Yaowu Xu [Thu, 16 Jul 2015 22:40:22 +0000 (15:40 -0700)]
Correctly report "Unsupported bitstream profile"
For vp9 decoder build without profile 2 and profile 3 support, this
commit changes to report error "Unsupported bitstream profile" for
input streams in profile 2 or 3, rather than other misleading error
information.
In addition, one of the invalid files in unit tests is actually coded
profile 2, this commit makes it tested only when the decoder is built
with vp9-highbitdepth.
This fixes issue #1028.
Change-Id: I8b6c1210787c8f89c703a546687dcf973ac20fc0
Marco [Thu, 16 Jul 2015 19:38:26 +0000 (19:38 +0000)]
Merge "Dynamic resize for 1 pass: update of golden frame."
Marco [Thu, 16 Jul 2015 01:00:03 +0000 (18:00 -0700)]
Dynamic resize for 1 pass: update of golden frame.
In aq-mode=3 under a resizing action (i.e., resize_pending != 0),
force an update of the golden reference frame.
Change-Id: I14806f6db71b5f8c827678cc5e1fc913c138a9a4
Frank Galligan [Thu, 16 Jul 2015 05:45:15 +0000 (05:45 +0000)]
Merge "Add vp9_int_pro_col_neon."
Jingning Han [Wed, 15 Jul 2015 22:41:52 +0000 (22:41 +0000)]
Merge "Add vpx_dsp_common.h file"
Jingning Han [Wed, 15 Jul 2015 22:41:10 +0000 (22:41 +0000)]
Merge "Remove redundant header files in vp9_loopfilter_filers.c"
Marco [Wed, 15 Jul 2015 21:43:05 +0000 (21:43 +0000)]
Merge "Fix to resize logic for 1 pass mode."
Jingning Han [Wed, 15 Jul 2015 20:02:01 +0000 (13:02 -0700)]
Add vpx_dsp_common.h file
Move the clamp functions to vpx_dsp_common.h file. Clear out the
dependency of vp9_loopfilter_filters.c on vp9_common.h file.
Change-Id: I9c4b928bcd7f597106b5aa96354356d3775a3431
Jingning Han [Wed, 15 Jul 2015 19:41:31 +0000 (12:41 -0700)]
Remove redundant header files in vp9_loopfilter_filers.c
This cleans out the unnecessary dependency on vp9 codec-specific
data structures.
Change-Id: Iadbe431174a0f9bf9423f39ab854fc18be554bea
Marco [Wed, 15 Jul 2015 18:45:52 +0000 (11:45 -0700)]
Adjust some logic for dynamic_resize 1 pass mode.
Use drop_frames_water_mark for threshold on buffer underflow,
and change threshold for resize down.
Change-Id: I2de19adce50abe9bcdc0b107528cec8cc1857fcc
Frank Galligan [Wed, 15 Jul 2015 16:04:28 +0000 (09:04 -0700)]
Add vp9_int_pro_col_neon.
BUG=https://code.google.com/p/webm/issues/detail?id=1023
Change-Id: I212a1d67b23ce3b5ce08800de369b25b9e375e7d
Marco [Tue, 14 Jul 2015 17:01:45 +0000 (10:01 -0700)]
Fix to source scaling for dynamic_resize.
The fast scaling for 1 pass mode was being used only on the
first frame after resizing event (because resize_scale_num/den
is set to 1 and only changed for first frame following resize event).
Change-Id: I723b63e21823eb858f25f5662d2bbe4f1842e61f
Marco [Tue, 14 Jul 2015 23:22:10 +0000 (16:22 -0700)]
Fix to resize logic for 1 pass mode.
Proper use/update of resize_state and resize_pending to constrain
the total amount of downsizing to be at most one scale down, for now.
Change-Id: Id18fc32499f2fbdbec16728dcdc9e4eac09098f0
Alex Converse [Tue, 14 Jul 2015 22:11:46 +0000 (22:11 +0000)]
Merge "Add an SSE2 version of vp9_iwht4x4_16_add"
Alex Converse [Mon, 13 Jul 2015 18:12:45 +0000 (11:12 -0700)]
Add an SSE2 version of vp9_iwht4x4_16_add
Roughly half as many cycles as plain C.
