Commit Graph

12199 Commits

Author SHA1 Message Date
Jingning Han
4bacaab46d Cosmetic naming change
Rename set_modeinfo_offsets as set_mode_info_offsets, to be more
consistent with naming convention.

Change-Id: I68ca1f36c4a78127d9439a50c1506a2afd07927d
2014-12-09 10:32:04 -08:00
Jingning Han
f051a7beab Take out redundant setting of mode_info from set_block_size
The later encoding process will take the top-left block's
mode_info for pre-determined block size.

Change-Id: I76a90f9ce7f3b2dbc2975b52442114e461c465b5
2014-12-09 10:27:18 -08:00
Jingning Han
1395ded2a7 Remove unused rd cost calculation from nonrd_use_partition
The per block rd cost calculation is not needed when partition
size is preset.

Change-Id: Ie5575248bbffb584e908aa13097f697ace6ec747
2014-12-08 18:45:19 -08:00
hkuang
f925e5ce0f Merge "Improve the performance by caching the left_mi and right_mi in macroblockd." 2014-12-08 10:24:17 -08:00
Paul Wilkins
127f65531b Merge "Use average mb energy from first pass in AQ2 test." 2014-12-08 09:01:39 -08:00
Frank Galligan
0f8e8330eb Merge "Fix potential integer overflow." 2014-12-07 21:37:39 -08:00
James Zern
da464c483f Merge "vp9 asserts: fix compile warning" 2014-12-05 21:09:42 -08:00
James Zern
0e00dbae02 Merge "fix building with --disable-spatial-resampling" 2014-12-05 20:36:44 -08:00
James Zern
0e8cc9a82f fix building with --disable-spatial-resampling
vpx_scale.c is only used by the vp8 encoder when spatial resampling is
enabled.

Change-Id: If3d3ad81e9ee6e0b59f8c040b9624ef52598fe03
2014-12-05 19:08:36 -08:00
James Zern
3db785facc Merge "vp9: fix frame-parallel encoding" 2014-12-05 19:00:48 -08:00
Deb Mukherjee
0d367474d0 Merge "Some internal-stats, vp9-highbitdepth bug fixes" 2014-12-05 17:49:52 -08:00
James Zern
6db81fd629 vp9: fix frame-parallel encoding
the flag in the header wasn't being set based on the encoder
configuration in non-intra only mode

broken since:
fbc2fbf Adding oxcf temp variable.

Change-Id: Ib4cff9901889824bc4e68d7f0f6deb1e41df2f53
2014-12-05 17:44:46 -08:00
Jingning Han
bd6bfb93b0 Merge "Remove redundant rdcost reset" 2014-12-05 17:35:07 -08:00
Jingning Han
296afb9440 Merge "Fix a motion search skip condition in vp9_pick_inter_mode" 2014-12-05 17:35:04 -08:00
Jingning Han
3d8d1e374e Merge "Remove redundant MB_MODE_INFO reset from vp9_pick_mode_inter" 2014-12-05 16:59:50 -08:00
hkuang
382f86f945 Improve the performance by caching the left_mi and right_mi in macroblockd.
This improve the deocde performance by ~2% on Nexus 7 2013.

Change-Id: Ie9c4ba0371a149eb7fddc687a6a291c17298d6c3
2014-12-05 16:25:42 -08:00
James Zern
616b3a810f vp9 asserts: fix compile warning
string literal to int within an assert

Change-Id: I76a173f96b9add5bf27c3f5ad5d72c6f30e51629
2014-12-05 16:20:42 -08:00
Jingning Han
17bedc54f5 Remove redundant rdcost reset
The initial reset of this_rdc in vp9_pick_inter_mode is not needed,
since it will be re-assign when used.

Change-Id: Ic0e12d741cbab292fc214c1eabb48b129af7839b
2014-12-05 16:06:17 -08:00
Jingning Han
eadffb2d6e Fix a motion search skip condition in vp9_pick_inter_mode
Compare the current best mode rate-distortion cost with the skip
threshold to decide if performing motion search.

