Dmitry Kovalev [Thu, 9 Jan 2014 01:40:35 +0000 (17:40 -0800)]
Adding {get, set}_rate_correction_factor() functions.
Change-Id: Ib3212832953a3445fc5f021af0e1de7886f09b4f
Dmitry Kovalev [Thu, 9 Jan 2014 00:29:29 +0000 (16:29 -0800)]
Merge "Renaming 'Mode' to 'mode'."
Alex Converse [Wed, 8 Jan 2014 22:43:46 +0000 (14:43 -0800)]
Merge "Add a C fallback for get_msb() and change inline to INLINE."
hkuang [Wed, 8 Jan 2014 22:41:43 +0000 (14:41 -0800)]
Merge "Add initial intra frame neon optimization. 1~2% gain."
Dmitry Kovalev [Wed, 8 Jan 2014 22:33:59 +0000 (14:33 -0800)]
Renaming 'Mode' to 'mode'.
Change-Id: I6cdd670d66288dbd66228f38bba6b30502d25362
Dmitry Kovalev [Wed, 8 Jan 2014 22:14:05 +0000 (14:14 -0800)]
Merge "Using struct twopass_rc* instead of VP9_COMP*."
Alex Converse [Wed, 8 Jan 2014 20:25:47 +0000 (12:25 -0800)]
Add a C fallback for get_msb() and change inline to INLINE.
For systems without __builtin_clz() or _BitScanReverse(), taken from libwep
Change-Id: Iead257efc1772c466c79e1dc0356ed571d38d43e
hkuang [Mon, 6 Jan 2014 22:34:09 +0000 (14:34 -0800)]
Add initial intra frame neon optimization. 1~2% gain.
More intra optimizations will be added.
Change-Id: I33ae8d93f6002bf7b64cc2669602d9e6bfa5a6e8
Yunqing Wang [Wed, 8 Jan 2014 19:33:42 +0000 (11:33 -0800)]
Merge "AVX2 Variance Optimization"
Johann [Wed, 8 Jan 2014 19:10:03 +0000 (11:10 -0800)]
Merge "Include gen_msvs_vcxproj.sh"
Alex Converse [Wed, 8 Jan 2014 19:06:54 +0000 (11:06 -0800)]
Merge "Replace RD modeling with a fixed point approximation."
levytamar82 [Sun, 29 Dec 2013 08:23:50 +0000 (01:23 -0700)]
AVX2 Variance Optimization
Optimizing the variance functions: vp9_variance16x16, vp9_variance32x32,
vp9_variance64x64, vp9_variance32x16, vp9_variance64x32,
vp9_mse16x16 by migrating to AVX2
some of the functions were optimized by processing 32 elements instead of 16.
some of the functions were optimized by processing 2 loop strides of 16
elements in a single 256 bit register
This optimization gives between 2.4% - 2.7% user level performance gain
and 42% function level gain.
Change-Id: I265ae08a2b0196057a224a86450153ef3aebd85d
Alex Converse [Fri, 3 Jan 2014 21:24:11 +0000 (13:24 -0800)]
Replace RD modeling with a fixed point approximation.
Change-Id: I44eb44eb3f36c05d916ef140ef42cc84f72f99ec
Jingning Han [Wed, 8 Jan 2014 18:06:03 +0000 (10:06 -0800)]
Merge "Fix an issue in motion vector prediction stage"
Johann [Wed, 8 Jan 2014 17:51:15 +0000 (09:51 -0800)]
Include gen_msvs_vcxproj.sh
Change-Id: I28e9cf9347acd7279df3b841863a248479633265
Deb Mukherjee [Wed, 8 Jan 2014 17:20:29 +0000 (09:20 -0800)]
Merge "Further rate control cleanups"
Johann [Wed, 8 Jan 2014 16:42:25 +0000 (08:42 -0800)]
Merge "Remove yasm.rules dependency"
Deb Mukherjee [Mon, 6 Jan 2014 23:19:22 +0000 (15:19 -0800)]
Further rate control cleanups
Some cleanups on frames_to_key, frames_since_key.
Also removes the unused fixed_q parameters in vp9.
Change-Id: If8743a32c71de30a8d17136477b53d607a7acda8
Jingning Han [Tue, 7 Jan 2014 17:53:38 +0000 (09:53 -0800)]
Fix an issue in motion vector prediction stage
The previous implementation stops motion vector prediction test when
the zero motion vector appears for the second time. This commit fixes
it by simply skipping the second time check on zero mv and continuing
on to next mv candidate.
It slightly improves stdhd in speed 2 by 0.06% on average. Most static
sequences are not affected. A few hard ones, like jet, ped, and riverbed
were improved by 0.1 - 0.2%.
Change-Id: Ia8d4e2ffb7136669e8ad1fb24ea6e8fdd6b9a3c1
Jingning Han [Tue, 7 Jan 2014 18:15:51 +0000 (10:15 -0800)]
Remove deprecated variable from rt_speed_feature
This resolves a merge error.
Change-Id: Ifb83acc0a08e80c82f7624f9c86f79d3a86cc871
Dmitry Kovalev [Tue, 7 Jan 2014 17:56:41 +0000 (09:56 -0800)]
Merge "Adding new_mv local variable."
Dmitry Kovalev [Tue, 7 Jan 2014 17:56:06 +0000 (09:56 -0800)]
Merge "Adding get_ref_frame_buffer() function."
Dmitry Kovalev [Tue, 7 Jan 2014 17:54:05 +0000 (09:54 -0800)]
Merge "Removing unused mvp_fill manipulation code."