Change-Id: I8c16c29940b76d54ee7e4fb874c328ce90bff5d4
paulwilkins [Tue, 14 Jul 2015 18:24:12 +0000 (18:24 +0000)]
Merge "Add extra resize trigger for frames above maximum allowed size."
Debargha Mukherjee [Fri, 10 Jul 2015 16:49:17 +0000 (09:49 -0700)]
Fixes part of merge regression from adding arf parameters.
From Change Ibf0c30b72074b3f71918ab278ccccc02a95a70a0
There is still an issue relating to one animated test clip with repeat
patterns where this change effectively increase the default maximum
arf interval by +1. This can be examined seperately.
Change-Id: Idd01d5480fc45202d8a059a0c3afc0997cc5bdd1
Jingning Han [Tue, 14 Jul 2015 16:22:40 +0000 (16:22 +0000)]
Merge "Refactor intra block prediction and reconstruction process"
Jingning Han [Tue, 14 Jul 2015 16:22:19 +0000 (16:22 +0000)]
Merge "Refactor intra block prediction function"
Jingning Han [Mon, 13 Jul 2015 18:49:57 +0000 (11:49 -0700)]
Refactor intra block prediction and reconstruction process
Flaten the intra block decoding process. It removes the legacy
foreach_transformed_block use in the decoder. This saves cycles
spent on retrieving the transform block position.
Change-Id: I21969afa50bb0a8ca292ef72f3569f33f663ef00
Jingning Han [Mon, 13 Jul 2015 17:40:01 +0000 (10:40 -0700)]
Refactor intra block prediction function
This commit simplifies the intra block boundary condition logic.
It removes the block index from the argument set.
Change-Id: If00142512eb88992613d6609356dfd73ba390138
Marco [Mon, 13 Jul 2015 19:06:25 +0000 (19:06 +0000)]
Merge "Dynamic resize for real-time: source scaling"
Yaowu Xu [Mon, 13 Jul 2015 16:27:09 +0000 (16:27 +0000)]
Merge "Revert "Add an SSE2 version of vp9_iwht4x4_16_add.""
Yaowu Xu [Mon, 13 Jul 2015 16:26:27 +0000 (16:26 +0000)]
Revert "Add an SSE2 version of vp9_iwht4x4_16_add."
This reverts commit
f8d35016408f3957c67945160d65be467ca97fdc.
Change-Id: If8c7af403c091b7fb447a6f0c73fecdbccbc51b3
Jim Bankoski [Mon, 13 Jul 2015 14:39:00 +0000 (14:39 +0000)]
Merge "Revert "Fill buffer speed up""
Jim Bankoski [Mon, 13 Jul 2015 13:47:46 +0000 (13:47 +0000)]
Revert "Fill buffer speed up"
This reverts commit
9b4f9f45eee4d63cef3cd10f24923ed0bdd5ab7b.
Change-Id: I23545ac8c7464127f7466fc6a58de517874fe0cf
Parag Salasakar [Mon, 13 Jul 2015 01:16:42 +0000 (01:16 +0000)]
Merge "mips msa vp8 loop filter optimization"
Marco [Thu, 9 Jul 2015 01:01:01 +0000 (18:01 -0700)]
Dynamic resize for real-time: source scaling
Use faster scaling on source.
Change-Id: I968df97239a86834c96126b86832d3d6d0875a53
Jim Bankoski [Thu, 9 Jul 2015 20:26:22 +0000 (20:26 +0000)]
Merge "Fill buffer speed up"
Jim Bankoski [Thu, 9 Jul 2015 18:41:30 +0000 (11:41 -0700)]
Fill buffer speed up
Eliminates the byte by byte read from bool decoder, by reading
in a size_t and then shifting it into place.
Change-Id: Id89241977103fc3b973e4ed172a5cbf246998e5d
paulwilkins [Thu, 9 Jul 2015 18:34:39 +0000 (18:34 +0000)]
Merge "Changes to use of rectangular partitions."
Yaowu Xu [Thu, 9 Jul 2015 17:49:16 +0000 (17:49 +0000)]
Merge "Remove clamp operations."
