Commit Graph

2531 Commits

Author SHA1 Message Date
Ronald S. Bultje
1761a6b55a Merge "Use full 32-pixel edge for superblock bestrefmv motion vector ordering." into experimental 2012-11-13 14:12:58 -08:00
Ronald S. Bultje
b147c64c16 Merge "Fix edge MV handling in SBs." into experimental 2012-11-13 14:12:48 -08:00
Deb Mukherjee
7de64f35d3 A fix in MV_REF experiment
This fix ensures that the forward prob update is not turned off for
motion vectors.

Change-Id: I0b63c9401155926763c6294df6cca68b32bac340
2012-11-13 08:27:04 -08:00
Yunqing Wang
e60478d46d Optimize 8x8 dequant and idct
Similar to 16x16 dequant and idct, based on the value of eobs, the
8x8 dequant and idct calculation was simplified to improve decorder
performance.

Combined vp9_dequant_idct_add_8x8 and vp9_dequant_dc_idct_add_8x8
to eliminate duplicate code.

Change-Id: Ia58e50ab27f7012b7379c495837c9c0b5ba9cf7f
2012-11-12 17:41:53 -08:00
Ronald S. Bultje
c79ae1713c Use full 32-pixel edge for superblock bestrefmv motion vector ordering.
Change-Id: I417e39867c020a17d85370972446a8ce2bbe9a6d
2012-11-12 17:06:56 -08:00
Ronald S. Bultje
722972454c Fix edge MV handling in SBs.
Change-Id: Ia1eddb108ec463835e9de8769572d698e21bca49
2012-11-12 17:06:52 -08:00
Paul Wilkins
2669f42b0d New inter mode context
This change is a fix / extension of the newbestrefmv
experiment. As such it is presented without IFDEF.

The change creates a new context for coding inter modes
in vp9_find_mv_refs(). This replaces the context that
was previously calculated in vp9_find_near_mvs().
The new context is unoptimized and not necessarily
any better at this stage (results pending), but eliminates
the need for a legacy call to vp9_find_near_mvs().
Based on numbers from Scott, this could help decode
speed by several %.

In a later patch I will add support for forward update of
context (assuming this helps) and refine the context as
necessary.

Change-Id: I1cd991b82c8df86cc02237a34185e6d67510698a
2012-11-12 15:50:02 +00:00
Paul Wilkins
6fb8953c19 Restrict ref mv search range.
Experiment to test speed trade off of reducing the
extent of the ref mv search.

Reducing the maximum number of tested candidates to 9 had
minimal net effect on quality in any of the tests sets.

Reduction to 7 has a small negative impact (worst was STD-HD
at about -0.2%).

This change is in response to the apparently high number of
decode cycles reported in regard to mv-ref selection.

Change-Id: I0e92e92e324337689358495a1ec9ccdeb23dc774
2012-11-12 11:31:12 +00:00
Deb Mukherjee
d01357bbad New b-intra mode where direction is contextual
Preliminary patch on a new 4x4 intra mode B_CONTEXT_PRED where the
dominant direction from the context is used to encode. Various decoder
changes are needed to support decoding of B_CONTEXT_PRED in conjunction
with hybrid transforms since the scan order and tokenization depends on
the actual direction of prediction obtained from the context. Currently
the traditional directional modes are used in conjunction with the
B_CONTEXT_PRED, which also seems to provide the best results.

The gains are small - in the 0.1% range.

Change-Id: I5a7ea80b5218f42a9c0dfb42d3f79a68c7f0cdc2
2012-11-10 07:12:30 -08:00
Yunqing Wang
71b1885403 Merge "Optimize 16x16 dequant and idct" into experimental 2012-11-09 08:30:53 -08:00
Yunqing Wang
6c17c9fae0 Optimize 16x16 dequant and idct
As suggested by Yaowu, simplified 16x16 dequant and idct. In decoder,
after detoken step, we know the number of non-zero dct coefficients
(eobs) in a macroblock. Idct calculation can be skipped or simplified
based on eobs, which improves the decoder performance.

Change-Id: I9ffa1cb134bcb5a7d64fcf90c81871a96d1b4018
2012-11-07 20:04:09 -08:00
John Koleszar
8959c8b11d Merge with upstream experimental changes (2)
Include upstream changes (variance fixes) into the merged code base.

Change-Id: I4182654c1411c1b15cd23235d3822702613abce1
2012-11-07 14:32:26 -08:00
James Zern
5338d983d6 Merge "Fix variance (signed integer) overflow" into experimental 2012-11-07 12:49:36 -08:00
John Koleszar
2c08c28191 Merge with upstream experimental changes
Include upstream changes (unit test fixes, in particular) into the
merged code base.

Change-Id: I096f8a9d09e2532fbec0c95d7a995ab22fa54b29
2012-11-07 11:46:23 -08:00
John Koleszar
7b8dfcb5a2 Rough merge of master into experimental
Creates a merge between the master and experimental branches. Fixes a
number of conflicts in the build system to allow *either* VP8 or VP9
to be built. Specifically either:

  $ configure --disable-vp9 $ configure --disable-vp8
  --disable-unit-tests

VP9 still exports its symbols and files as VP8, so that will be
resolved in the next commit.

