diff options
| author | Jingning Han <jingning@google.com> | 2014-07-01 16:10:44 -0700 |
|---|---|---|
| committer | Jingning Han <jingning@google.com> | 2014-07-01 17:00:07 -0700 |
| commit | 9ac2f6632044ad2ec268ed4921c31391f30ecd0c (patch) | |
| tree | 9eba14581dabcbdbf14534ace311009b8860f1c0 /vp9/encoder/vp9_speed_features.c | |
| parent | 6643b8868da0efc6ec8217cb16c8b3c9bb8ef62f (diff) | |
| download | libvpx-9ac2f6632044ad2ec268ed4921c31391f30ecd0c.tar libvpx-9ac2f6632044ad2ec268ed4921c31391f30ecd0c.tar.gz libvpx-9ac2f6632044ad2ec268ed4921c31391f30ecd0c.tar.bz2 libvpx-9ac2f6632044ad2ec268ed4921c31391f30ecd0c.zip | |
Re-design quantization process
This commit re-designs the quantization process for transform
coefficient blocks of size 4x4 to 16x16. It improves compression
performance for speed 7 by 3.85%. The SSSE3 version for the
new quantization process is included.
The average runtime of the 8x8 block quantization is reduced
from 285 cycles -> 255 cycles, i.e., over 10% faster.
Change-Id: I61278aa02efc70599b962d3314671db5b0446a50
Diffstat (limited to 'vp9/encoder/vp9_speed_features.c')
| -rw-r--r-- | vp9/encoder/vp9_speed_features.c | 2 |
1 files changed, 2 insertions, 0 deletions
diff --git a/vp9/encoder/vp9_speed_features.c b/vp9/encoder/vp9_speed_features.c index 88a750276..be68bebf5 100644 --- a/vp9/encoder/vp9_speed_features.c +++ b/vp9/encoder/vp9_speed_features.c @@ -279,6 +279,7 @@ static void set_rt_speed_feature(VP9_COMP *cpi, SPEED_FEATURES *sf, sf->reuse_inter_pred_sby = 1; } if (speed >= 7) { + sf->use_quant_fp = cm->frame_type == KEY_FRAME ? 0 : 1; sf->lpf_pick = LPF_PICK_MINIMAL_LPF; sf->encode_breakout_thresh = (MIN(cm->width, cm->height) >= 720) ? 800 : 300; @@ -314,6 +315,7 @@ void vp9_set_speed_features(VP9_COMP *cpi) { sf->use_lp32x32fdct = 0; sf->adaptive_motion_search = 0; sf->adaptive_pred_interp_filter = 0; + sf->use_quant_fp = 0; sf->reference_masking = 0; sf->partition_search_type = SEARCH_PARTITION; sf->less_rectangular_check = 0; |