Commit Graph

2649 Commits

Author SHA1 Message Date
Yaowu Xu
9158b8956f Merge "make bsize requirement for SEG_LVL_SKIP explicit" 2013-09-05 08:15:03 -07:00
Jim Bankoski
2e4ca9d1a5 resolve clang warnings : uninitialized vars in vp9_entropy.h
This helps clear out some of the warnings

Change-Id: Ie7ccaca8fd92542386a7f1b257398e1bdf2f55dc
2013-09-04 18:38:41 -07:00
Jim Bankoski
e8feb2932f Merge "wrap non420 loop filter code in macro" 2013-09-04 17:20:53 -07:00
Paul Wilkins
e5deed06c0 Merge "Attempt to fix speed 4" 2013-09-04 17:19:22 -07:00
Yaowu Xu
1ee66933c1 make bsize requirement for SEG_LVL_SKIP explicit
The segment feature SEG_LVL_SKIP requires the prediction unit size
to be at least BLOCK_8X8. This commit makes the requirement to be
explicit. This is to prevent future encoder implementations from
making wrong choices.

Change-Id: I0127f0bd4c66e130b81f0cb0a8d3dbfe3b2da5c2
2013-09-04 16:32:26 -07:00
hkuang
01c4e04424 Speed up idct8x8 by rearrange instructions.
Speed improve from 264% ~ 270% to 280% ~ 300% base on assembly-perf.

Change-Id: I3e2cc818ec14b432204ff43732f39b6438db685d
2013-09-04 15:57:22 -07:00
Yaowu Xu
72872d3d8c Merge "Fixing problem with invalid delta_q reading." 2013-09-04 14:21:30 -07:00
hkuang
3c05bda058 Merge "Add neon optimize vp9_short_iht4x4_add." 2013-09-04 13:35:09 -07:00
hkuang
3b8614a8f6 Add neon optimize vp9_short_iht4x4_add.
Change-Id: I42c497b68ae1ee645b59c9968ad805db0a43e37e
2013-09-04 12:37:58 -07:00
Dmitry Kovalev
890eee3b47 Fixing problem with invalid delta_q reading.
This is a bitstream change but no currently produces videos should
be affected. https://code.google.com/p/webm/issues/detail?id=610

Change-Id: Ic85a6477df6c201cdf7f70f6bd84607b71f4593c
2013-09-04 11:25:43 -07:00
Yaowu Xu
76a437a31b Merge "Replacing init_dequantizer() with setup_plane_dequants()." 2013-09-04 10:42:12 -07:00
Jim Bankoski
872c6d85c0 Merge "speed up inc_mv_component" 2013-09-04 10:35:51 -07:00
Jim Bankoski
bb2313db28 Merge "make vp9 postproc a config option" 2013-09-04 10:35:26 -07:00
Yunqing Wang
9fd2767200 Merge "Use correct bit cost while static-thresh is on" 2013-09-04 10:26:37 -07:00
Jim Bankoski
c3c21e3c14 wrap non420 loop filter code in macro
Change-Id: I62bca0e7a4bffc1a78b750dbb9df9d2378e92423
2013-09-04 10:24:42 -07:00
Jim Bankoski
79401542f7 make vp9 postproc a config option
Vp9 postproc is disabled for now as its not been shown to help and
may be merged with vp8.

Change-Id: I25620d6cd34c6e10331b18c7b5ef7482e39c6057
2013-09-04 10:02:08 -07:00
Jim Bankoski
532179e845 faster accounting of inc_mv
Moves counting of mv branches to where we have a new mv, instead of after
the whole frame is summed.

Change-Id: I945d9f6d9199ba2443fe816c92d5849340d17bbd
2013-09-04 09:47:57 -07:00
Dmitry Kovalev
d6606d1ea7 Replacing init_dequantizer() with setup_plane_dequants().
Change-Id: Ib67e996b4a6dcb6f481889f5a0d84811a9e3c5d1
2013-09-04 09:22:59 -07:00
Jim Bankoski
5dda1d2394 speed up inc_mv_component
Convert mv_class if statements to look up.  re order to avoid ifs...