Yaowu Xu [Thu, 9 Jul 2015 17:49:06 +0000 (17:49 +0000)]
Merge "Clean out more MSVC warnings"
Scott LaVarnway [Thu, 9 Jul 2015 17:16:21 +0000 (17:16 +0000)]
Merge "Eliminate num_8x8 and num_4x4 width/height lookups"
Alex Converse [Thu, 9 Jul 2015 16:57:01 +0000 (16:57 +0000)]
Merge "Add an SSE2 version of vp9_iwht4x4_16_add."
paulwilkins [Thu, 9 Jul 2015 16:38:37 +0000 (16:38 +0000)]
Merge "Change speed and rd features for formatting bars."
Scott LaVarnway [Thu, 9 Jul 2015 12:30:46 +0000 (05:30 -0700)]
Eliminate num_8x8 and num_4x4 width/height lookups
Also some log2 lookups.
Pass in 8x8 block width/height and log2 num4x4s instead.
Change-Id: I8ea9a1ec1e0bbab23f8ba556954a1b5433f4d613
Yaowu Xu [Wed, 8 Jul 2015 01:29:14 +0000 (18:29 -0700)]
Remove clamp operations.
The clamp calls with INT32_MIN and INT32_MAX have no effect at all on
int values passed in, therefore this commit removes those effectless
clamps and also adds more const intermediate results to make the code
more readable.
Change-Id: I66d8811f58bb74ec31cbec9a6c441983a662352e
Jingning Han [Wed, 8 Jul 2015 21:26:00 +0000 (14:26 -0700)]
Format fixes in vp9_encodeframe.c and vp9_encodemb.c
Change-Id: Ib1303dac9043ab1b1f8fce54611cf4ea8a208038
Jingning Han [Wed, 8 Jul 2015 16:15:39 +0000 (09:15 -0700)]
Refactor transform block loop for inter mode decoding
Rework the inter mode transform block decoding loop. Replace the
block index with the row and col index as the input argument. It
saves function call to compute the row and col index according to
the block index and overall block size, and many if statements
associated with the transform block position relative to the coding
block. For the test bit-stream pedestrian_area 1080p at 5 Mbps,
the decoding speed goes up from 81.13 fps to 81.92 fps.
Note that the intra coded block decoding needs more refactoring
work than the inter ones. So keep it using foreach_transforme_block
as for now.
Change-Id: I5622bdae7be28ed5af96693274057f55ba9b4fb4
Yaowu Xu [Tue, 7 Jul 2015 21:22:07 +0000 (14:22 -0700)]
Clean out more MSVC warnings
Change-Id: I1bab0c104df2ec4825d050cd516e26ab635a7b3e
Alex Converse [Tue, 7 Jul 2015 22:27:54 +0000 (15:27 -0700)]
Add an SSE2 version of vp9_iwht4x4_16_add.
80% fewer cycles than C
Change-Id: I841bde1e268ddd33ae2ee75eee94737a400e2cde
Alex Converse [Wed, 8 Jul 2015 20:42:34 +0000 (20:42 +0000)]
Merge "Don't allocate dqcoeff in MACROBLOCKD."
Alex Converse [Wed, 8 Jul 2015 04:19:21 +0000 (21:19 -0700)]
Don't allocate dqcoeff in MACROBLOCKD.
The encoder gets its dqcoeff from the context tree. In the decoder move
it to directly after MACROBLOCKD.
Change-Id: I46c9b76f26956a360d17de0b26ecb994dae34ecb
Jingning Han [Wed, 8 Jul 2015 19:28:24 +0000 (19:28 +0000)]
Merge "Refactor inverse_transform_block argument list"
Jingning Han [Wed, 8 Jul 2015 19:27:21 +0000 (19:27 +0000)]
Merge "Reset dqcoeff[0] only if eob is 1"
Frank Galligan [Wed, 8 Jul 2015 18:15:38 +0000 (18:15 +0000)]
Merge "VP9_LPF_VERTICAL_16_DUAL_SSE2 optimization"
Johann [Wed, 8 Jul 2015 17:50:32 +0000 (17:50 +0000)]
Merge "Use 'unsigned long' for _BitScanReverse parameter"
paulwilkins [Tue, 30 Jun 2015 09:18:42 +0000 (10:18 +0100)]
Add extra resize trigger for frames above maximum allowed size.
Even if the recode loop is not enabled for the current frame type
trap the case where the projected size of a a frame is above the
maximum allowed in recode_loop_test()
Change-Id: I453004694b8f8699e3c2a83252e9f83adccdda4e
paulwilkins [Wed, 24 Jun 2015 10:36:51 +0000 (11:36 +0100)]
Changes to use of rectangular partitions.
