18311 Commits

Author SHA1 Message Date
Debargha Mukherjee
1b17559327 Adds 1D transforms for ADST/FlipADST to make 16
Makes a set of 16 transforms total, adding all 1D
combinations of ADST and FlipADST, and removng all DST
transforms.

lowres, midres both improve by about 0.1% and hdres by
-0.378% in BDRATE but with fewer transforms that are also
simpler.

Further experiments to continue later.

Change-Id: I7348a4c0e12078fdea5ae3a2d36a89a319ffcc6e
2016-03-21 11:19:36 -07:00
Yaowu Xu
c96c3fa2b3 Properly set rate_nocoef when pallete mode is used
Change-Id: Iff04c82b3d3b5cf2c7700717c3c3d678bbbb9f9b
2016-03-21 11:07:53 -07:00
Angie Chiang
abd447e339 mv vp10_fwd_txfm2d_#x# into vp10_rtcd.h
Change-Id: Iad7352698786791b0fd7c005a7edfd1724b71599
2016-03-21 10:51:54 -07:00
Angie Chiang
40ef86f27d Passing TXFM_TYPE instead of func pointer
This is to facilitate sse2 implementation

Change-Id: Id2f53e83c5508c4445d9b1bba00a649cb4da6b74
2016-03-21 10:50:59 -07:00
Jingning Han
66df6e7c7f Refactor sub8x8 reference motion vector search function
Rework the interface to allow codec store the reference motion
vector list information for coding process.

Change-Id: I47e26587f6c0808655e4626f316ec7614a7ad8ed
2016-03-21 10:02:08 -07:00
Jingning Han
5c9d315572 Rework the DRL syntax entropy coding system
This commit re-designs the probability model for the syntax elements
of the dynamic motion vector referencing system.

Change-Id: Icfb8203c7e8f64e10e99f5890e25e6f6b15fe5d1
2016-03-21 09:52:33 -07:00
Jingning Han
4914ae4622 Merge "Enable dynamic motion vector referencing for newmv mode" into nextgenv2 2016-03-19 00:40:04 +00:00
Debargha Mukherjee
6074403d2d Merge "Add ext_partition_types config option" into nextgenv2 2016-03-18 23:51:55 +00:00
Debargha Mukherjee
3c065ac46a Merge "Refactor bsse and skip_txfm in MACROBLOCK." into nextgenv2 2016-03-18 23:51:40 +00:00
Debargha Mukherjee
05029a47a1 Merge "Refactor save_context restore_context in rd_pick_partition." into nextgenv2 2016-03-18 23:51:06 +00:00
Debargha Mukherjee
0ac48f8f65 Merge "Refactor mbmi->inter_tx_size to 2D array." into nextgenv2 2016-03-18 23:50:25 +00:00
Sarah Parker
0adb805db9 Merge "Remove prune three from speed features" into nextgenv2 2016-03-18 21:29:24 +00:00
Sarah Parker
fab5454a16 Remove prune three from speed features
Not getting good results for this feature, will try again
when transforms are frozen.

Change-Id: Id12396786cb9369ad34d0bd845f7beba3a037726
2016-03-18 13:06:40 -07:00
Alex Converse
44ce668063 Add a placeholder forward buffered ANS coder.
This buffered ANS coder supports coding the symbols in forward (decode)
order. Rather than windowing or growing the buffer, right now this
coder merely asserts that the buffer will never overflow.

This approach should allow ANS to be used as a drop in replacement for
other entropy coders rather than requiring complicated reversal logic
throughout the codebase.

Change-Id: I6689271233d0e22fea94c51950415dad5af96598
2016-03-18 19:33:45 +00:00
Yaowu Xu
42e5c2ad8a Two minor logic fixes
Change-Id: I1d5624fb2f34f87a55613036851034ec7c2d0b76
2016-03-18 11:48:19 -07:00
Marco Paniconi
bfc2a7e3a0 Merge "vp9: Improvement to skin detection." 2016-03-18 17:58:03 +00:00
Jingning Han
93748c3e4f Enable dynamic motion vector referencing for newmv mode
This commit enables the dynamic motion vector predictor for NEWMV
mode. It allows the codec to select the best motion vector predictor
in a rate-distortion optimization framework for motion vector
residual coding. The compression performance is improved:
lowres  0.14%
midres  0.27%
hdres   0.24%

Change-Id: I6a601c74eb6cb0b71a613336d40363359f2edecd
2016-03-18 09:51:37 -07:00
hui su
30d2d9783e Small speed up for super_block_uvrd
Eearly termination if U plane RD cost is large enough.