Change-Id: I76966a21bf517bb1f9a7957c08c476c7bb3e9a63
2013-09-04 07:11:30 -07:00
James Zern
1cf2272347 Merge "Fix intermediate height in convolve_c" 2013-09-03 15:50:33 -07:00
Paul Wilkins
49317cddad Attempt to fix speed 4
Speed 4 fixed partition size. Use fixed size unless it does not
fit inside image, in which case use the largest size that does.

Change-Id: I250f7a80506750dd82ab355721624a1344247223
2013-09-03 17:46:25 +01:00
Jingning Han
010c0ad0eb Merge "Fix 32x32 forward transform SSE2 version" 2013-09-03 08:58:03 -07:00
Scott LaVarnway
948aaab4ca Merge "Improved mb_lpf_horizontal_edge_w_sse2_8" 2013-09-03 05:44:01 -07:00
Jingning Han
3cf46fa591 Fix 32x32 forward transform SSE2 version
This commit fixed the potential overflow issue in the SSE2
implementation of 32x32 forward DCT. It resolved the corrupted
coded frames in the border of scenes.

Change-Id: If87eef2d46209269f74ef27e7295b6707fbf56f9
2013-08-31 18:47:08 -07:00
Yunqing Wang
0ca7855f67 Use correct bit cost while static-thresh is on
While static-thresh is on, we only need to transmit skip
flag if skip = 1. The cost of skip bit is added to the
total rate cost.

Change-Id: I64e73e482bc297eba22907026298a15fa8cc3920
2013-08-30 15:25:13 -07:00
Paul Wilkins
2b9baca4f0 Merge "Added per pixel inter rd hit count stats" 2013-08-30 08:56:01 -07:00
Tero Rintaluoma
e326cecf18 Fix intermediate height in convolve_c
- Intermediate height was not correct i.e. when block size is 4 and
  y_step_q4 is 6. In this case intermediate height was
  (4*6) >> 4 = 1 and vertical interpolation needs two source pixels
  plus 7 extra pixels for taps.
- Also if the current output block is 16x16 and we are using 4x upscaling
  we need only 12 rows after horizontal filtering instead of 16.

  Patch Set 2: Intermediate_height updated after CL 66723
               "Fix bug in convolution functions (filter selection)"

Change-Id: I5a1a1bc2ac9d5edb3a6e0818de618bf318fdd589
2013-08-30 10:31:21 +03:00
Jim Bankoski
1d44fc0c49 Merge "rework filter_block_plane" 2013-08-29 20:11:09 -07:00
Jim Bankoski
bc50961a74 rework filter_block_plane
Change-Id: I55c3b60c4c0f4910d3dfb70e3edaae00cfa8dc4d
2013-08-29 17:00:05 -07:00
Jingning Han
c86c5443eb Merge "Fix overflow issue in SSSE3 32x32 quantization" 2013-08-29 16:49:04 -07:00
Paul Wilkins
1f4bf79d65 Added per pixel inter rd hit count stats
Added some code to output normalized rd hit count stats.
In effect this approximates to the average number of rd
operations/tests per pixel for the sequence.

The results are not quite accurate and I have not bothered
to account for partial SB64s at frame edges and for key frames
However they do give some idea of the number of modes /
prediction methods being tested for each pixel across the
different partition sizes. This indicates how much scope their
is for further gains either by reducing the number of partitions
examined or the modes per partition through heuristics.

Patch 3 moved place where count incremented so partial rd
tests that are aborted with INT_MAX return are also counted.

Example numbers for first 50 frames of Akiyo.
Speed 0 ~84.4 rd operations / pixel
Speed 1 ~28.8
Speed 2 ~11.9

Change-Id: Ib956e787e12f7fa8b12d3a1a2f6cda19a65a6cb8
2013-08-30 00:13:51 +01:00
Deb Mukherjee
b6dbf11ed5 Merge "Adds a speed feature for fast 1-loop forw updates" 2013-08-29 15:54:04 -07:00
James Zern
e83e8f0426 Merge changes Ib1e853f9,Ifd75c809,If3e83404
* changes:
  consistently name VP9_COMMON variables #3
  consistently name VP9_COMMON variables #2
  consistently name VP9_COMMON variables #1
2013-08-29 15:50:56 -07:00
Yaowu Xu
ee961599e1 Merge "Fixed potential overflows" 2013-08-29 15:43:26 -07:00
James Zern
d765df2796 consistently name VP9_COMMON variables #3
stragglers