Changes to allow more use of rectangular partitions at
speeds 1 and 2 for content classed by the first pass as
animation and for blocks near the active image edge.
This has quite a big impact in quality for the animated
test sequence but also hurts encode speed for speed 2.
For other content types the impact on both speed and
quality is small.
Added some plumbing for detection of internal vertical
image edges.
Change-Id: I3fc48de2349f8cb87946caaf0b06dbb0ea261a9a
paulwilkins [Wed, 24 Jun 2015 10:36:51 +0000 (11:36 +0100)]
Change speed and rd features for formatting bars.
Change speed features / behavior for split mode when there
is an internal active edge (e.g. formatting bars).
Remove some threshold constraints in rd code near the active
edge of the image.
Add some plumbing for left and right active edge detection.
Patch set 5. Limit rd pass through for sub 8x8 to internal active edges.
This takes away any speed penalty for most clips but keeps the enhanced
edge coding for the more critical case of internal image edges
Change-Id: If644e4762874de4fe9cbb0a66211953fa74c13a5
Jingning Han [Tue, 7 Jul 2015 22:32:27 +0000 (15:32 -0700)]
Refactor inverse_transform_block argument list
Replace block index with transform type in the argument list. This
allows to save an extra fetch to the prediction mode. For pedestrian
area 1080p coded at 5 Mbps with single tile, the average decoding
speed goes up from 80.55 fps (before the refactoring series) to
81.13 fps.
Change-Id: Icbebf84ce63c19c0c92f3690ed201f6c3eab7881
Parag Salasakar [Wed, 8 Jul 2015 07:11:00 +0000 (12:41 +0530)]
mips msa vp8 loop filter optimization
average improvement ~2x-4x
Change-Id: I20c4f900ef95d99b18f9cf4db592cd352c2212eb
James Zern [Wed, 8 Jul 2015 01:27:11 +0000 (01:27 +0000)]
Merge "vp9_entropymv: remove vp9_get_mv_mag()"
Johann [Wed, 8 Jul 2015 00:45:06 +0000 (17:45 -0700)]
Use 'unsigned long' for _BitScanReverse parameter
New clang warnings in chromium reveal an incompatible pointer being
passed.
Change-Id: I0d893993726ecf899f467c16175bfb8c76329731
https://code.google.com/p/chromium/issues/detail?id=504696
Frank Galligan [Wed, 8 Jul 2015 00:16:02 +0000 (00:16 +0000)]
Merge "Add vp9_int_pro_row_neon."
Johann [Tue, 7 Jul 2015 23:57:17 +0000 (23:57 +0000)]
Merge "Move sub pixel variance to vpx_dsp"
Jingning Han [Tue, 7 Jul 2015 23:09:39 +0000 (23:09 +0000)]
Merge "Rework scan order fetch logic for decoder"
Johann [Fri, 5 Jun 2015 16:54:19 +0000 (09:54 -0700)]
Move sub pixel variance to vpx_dsp
Change-Id: I66bf6720c396c89aa2d1fd26d5d52bf5d5e3dff1
Marco [Tue, 7 Jul 2015 22:47:43 +0000 (22:47 +0000)]
Merge "Update to speed 5 non-rd mode partition search."
Jingning Han [Tue, 7 Jul 2015 22:35:44 +0000 (22:35 +0000)]
Merge "Add vp9_ prefix to init_macroblockd"
Jingning Han [Tue, 7 Jul 2015 22:35:26 +0000 (22:35 +0000)]
Merge "Reduce dqcoeff array size in decoder"
Jingning Han [Tue, 7 Jul 2015 22:18:30 +0000 (15:18 -0700)]
Reset dqcoeff[0] only if eob is 1
If only the first dequantized coefficient is non-zero, reset
dqcoeff[0] to zero directly.
Change-Id: I0197ba72028a8ec436f0b1b9abcc1c0ae5d70abe
Jingning Han [Tue, 7 Jul 2015 20:52:56 +0000 (13:52 -0700)]
Rework scan order fetch logic for decoder
Save redundant call for getting prediction mode to obtain scan
order for detokenization.