No notable compression performance changes.

Change-Id: Ieeefc5859cb55d94391b502b4bd840bc8bcb2578
2016-03-18 09:28:10 -07:00
Marco
a3562b60d5 vp8-denoiser: Update on the sum_diff threshold.
Change-Id: I0d459c7ba358da287aff1c706279603d6d4d3c48
2016-03-18 08:39:42 -07:00
Geza Lore
f8cfb72a32 Refactor bsse and skip_txfm in MACROBLOCK.
Simple refactoring to 2 dimensional arrays, in preparation for 128
wide superblocks.

Change-Id: I40d447bd9fbd4f755534ea3cc82fc8f4676cea07
2016-03-18 15:30:10 +00:00
Geza Lore
efe7d4e5a2 Refactor mbmi->inter_tx_size to 2D array.
This is in preparation of increasing the superblock size.

Change-Id: I9197e397399fbe8aec1178a45ea0337dd90412d7
2016-03-18 15:30:09 +00:00
Julia Robson
8bc8f2711a Add ext_partition_types config option
Change-Id: I91d280a64d1da77be31004ea91d08d1a70529e42
2016-03-18 15:22:42 +00:00
Hui Su
cb61bfa695 Merge "Minor bug fix in ex-intra experiment" into nextgenv2 2016-03-18 05:01:56 +00:00
James Zern
e1ad39d754 Merge "altref_test: mark AltRefForcedKeyTest as large" 2016-03-18 02:38:14 +00:00
hui su
507c55b227 Minor bug fix in ex-intra experiment
No performance change observed from borg tests.

Change-Id: I20c232c5dde8cfc84452a4c7185389bd8b812ffd
2016-03-17 19:37:21 -07:00
James Zern
0dcd21fb1b altref_test: mark AltRefForcedKeyTest as large
Change-Id: I2bbd7cd2f3b22d3dbfd40d9af430a5d276cd1855
2016-03-17 18:46:10 -07:00
Marco
ce39add465 vp9: Improvement to skin detection.
Add consec_zeromv to skin detection.
Reduces false detection in steady background areas.

Change-Id: Ic31c831537d29f082f58477f82ed78f313d7dd30
2016-03-17 16:26:10 -07:00
Sarah Parker
99235fd908 Merge "Fix hbd segfault for prune_one and prune_two" into nextgenv2 2016-03-17 22:40:47 +00:00
Marco
32c38a20ba vp9: Fix the threshold for gf update in cyclic refresh.
Fix the threshold for forcing refresh of golden frame based
on high motion. The current comparison was incorrect and
prevented this (force update of gf on high motion) from being used.
For now keep this logic under a flag (and off for now) so as to
not change behavior, until further testing.

Change-Id: Ib5f0082159a428b0603b9534e4bcb6f83e4ccb25
2016-03-17 12:24:58 -07:00
Angie Chiang
cb3072ca4a Refactor: call inv_txfm_add
Change-Id: I52c209a5db1b4a6525c04b70291a08ab5a68c6fe
2016-03-17 11:19:00 -07:00
Angie Chiang
6b55b8dc93 Refactor:Merge inverse_transform_block_inter/intra
Merge inverse_transform_block_inter inverse_transform_block_intra
to inverse_transform_block