Change-Id: Ib1e853f9a331b7b66639dc34d79568d84d1930f1
2013-08-29 13:27:41 -07:00
James Zern
aa05321262 consistently name VP9_COMMON variables #2
oci -> cm

Change-Id: Ifd75c809d9cc99034d3c2fccc4653a78b3aec21f
2013-08-29 13:25:58 -07:00
James Zern
924d74516a consistently name VP9_COMMON variables #1
pc -> cm

Change-Id: If3e83404f574316fdd3b9aace2487b64efdb66f3
2013-08-29 13:25:57 -07:00
Dmitry Kovalev
e80bf802a9 Merge "Renaming txfm_size to tx_size." 2013-08-29 12:30:18 -07:00
Jingning Han
abff678866 Fix overflow issue in SSSE3 32x32 quantization
The 32x32 quantization process can potentially have the intermediate
stacks over 16-bit range, thereby causing enc/dec mismatch. This commit
fixes this overflow issue in the SSSE3 implementation, as well as the
prototype, of 32x32 quantization.

This fixes issue 607 from webm@googlecode.

Change-Id: I85635e6ca236b90c3dcfc40d449215c7b9caa806
2013-08-29 11:00:54 -07:00
Yaowu Xu
aaa7b44460 Fixed potential overflows
The two arrays are typically initialized to INT64_MAX, if they are not
filled with valid values before the addition, the values can overflow
and lead to wrong results.

Change-Id: I515de22cf3e8f55af4b74bdb2c8eb821a02d3059
2013-08-29 10:26:52 -07:00
Scott LaVarnway
22dc946a7e Improved mb_lpf_horizontal_edge_w_sse2_8
This patch is a reformatted version of optimizations done by
engineers at Intel (Erik/Tamar) who have been providing
performance feedback for VP9.  For the test clips used (720p, 1080p),
up to 1.2% performance improvement was seen.

Change-Id: Ic1a7149098740079d5453b564da6fbfdd0b2f3d2
2013-08-29 08:30:17 -04:00
Dmitry Kovalev
b71807082c Merge "General code cleanup." 2013-08-28 12:57:49 -07:00
Dmitry Kovalev
db20806710 Merge "Removing unnecessary call to vp9_setup_interp_filters." 2013-08-28 12:31:08 -07:00
Dmitry Kovalev
b62ddd5f8b General code cleanup.
Switching from mi_{width, height}_log2 and b_{width, height}_log2 to
num_8x8_blocks_{wide, high} and num_4x4_blocks_{wide, high}. Removing
redundant code, adding const.

Change-Id: Iaab2207590fd24d0b76999071778d1395dc5cd5d
2013-08-28 12:22:37 -07:00
Deb Mukherjee
e02dc84c1a Adds a speed feature for fast 1-loop forw updates
Incorporates a speed feature for fast forward updates of
coefficients. This feature takes 3 values:
0 - use standard 2-loop version
1 - use a 1-loop version
2 - use a 1-loop version with reduced updates

Results: derfraw300 +0.007% (on speed 0) at feature value = 1
                    -0.160% (on speed 0) at feature value = 2

There is substantial speed up at speeds 2 and above for low
resolution sequences where the entropy updates are a big part
of the overall computations.

Change-Id: Ie96fc50777088a5bd441288bca6111e43d03bcae
2013-08-28 10:56:52 -07:00
Dmitry Kovalev
851a2fd72c Renaming txfm_size to tx_size.
Change-Id: I752e374867d459960995b24d197301d65ad535e3
2013-08-27 19:47:53 -07:00
Jingning Han
eb7acb5524 Merge "Fix buf alignment in sub8x8 comp inter-inter pred" 2013-08-27 19:03:12 -07:00
Dmitry Kovalev
1d3f94efe2 Merge "Adding get_entropy_context function." 2013-08-27 17:02:36 -07:00
Frank Galligan
7d058ef86c Merge "Fix winodws warning." 2013-08-27 15:39:58 -07:00
Frank Galligan
f1560ce035 Fix winodws warning.
Const is not needed on the function parameter.