Change-Id: I0683ef119f1579d1261ed5d59052a1745b68ef6f
Marco [Tue, 7 Jul 2015 21:34:08 +0000 (21:34 +0000)]
Merge "Update to vpx_temporal_svc_encoder:"
Jingning Han [Tue, 7 Jul 2015 20:42:18 +0000 (20:42 +0000)]
Merge "Unify subtract function used in VP8/9"
Yaowu Xu [Tue, 7 Jul 2015 19:28:56 +0000 (19:28 +0000)]
Merge "Allows using optimzed version vp9_fdct8x8"
Johann [Tue, 7 Jul 2015 19:14:32 +0000 (19:14 +0000)]
Merge "Remove comment about disabled tests"
Jingning Han [Tue, 7 Jul 2015 19:00:01 +0000 (12:00 -0700)]
Add vp9_ prefix to init_macroblockd
Change-Id: I202d4924e627eec94838741df004ed9259d38b88
Marco [Tue, 7 Jul 2015 18:43:15 +0000 (11:43 -0700)]
Update to speed 5 non-rd mode partition search.
If the pre-selected partition size (from variance partition) is
32x32, also apply nonrd partition search for 32x32 and 16x16 size.
Overall small positive gain in metrics, average ~1%.
Some visual improvement, for lower resolutions.
Change-Id: I69cb425bda94f7d13d34c451ab30e9276335a30e
Jingning Han [Tue, 7 Jul 2015 18:36:05 +0000 (11:36 -0700)]
Reduce dqcoeff array size in decoder
The decoding process handles detokenization and reconstruction per
transform block sequentially. There is no need to offset the dqcoeff
buffer according to the transform block index. This allows to
reduce the memory spill and improve cache performance.
Change-Id: Ibb8bfe532a7a08fcabaf6d42cbec1e986901d32d
Parag Salasakar [Tue, 7 Jul 2015 18:14:18 +0000 (18:14 +0000)]
Merge "mips msa vp8 idct optimization"
Marco [Tue, 7 Jul 2015 18:07:10 +0000 (11:07 -0700)]
Update to vpx_temporal_svc_encoder:
Add set_tune_content control setting.
Change-Id: I8f7af8e8e9bc35a2d1546c5a0b2dd14be3cce750
Yaowu Xu [Tue, 7 Jul 2015 17:28:42 +0000 (10:28 -0700)]
Allows using optimzed version vp9_fdct8x8
Change-Id: I59cecb7178a93cdee7ad535fa996ef0caa6e988c
Jingning Han [Mon, 6 Jul 2015 23:52:24 +0000 (16:52 -0700)]
Unify subtract function used in VP8/9
This commit replaces the vp8_ prefixed subtract function with the
common vpx_subtract_block function. It removes redundant SIMD
optimization codes and unit tests.
Change-Id: I42e086c32c93c6125e452dcaa6ed04337fe028d9
Debargha Mukherjee [Tue, 7 Jul 2015 14:25:34 +0000 (14:25 +0000)]
Merge "Adds macros for new encoder controls"
paulwilkins [Tue, 7 Jul 2015 12:44:27 +0000 (12:44 +0000)]
Merge "Alter partition search at image edge."
paulwilkins [Tue, 7 Jul 2015 08:44:33 +0000 (08:44 +0000)]
Merge "Error score recalibration for inactive regions."
paulwilkins [Tue, 7 Jul 2015 08:44:15 +0000 (08:44 +0000)]
Merge "ARF Boost correction for inactive regions."
Parag Salasakar [Thu, 2 Jul 2015 12:38:06 +0000 (18:08 +0530)]
mips msa vp8 idct optimization
average improvement ~2x-5x
Change-Id: I19e82f78772993bcd67fcf975fe180232172f86d
James Zern [Tue, 7 Jul 2015 03:49:28 +0000 (20:49 -0700)]
vp9_entropymv: remove vp9_get_mv_mag()
inline the code directly in read_mv_component(), the only place where it
was being used; this removes a function call in a hot function
Change-Id: I66f99c0c9ce3bc310101dbca4a470f023cc6fb55
Parag Salasakar [Tue, 7 Jul 2015 02:10:21 +0000 (02:10 +0000)]
Merge "mips msa vpx_dsp sadx3 sadx8 optimization"
James Zern [Tue, 7 Jul 2015 00:48:57 +0000 (00:48 +0000)]
Merge "inline vp9_reader_has_error()"
Johann [Tue, 30 Jun 2015 16:23:23 +0000 (09:23 -0700)]
Remove comment about disabled tests
The tests were re-enabled in
1fe643cf03f45198e5a033743d7c4dc027b8dfd9
Change-Id: I17b45a4bbe720a56605aed338446ff77df71abad
Debargha Mukherjee [Mon, 6 Jul 2015 23:07:53 +0000 (16:07 -0700)]
Adds macros for new encoder controls
To aid version management for integration with ffmpeg by use
of:
#ifdef VPX_CTRL_<CTRL_ID>
...