Change-Id: Ia071824f8dd3b7db485f424692a485a2da6a1a9f
2014-12-05 15:58:36 -08:00
Jingning Han
732d57c2b5 Remove redundant MB_MODE_INFO reset from vp9_pick_mode_inter
Change-Id: I0222f7abc61202f4a83b117bbfb042ada6304562
2014-12-05 15:51:11 -08:00
hkuang
eaa6deee5b Merge "Merge set_prev_mi function into encoder function." 2014-12-05 15:12:50 -08:00
Deb Mukherjee
37448d3e1f Some internal-stats, vp9-highbitdepth bug fixes
Change-Id: I0363d98f6f6558a43276aec48f27dca37c93f5ad
2014-12-05 13:40:50 -08:00
Jingning Han
6ae829088f Merge "Remove redundant vp9_zero in choose_partitioning" 2014-12-05 11:47:58 -08:00
Jingning Han
69a9dc5cd3 Merge "Enable conditional skip path in rd_pick_intra_sby_mode" 2014-12-05 11:25:30 -08:00
Jingning Han
62c7356098 Merge "Use hybrid RD and non-RD coding flow for key frame coding" 2014-12-05 11:25:19 -08:00
Jingning Han
9d88b30854 Remove redundant vp9_zero in choose_partitioning
It makes the overall speed -6 about 2% faster with no compression
performance change.

Change-Id: I680a967b421caa2c5a5cdb821311c4726a2df45a
2014-12-05 10:39:39 -08:00
Jingning Han
74ded4863e Enable conditional skip path in rd_pick_intra_sby_mode
These speed-up features for key frame coding are only turned on
in the settings of hybrid non-RD and RD mode decision. It provides
about 20% speed-up to the hybrid key frame coding at the expense
of certain compression performance loss. For vidyo1, the key frame
coding statistics are changed
9838F, 35.020 dB, 61677 us -> 9920F, 34.834 dB, 47556 us

Overall rtc set compression performance is down by -0.257%.

Change-Id: I0025447fda26bb7855e982955642b5f55d71b51f
2014-12-05 09:36:09 -08:00
Jingning Han
07711e9b27 Use hybrid RD and non-RD coding flow for key frame coding
When block size is below 16x16, the encoder swap from non-RD to
RD mode for key frame coding. This largely brough back the key
frame compression performance. For vidyo1 at 1000 kbps, the key
frame coding statistics are changed

9978F, 34.183 dB, 36807 us -> 9838F, 35.020 dB, 61677 us

As compared to the full RD case
7187F, 34.930 dB, 214470 us

The overall rtc set coding performance (single key frame setting)
is improved by 1.5%.

Change-Id: I78a4ecf025d7b24ec911e85be94e01da05e77878
2014-12-05 09:35:27 -08:00
Yunqing Wang
a3a4a34c60 Merge "vp9_ethread: the tile-based multi-threaded encoder" 2014-12-05 08:23:49 -08:00
Frank Galligan
4c4d7261e4 Fix potential integer overflow.
ioc found a potential integer overflow in the rate control.

This is related to https://code.google.com/p/webm/issues/detail?id=821

Change-Id: Ib6c4acd6e964972f932fce7490592eb134f2b7ea
2014-12-05 08:02:12 -08:00
Paul Wilkins
bb6e47c1c9 Merge "Increase strength of AQ1." 2014-12-05 04:11:43 -08:00
Debargha Mukherjee
15cf55b3ca Merge "Use the RTC optimizations when in high bitdepth mode." 2014-12-04 19:22:27 -08:00
James Zern
b43c27ab6e Merge "vp9_reader: reorder struct members" 2014-12-04 16:08:08 -08:00
Debargha Mukherjee
4bfde1071e Merge "Corrected the renaming of CONFIG_VP9_HIGH ro CONFIG_VP9_HIGHBITDEPTH." 2014-12-04 15:52:35 -08:00
Peter de Rivaz
a306bd8274 Use the RTC optimizations when in high bitdepth mode.
Change 72193 made the encoder behave differently
when configured with and without high bitdepth.
This change means the same algorithm is used for both.