Change-Id: I38c2a7317cb6f42f70bbddfde9a2cd18d65ceb1c
2013-08-27 15:19:55 -07:00
Dmitry Kovalev
a93992e725 Adding get_entropy_context function.
Moving common code from encoder and decoder to this function.

Change-Id: I60fa643fb1ddf7ebbff5e83b6c4710137b0195ef
2013-08-27 14:17:53 -07:00
hkuang
3a679e56b2 Add neon optimize vp9_short_idct16x16_1_add.
Change-Id: Ib9354c1d975d03e8081df20d50b6a77dfe2dc7e5
2013-08-27 14:00:27 -07:00
hkuang
ce04b1aa62 Merge "Add neon optimize vp9_short_idct8x8_1_add." 2013-08-27 12:10:07 -07:00
Dmitry Kovalev
7b95f9bf39 Renaming BLOCK_SIZE_TYPE to BLOCK_SIZE in the encoder.
Change-Id: I62bb07c377f947cb72fac68add7a6b199e42c6b9
2013-08-27 11:05:08 -07:00
Dmitry Kovalev
ba10aed86d Merge "Using num_8x8_* lookup tables instead of mi_*_log2." 2013-08-27 10:49:36 -07:00
Dmitry Kovalev
12e5931a9a Merge "Using existing functions instead of raw expressions." 2013-08-27 10:33:34 -07:00
Dmitry Kovalev
f77c6973a1 Merge "Cleaning up decode_block_intra function." 2013-08-27 10:17:56 -07:00
Dmitry Kovalev
f389ca2acc Merge "Cleaning up model_rd_for_sb_y_tx." 2013-08-27 10:17:10 -07:00
Dmitry Kovalev
bfebe7e927 Merge "Renaming BLOCK_SIZE_TYPE to BLOCK_SIZE in the common/decoder." 2013-08-27 10:15:21 -07:00
Dmitry Kovalev
78e670fcf8 Merge "Renaming D27 to D207." 2013-08-27 10:03:57 -07:00
Jingning Han
2d6aadd7e2 Fix buf alignment in sub8x8 comp inter-inter pred
This commit resolved a mis-alignment issue in compound inter-inter
prediction of sub8x8. This patch follows solution from dkovalev@.

Change-Id: I3cc0cf7e55b84110e0c42ef4b2e6ca7ac3f8f932
2013-08-27 09:28:05 -07:00
Yaowu Xu
45125ee573 Merge "fixed the reading too many bytes" 2013-08-27 09:09:18 -07:00
Yaowu Xu
9482c07953 fixed the reading too many bytes
In subpel_avg_variance functions, code similar to the following

punpkldq m2, [addr]

actually reads 8 bytes. For functions that are supposed to work on
buffers only have less 8 bytes a line, this caused valgrind error
of reading uninitialized memory.

Change-Id: I2a4c079dbdbc747829bd9e2ed85f0018ad2a3a34
2013-08-27 08:39:20 -07:00
Dmitry Kovalev
44b7854c84 Removing unnecessary call to vp9_setup_interp_filters.
vp9_setup_interp_filters before each inter block decoding, it is not
necessary to call it just before the whole frame decoding.

Change-Id: Id1b0ee62f987474e27eafba0013a4896b492c400
2013-08-26 17:25:49 -07:00
hkuang
36e9b82080 Add neon optimize vp9_short_idct8x8_1_add.
Change-Id: I0b15d5e3b0eb97abb9ab5ec08e88b61f8723aaf4
2013-08-26 16:28:57 -07:00
hkuang
ba8fc71979 Merge "Add neon optimize vp9_short_idct4x4_1_add." 2013-08-26 16:26:38 -07:00
Dmitry Kovalev
657ee2d719 Cleaning up model_rd_for_sb_y_tx.
Removing references to plane_block_width and plane_block_height (we are
going to delete the latter ones).