#endif
Change-Id: If550e06de4d3aa3685881f312ce6a86fa9de083b
James Zern [Mon, 6 Jul 2015 22:52:38 +0000 (22:52 +0000)]
Merge "vp9_variance*.c: make static tables const"
James Zern [Mon, 6 Jul 2015 22:43:01 +0000 (22:43 +0000)]
Merge "mips msa vp9 subpel variance optimization"
Jingning Han [Mon, 6 Jul 2015 22:39:26 +0000 (22:39 +0000)]
Merge "Move subtract functions from vp9 to vpx_dsp"
James Zern [Mon, 6 Jul 2015 22:04:37 +0000 (15:04 -0700)]
vp9_variance*.c: make static tables const
Change-Id: Ia5044d13c09685c401191fe87fbf90d36203aadd
Parag Salasakar [Sun, 5 Jul 2015 16:14:06 +0000 (21:44 +0530)]
mips msa vp9 subpel variance optimization
Change-Id: If88401bf8c5d8ee58200278734d7a5058d1585d0
Debargha Mukherjee [Mon, 6 Jul 2015 21:36:38 +0000 (21:36 +0000)]
Merge "Expose params min-gf-interval/max-gf-interval"
James Zern [Mon, 6 Jul 2015 21:36:36 +0000 (21:36 +0000)]
Merge "remove vp9_get_interp_kernel()"
James Zern [Thu, 2 Jul 2015 02:01:58 +0000 (19:01 -0700)]
remove vp9_get_interp_kernel()
expose filter_kernels[] and do the table lookup directly
Change-Id: I0b10bff0327c3e01a723736141a9ffd377cd3d20
Debargha Mukherjee [Wed, 1 Apr 2015 23:39:06 +0000 (16:39 -0700)]
Expose params min-gf-interval/max-gf-interval
Adds two new vp9 parameters --min-gf-interval and --max-gf-interval
to enable testing based on frequency of alt-ref frames.
Also adds a unit-test to test enforcement of min-gf-interval.
For both these parameters the default value is 0, which indicates
they are picked by the encoder, based on resolution and framerate
considerations. If they are greater than zero, the specified
parameter is honored.
(Additional note by paulwilkins)
Note that there is a slight oddity in that key frames are also GFs and
considered part of GF only group. However they are treated as not
being part of an arf group because for arf groups the previous GF is
assumed to be the terminal or overlay frame for the previous group.
(end note)
Change-Id: Ibf0c30b72074b3f71918ab278ccccc02a95a70a0
Jingning Han [Mon, 6 Jul 2015 16:33:27 +0000 (09:33 -0700)]
Move subtract functions from vp9 to vpx_dsp
Factor out the subtraction operator as common function.
Change-Id: I526e703477c6a290e0e3e3c8898f8bb1ca82779b
Jingning Han [Mon, 6 Jul 2015 17:01:29 +0000 (17:01 +0000)]
Merge "Rename vpx_thread to vpx_util"
James Zern [Fri, 3 Jul 2015 18:44:00 +0000 (18:44 +0000)]
Merge "Revert "Correct the inter prediction coordinate...""
hkuang [Wed, 1 Jul 2015 23:22:47 +0000 (16:22 -0700)]
Revert "Correct the inter prediction coordinate..."
Change in
92b199061a13173d29247257f4ef906d38b012bf leads to frame parallel
decode failure in extreme case.
addresses issue #1010
Change-Id: I4fa488dac8e8c584f5eef4cae1640a579130d387
James Zern [Thu, 2 Jul 2015 21:07:31 +0000 (21:07 +0000)]
Merge "Revert "mips msa vp9 subpel variance optimization""
James Zern [Thu, 2 Jul 2015 21:07:22 +0000 (21:07 +0000)]
Merge "Revert "mips msa vp9 avg subpel variance optimization""