
+ // skip searching the other filters if the first is
+ // already substantially larger than the best so far
+ tmp_best_filter = mbmi->interp_filter;
+ tmp_best_rdu = INT64_MAX;
+ break;
}
+ }
+ }
} // switchable_filter_index loop
if (tmp_best_rdu == INT64_MAX)
continue;
--
GitLab