Change-Id: I7982da4d373aebb54d2209dc8886f6192df4d287
2013-08-26 16:18:28 -07:00
hkuang
69384f4fad Add neon optimize vp9_short_idct4x4_1_add.
Change-Id: I6ecb5c4a1a472feb8e84e9f3352b536d5e28a4a5
2013-08-26 15:55:16 -07:00
Dmitry Kovalev
242460cb66 Cleaning up decode_block_intra function.
Change-Id: Ia41ea5d526d15fcbc9b56d74079593cf8b2fdf66
2013-08-26 15:24:12 -07:00
Dmitry Kovalev
b25589c6bb Using num_8x8_* lookup tables instead of mi_*_log2.
Change-Id: I8a246b3d056c98be614d05a90bc261e2441ffc10
2013-08-26 14:22:54 -07:00
Yaowu Xu
4505e8accb Merge "Fix the reading of too many input pixels" 2013-08-26 14:01:50 -07:00
Paul Wilkins
aa823f8667 Merge "Changes to adaptive inter rd thresholds." 2013-08-26 12:48:11 -07:00
Yaowu Xu
6c5433c836 Fix the reading of too many input pixels
in VP9_get4x4var_mmx

Change-Id: I4b4a8f45f25ebdfad281f169cc87aba5e2d6f227
2013-08-26 12:35:27 -07:00
Paul Wilkins
642696b678 Merge "Limit Key frame Intra modes checks." 2013-08-26 12:34:56 -07:00
Dmitry Kovalev
45870619f3 Renaming BLOCK_SIZE_TYPE to BLOCK_SIZE in the common/decoder.
Adding temporary "typedef BLOCK_SIZE BLOCK_SIZE_TYPE" which will go away
after encoder's patch.

Change-Id: I06ec6a6f079401439843ec981d1496234fd7775c
2013-08-26 11:33:16 -07:00
Jingning Han
4681197a58 Merge "Temporarily disable SSSE3 quant_32x32" 2013-08-26 11:19:53 -07:00
Dmitry Kovalev
5eed6e2224 Merge "Removing redundant calls to clamp_mv2." 2013-08-26 10:48:37 -07:00
Jingning Han
166dc85bed Temporarily disable SSSE3 quant_32x32
Make the current head working properly, while working on fixing an
issue in the SSSE3 implementation of 32x32 quantization.

Change-Id: Ic029da3fd7f1f5e58bc641341cbd226ec49a16bc
2013-08-26 10:45:59 -07:00
James Zern
c8ba8c513c cosmetics: strip 'VP9_' from defines in vp9 only code
Change-Id: I481d9bb2fa3ec72b6a83d5f04d545ad8013f295c
2013-08-23 19:16:49 -07:00
James Zern
2c6ba737f8 Merge "vp9: remove unnecessary wait w/threaded loopfilter" 2013-08-23 18:52:10 -07:00
Dmitry Kovalev
50ee61db4c Renaming D27 to D207.
I've already renamed d27_predictor to d207_predictor but forgot about the
corresponding constant.

Change-Id: Id312aa80fc5b5a1ab8a709a33418a029552a6857
2013-08-23 17:33:48 -07:00
Dmitry Kovalev
480dd8ffbe Using existing functions instead of raw expressions.
Change-Id: Ifa50b04bac1a6ff2abef989073cbf1f37a89eb50
2013-08-23 17:26:53 -07:00
Dmitry Kovalev
e6c435b506 Merge "Cleanup in mvref_common.{h, c}." 2013-08-23 17:09:49 -07:00
Dmitry Kovalev
7194da2167 Merge "Fixing display size setting problem." 2013-08-23 17:08:51 -07:00
Yaowu Xu
13930cf569 Limit mv range to be based on partition size
Previous change c4048dbd limits the mv search range assuming max block
size of 64x64, this commit change the search range using actual block
size instead.

Change-Id: Ibe07ab02b62bf64bd9f8675d2b997af20a2c7e11
2013-08-23 15:43:57 -07:00
Dmitry Kovalev
cd2cc27af1 Removing redundant calls to clamp_mv2.
We could avoid calling clamp_mv2 because it has been already called
inside vp9_find_best_ref_mvs function.

Change-Id: I08edeaf3e11e98c19e67b9711b2523ca5fb1416e
2013-08-23 15:18:35 -07:00
Yaowu Xu
8e04257bc5 Merge "Added border extension" 2013-08-23 14:43:58 -07:00
Adrian Grange
78debf246b Merge "Fix bug in convolution functions (filter selection)" 2013-08-23 13:41:47 -07:00
Dmitry Kovalev
fb481913f0 Merge "Removing useless calls to setup_{pre, dst}_planes." 2013-08-23 13:37:32 -07:00
Dmitry Kovalev
11e3ac62a5 Fixing display size setting problem.
Fix of https://code.google.com/p/webm/issues/detail?id=608. We could have
used invalid display size equal to the previous frame size (not to the
current frame size).