Change-Id: I707a44a94afca773a9e0c2f7ebeeea83030257c5
2014-12-04 15:48:42 -08:00
James Zern
dd26b9a6d0 Merge "Loop filter test fix" 2014-12-04 15:14:20 -08:00
hkuang
62de07c8c6 Merge set_prev_mi function into encoder function.
Change-Id: Ifcf2efbb232ea4cabcdebbe77e0820d121e4a6da
2014-12-04 14:44:23 -08:00
Yunqing Wang
eba9c762a1 vp9_ethread: the tile-based multi-threaded encoder
Currently, VP9 supports column-tile encoding, which allows a frame
to be encoded in multiple column tiles independently. The number of
column tiles are set by encoder option "--tile-columns". This
provides a way to encode a frame in parallel.

Based on previous set of patches, this patch implemented the tile-
based multi-threaded encoder. Each thread processes one or more
tiles.

Usage:
For HD clips:
--tile-columns=2 --threads=1/2/3/4

While using 4 threads, tests showed that the encoder achieved
2.3X - 2.5X speedup at good-quality speed 3, and 2X speedup at
realtime speed 5.

Change-Id: Ied987f8f2618b1283a8643ad255e88341733c9d4
2014-12-04 11:21:34 -08:00
Deb Mukherjee
4f860dba78 Merge "Fixes a missing highbitdepth convolve call bug" 2014-12-04 11:19:59 -08:00
Adrian Grange
9065da983f Merge "Free motion vector array before re-allocating" 2014-12-04 07:08:37 -08:00
Peter de Rivaz
f610f88be4 Corrected the renaming of CONFIG_VP9_HIGH ro CONFIG_VP9_HIGHBITDEPTH.
Change 71789 renamed CONFIG_VP9_HIGH to CONFIG_VP9_HIGHBITDEPTH.
However, one use of CONFIG_VP9_HIGH was missed.

Change-Id: I0ebb9c71380c6d810a25708d15471abf9533e695
2014-12-04 11:01:46 +00:00
Deb Mukherjee
072ed17d1e Loop filter test fix
Related to fixing random number range.

Change-Id: I84d004600e1f6918badf08d08da421b4e8cc4826
2014-12-04 01:26:06 -08:00
Marco
7cb7588b1e Merge "Various updates to vp8." 2014-12-03 18:12:47 -08:00
James Zern
33e61df7d7 Merge "acm_random: provide full range for Rand16()" 2014-12-03 18:05:36 -08:00
Tom Finegan
7339681ee9 Merge "sse2 visual studio build fix" 2014-12-03 18:05:03 -08:00
Deb Mukherjee
70d9dbd818 Fixes a missing highbitdepth convolve call bug
Bug was introduced in https://gerrit.chromium.org/gerrit/#/c/72122/

Change-Id: Idb500ea619a30e7bc50e22fb8ee03be5282f41db
2014-12-03 17:48:50 -08:00
James Zern
615230b810 acm_random: provide full range for Rand16()
the gtest implementation used only returns values between 0 and 2^31-1

+ temporarily disable some tests in fdct8x8_test which misbehave with the
  new range

Change-Id: I45381076f0bea3317cc6728305890e4fd2f2facd
2014-12-03 16:59:40 -08:00
James Zern
276ab21562 Merge "acm_random: provide full range for Rand8()" 2014-12-03 16:58:42 -08:00
Adrian Grange
b56451f488 Merge "Use memset for initialization to 0" 2014-12-03 16:50:39 -08:00
Deb Mukherjee
6615706af2 sse2 visual studio build fix
Change-Id: Id8c8c3be882bcd92afea3ccec6ebdf3f208d28ef
2014-12-03 16:35:26 -08:00