Unit tests are broken in VP9, but this isn't a new issue. They are
fixed upstream on origin/experimental as of this writing, but rebasing
this merge proved difficult, so will tackle that in a second merge
commit.

Change-Id: I2b7d852c18efd58d1ebc621b8041fe0260442c21
2012-11-07 11:30:16 -08:00
Yaowu Xu
0cedaa3631 merge full pixel refmv experiment
Change-Id: Ib39ad47a7d188f3b45416937b7eeb28c3e79b74c
2012-11-07 10:52:45 -08:00
James Zern
984734436d Fix variance (signed integer) overflow
In the variance calculations the difference is summed and later squared.
When the sum exceeds sqrt(2^31) the value is treated as a negative when
it is shifted which gives incorrect results.

To fix this we force the multiplication to be unsigned.

The alternative fix is to shift sum down by 4 before multiplying.
However that will reduce precision.

For 16x16 blocks the maximum sum is 65280 and sqrt(2^31) is 46340 (and
change).

This change is based on:
1698234 Missed some variance casts
fea3556 Fix variance overflow

Change-Id: I2c61856cca9db54b9b81de83b4505ea81a050a0f
2012-11-06 23:06:44 -08:00
Yaowu Xu
a879b4e6d4 fixed function prototype
so they are consistent with actual definitions of the functions

Change-Id: Ie4b4e81b3da3e288fc2edbbd2b393a5c54d2556b
2012-11-06 15:55:11 -08:00
Yaowu Xu
acadcec5c5 group refmv experiment related functions
Change-Id: Iedaa108ddb65f54d768424f9c47ad4d069b656fd
2012-11-06 15:54:47 -08:00
James Zern
182f99f0c6 Merge "fix test builds" into experimental 2012-11-06 12:18:01 -08:00
James Zern
2e3e685799 fix test builds
s/([vV][pP])8/$19/
additionally dct.h was removed; declare the _c functions that are used
in the tests. the TODO for conversion to parameterized tests still
remains.

Change-Id: I73db9425a57075bbb78a92693ba6b320578981cd
2012-11-06 12:12:58 -08:00
Yunqing Wang
4626faf1e7 Convert 16x16 dct/idct to integer forms
Converted vp9_short_fdct16x16_c and vp9_short_idct16x16_c to
integer versions.

Change-Id: Ie3ec985a890ac0f4f4f5818e6f0122e00c8af69f
2012-11-06 11:25:25 -08:00
Yaowu Xu
55f2f14f10 Merge "silent a lot of MSVC compiler warnings" into experimental 2012-11-06 09:39:47 -08:00
Yaowu Xu
8a336b0d0d silent a lot of MSVC compiler warnings
there are still a couple type of warning left, which are related to
double constants assigned to float type. As those would be addressed
by the conversion of transforms into integer version. This commit
has left those un-dealt with.

Change-Id: I48fd9b489c0c27ad6b543f4177423419f929f2bb
2012-11-06 09:09:25 -08:00
Jim Bankoski
7849aa20ed remove invoke_search macro
Removed invoke search from encoder

Change-Id: I3d809b795abe6df0e71366edfe94026aaede14fb
2012-11-05 16:58:03 -08:00
James Zern
ee38c4184b loopfilter: prevent signed integer overflow
use unsigned ints to extended filter values in
vp9_mbloop_filter_horizontal_edge_c_sse2

Change-Id: I55ec3ac2bcb9baf55626b0384d151b07fc8e087d
2012-11-03 09:45:21 -07:00
Yunqing Wang
28826a913c Merge "Fix eobs data type" into experimental 2012-11-02 16:00:56 -07:00
Yunqing Wang
d41b0e6498 Fix eobs data type
The block sizes for decoding tokens are up to 16x16, which means
eobs is within [0, 256]. Using (signed) char is not enough. Changed
eobs data type to unsigned short to fix the problem.

Change-Id: I88a7d3098e1f1604c336d6adb88ffec971fb03a6
2012-11-02 13:22:29 -07:00
Ronald S. Bultje
3c4f47e843 Place non-static function prototypes in a header file.
Change-Id: I7cd21b9f1e69f4e0b3338bfe27b3c67e4b47de58
2012-11-02 11:22:57 -07:00
John Koleszar
06f3e51da6 vpx_scale: sync from master
Update vpx_scale from current code in master, run style transform, fix
lint warnings.

Change-Id: I47eadeb5b6881d448ea3728537f9b8a5b5aac78e
2012-11-02 08:44:54 -07:00
Ronald S. Bultje
4b2c2b9aa4 Rename vp8/ codec directory to vp9/.
Change-Id: Ic084c475844b24092a433ab88138cf58af3abbe4
2012-11-01 16:31:22 -07:00