Change-Id: I0f561830e639e3bf8d831c28a7c784cb0c4c8b09
2016-03-17 11:19:00 -07:00
Angie Chiang
e8bcce1255 Merge "add dct 64x64 transform" into nextgenv2 2016-03-17 18:15:11 +00:00
Sarah Parker
79dd1ee907 Fix hbd segfault for prune_one and prune_two
Change-Id: I71da102550aa7d81961e5f10c71058d5fa8dc6c4
2016-03-17 09:58:29 -07:00
Alex Converse
e6aebcdf08 Merge "Use whole pixel only at speed 8 screen content." 2016-03-17 16:25:28 +00:00
Alex Converse
54e5ff00a6 Merge "Reuse computation in build_nmv_component_cost_table." 2016-03-17 16:25:02 +00:00
Alex Converse
ed5d2e71c0 Merge "Fix using screen content tuning in all the cpu speed tests." 2016-03-17 16:24:48 +00:00
Geza Lore
24659dbca3 Refactor save_context restore_context in rd_pick_partition.
This is a cosmetic patch that removes a great deal of conditional
compilation around CONFIG_VAR_TX from the partition search function.

Change-Id: I9dcef9d4fe6847b793c77bdf565a5cacbdfacd59
2016-03-17 15:53:59 +00:00
Marco Paniconi
825e754288 Merge "vp9-svc: Bugfix on usage of svc_force_zeromv in non-rd pickmode." 2016-03-17 15:14:45 +00:00
Yaowu Xu
51f951292a Merge "Properly save and restore skip related variables" into nextgenv2 2016-03-17 14:41:03 +00:00
Alex Converse
e17f6347b9 Fix using screen content tuning in all the cpu speed tests.
Change-Id: Ie302283a52c563be0f1a84650f5f29be30d2bbe4
2016-03-16 21:57:00 -07:00
Alex Converse
0271bafb4a Reuse computation in build_nmv_component_cost_table.
Saves 1% CPU on remoting_perftest with speed 8.

Change-Id: I28dc2f602bd3bf5639a40c9f21f5a9833f01d4be
2016-03-16 21:55:33 -07:00
Angie Chiang
ed2514a22c add dct 64x64 transform
Change-Id: I131c4d1216cd156e520b8a91c4438c2d3c6602cb
2016-03-16 19:37:21 -07:00
Yunqing Wang
a129905c70 Merge "Optimize HBD up-sampled prediction functions" into nextgenv2 2016-03-16 23:52:08 +00:00
Alex Converse
bd924fbd21 choose_partitioning: Avoid computing segment_id twice.
1% faster remoting_perftest at speed -8.

Change-Id: I78d591e14b8f890c1e69cacb6d490dfeb9c37900
2016-03-16 16:32:49 -07:00
Yaowu Xu
4fcc4f816e Properly save and restore skip related variables
Change-Id: Id52744e140585c08d047fb395b75ac9318a0b4c3
2016-03-16 16:28:52 -07:00
Marco
f03dfcb118 vp9-svc: Bugfix on usage of svc_force_zeromv in non-rd pickmode.
Only affects 1 pass cbr svc mode.

Change-Id: Ifc0948b4aabffa64a9701295d8dbc58a7208a30f
2016-03-16 16:12:16 -07:00
hui su
83b47af18d Add "entropy" experiment
This patch added two features to improve entropy coding efficiency
for coefficient tokens.

1. Choose 1 of 4 default probability tables based on q-index for
key-frames.
It is ported from nextgen branch:
https://chromium-review.googlesource.com/#/c/280586/

2. Do backward update after each superblock (64X64) row using
subframe token counts.

Coding gain: 0.1% on lowres; 0.42% on midres; 0.36% on hdres.
Much larger gain for key-frames: 2.6%, 2.3%, 1.7%.

Design doc: go/huisu-entropy

Change-Id: Ia3b6a615636be09247d70e4c520405637561532b
2016-03-16 11:55:50 -07:00
Angie Chiang
f330444a09 Merge "remove filter_cache" into nextgenv2 2016-03-16 17:21:53 +00:00
Geza Lore
c2005c578b Factor out zeroing above and left context.
Change-Id: I6e5d8cff869c7415a924f845c9e6ccaabe2b7a9b
2016-03-16 13:08:29 +00:00
Geza Lore
8735c17ceb Fix build without supertx.
Change-Id: Ib60821487710f6cf06aaa8ddcbdd5487ba1cbe4f
2016-03-16 13:08:13 +00:00