Change-Id: I91b576be5032e47084214052a1990dc51213e2f0
2013-08-23 13:12:46 -07:00
Dmitry Kovalev
21d8e8590b Cleanup in mvref_common.{h, c}.
Making code more compact, adding consts, removing redundant arguments,
adding do/while(0) for macros.

Change-Id: Ic9ec0bc58cee0910a5450b7fb8cfbf35fa9d0d16
2013-08-23 12:00:30 -07:00
Yaowu Xu
656632b776 Added border extension
To the source buffer to be encoded as an alt ref frame. This is to fix
the problem of using uninitialized memory in encoder.

See https://code.google.com/p/webm/issues/detail?id=605

Change-Id: I97618a2fc207e08abcf5301b734aa9e3ad695e2c
2013-08-23 11:31:28 -07:00
Adrian Grange
3f10831308 Fix bug in convolution functions (filter selection)
(In response to Issue 604:
 https://code.google.com/p/webm/issues/detail?id=604)

There were bugs in the convolution code for two cases:

1. Where the filter table was assumed to be aligned to a
   256 byte boundary. The offset of the pixel in the
   source buffer was computed incorrectly.

2. Where no such alignment assumption was made. An
   incorrect address for the filter table base was used.

To fix both problems, I now assume that the filter table is
256-byte aligned and modify the pixel offset calculation to
match.

A later patch should remove the restriction that the filter
table is aligned to a 256-byte boundary.

There was also a bug in the ConvolveTest unit test
(convolve_test.cc).

(Bug & initial fix suggestion submitted by Tero Rintaluoma
and Sami Pietilä).

Change-Id: I71985551e62846e55e40de9e7e3959d4805baa82
2013-08-23 11:16:08 -07:00
Dmitry Kovalev
1c159c470a Merge "Checking scale factors on access." 2013-08-23 11:05:17 -07:00
hkuang
b85367a608 Merge "Optimise idct4x4: rearrange the instructions a bit to improve instruction scheduling." 2013-08-23 10:08:43 -07:00
Paul Wilkins
aa5b67add0 Changes to adaptive inter rd thresholds.
Values now carried over frame to frame.
Change to algorithm for decreasing threshold after
a hit and to max threshold (now based on speed)

Removed some old commented out code relating to
VP8 adaptive thresholds.

The impact of these changes tested on Akiyo (50 frames)
and measured in terms of unit rd hits is as follows:

Speed 0 84.36 -> 84.67
Speed 1 29.48 -> 22.22
Speed 2 11.76 -> 8.21
Speed 3 12.32 -> 7.21

Encode speed impact is broadly in line with these.

Change-Id: I5b886efee3077a11553fa950d796fd6d00c8cb19
2013-08-23 16:18:45 +01:00
Paul Wilkins
f76f52df61 Limit Key frame Intra modes checks.
Most of the focus so far has been on inter frames.

At high speed settings the key frame is now taking a high %
of the cycles.

This patch puts in some masking to reduce the number
of INTRA modes searched during key frame coding (as already
happens for inter frames) at higher speed settings

TODO: Develop this further with either adaptive rd thresholds
when choosing which intra modes to consider or some other
heuristic.

Impact.
At high speed settings on some clips the key frame was starting
to dominate. In a coding of the first 50 frames of AKIYO at speed
2 limiting the key frame intra modes to DC or TM_PRED resulted in
~30% overall speedup. For Bus the number was lower at ~4-5%.

Change-Id: I7bde68aee04995f9d9beb13a1902143112e341e2
2013-08-23 16:10:30 +01:00
Jingning Han
9655c2c7a6 Merge "Fix rectangular partition check flag" 2013-08-22 18:59:18 -07:00
Dmitry Kovalev
33104cdd42 Merge "vp9_encodeframe.c cleanup." 2013-08-22 18:07:35 -07:00
James Zern
711aff9d9d Merge "vp9/encoder: fix last_frame_seg_map mem leak" 2013-08-22 18:04:03 -07:00