Jingning Han [Tue, 7 Jan 2014 17:31:17 +0000 (09:31 -0800)]
Merge "Remove avoid_frame_with_high_error from RD loop"
Dmitry Kovalev [Tue, 7 Jan 2014 10:40:02 +0000 (02:40 -0800)]
Merge "Reusing ROUND_POWER_OF_TWO macro."
Paul Wilkins [Tue, 7 Jan 2014 10:27:20 +0000 (02:27 -0800)]
Merge "Clean up: unused function and variables"
Dmitry Kovalev [Tue, 7 Jan 2014 02:07:12 +0000 (18:07 -0800)]
Removing unused mvp_fill manipulation code.
The code can be removed because mvp_full will be overridden after that.
Change-Id: I89559b1b6914c86bcd02b7359d37241948ac11d3
Dmitry Kovalev [Tue, 7 Jan 2014 01:58:26 +0000 (17:58 -0800)]
Merge "Replacing &cpi->common with cm."
Dmitry Kovalev [Tue, 7 Jan 2014 01:58:01 +0000 (17:58 -0800)]
Adding new_mv local variable.
Change-Id: I9631b35810c232c134f39dc0edadb1b3860a45ae
Dmitry Kovalev [Tue, 7 Jan 2014 01:38:57 +0000 (17:38 -0800)]
Reusing ROUND_POWER_OF_TWO macro.
Change-Id: I064ba32d5358bfbf080a4300fc1793b345080006
Dmitry Kovalev [Tue, 7 Jan 2014 01:29:16 +0000 (17:29 -0800)]
Replacing &cpi->common with cm.
Change-Id: Ic5bf5682ccdb8d2fbad6bba0d7db19a4f47b62a1
Alex Converse [Tue, 7 Jan 2014 01:22:21 +0000 (17:22 -0800)]
Merge "Fix encoding Raw yv12 and i420 from a pipe."
Marco Paniconi [Tue, 7 Jan 2014 00:35:52 +0000 (16:35 -0800)]
Merge "Code cleanup: remove unneeded lines."
Alex Converse [Tue, 7 Jan 2014 00:29:09 +0000 (16:29 -0800)]
Fix encoding Raw yv12 and i420 from a pipe.
rewind() does not work on pipes.
https://code.google.com/p/webm/issues/detail?id=678
Change-Id: I057f1e25c3f5662012d6e33ff4c97c88f50df357
Yaowu Xu [Tue, 7 Jan 2014 00:26:57 +0000 (16:26 -0800)]
Merge "Added placeholder for real time mode"
Marco Paniconi [Tue, 7 Jan 2014 00:03:13 +0000 (16:03 -0800)]
Code cleanup: remove unneeded lines.
Change-Id: I44a89b822a436299b9dd4ff26ad2e35767c29c58
Charles 'Buck' Krasic [Mon, 6 Jan 2014 23:18:04 +0000 (15:18 -0800)]
Merge "Write correct resolution to the IVF file header (b/
11270652)"
Johann [Mon, 6 Jan 2014 23:12:42 +0000 (15:12 -0800)]
Remove yasm.rules dependency
The file was removed by
9152f4851dd6cf5800b5db82a03ce39a08095ecf after
the solution files were changed.
Change-Id: I868c56fd609f45fb3e21afd085b9e6c268aac038
Dmitry Kovalev [Mon, 6 Jan 2014 23:02:31 +0000 (15:02 -0800)]
Merge "Combining ref_frame and second_ref_frame into ref_frames[2]."
Dmitry Kovalev [Mon, 6 Jan 2014 23:01:54 +0000 (15:01 -0800)]
Merge "Moving reset_segment_features() to encoder/vp9_segmentation.h."
Dmitry Kovalev [Mon, 6 Jan 2014 22:50:54 +0000 (14:50 -0800)]
Adding get_ref_frame_buffer() function.
Encapsulating direct references to lst_fb_idx, gld_fb_idx, alt_fb_idx.
Change-Id: I7e65ba3f131286e433e6651970c5647311fa4687
Dmitry Kovalev [Mon, 6 Jan 2014 22:19:44 +0000 (14:19 -0800)]
Merge "Adding RefBuffer struct."
Charles 'Buck' Krasic [Sat, 4 Jan 2014 00:06:35 +0000 (16:06 -0800)]
Write correct resolution to the IVF file header (b/
11270652)
also:
o remove dead code, create_dummy_frame
o Fix a bug in command line handling that caused a segfault if wrong
number of arguments were given.
Change-Id: I78f026aee4e363967b750e6cde0982659c558a1f
Jingning Han [Mon, 6 Jan 2014 21:29:16 +0000 (13:29 -0800)]
Remove avoid_frame_with_high_error from RD loop
The feature undergoes prior assumption that the recursive partition
size search from 4x4 to 64x64, hence utilizing information from small
blocks to determine early termination in large block rate-distortion
optimization search. The current codebase is now going from top down.
The previous function might go with not properly initialized values,
hence removed.
Tested on pedestrian_area_1080p at 4000 kbps running under speed 2.
No visible difference in runtime observed.
Change-Id: I553df415c6191413762db7ae34e8790c71d8118e
Dmitry Kovalev [Mon, 6 Jan 2014 20:46:23 +0000 (12:46 -0800)]
Using struct twopass_rc* instead of VP9_COMP*.
Change-Id: Id9ff7772aa3a3fb5d6cf94aff7dc9489bd964340
Dmitry Kovalev [Mon, 6 Jan 2014 20:23:36 +0000 (12:23 -0800)]
Combining ref_frame and second_ref_frame into ref_frames[2].
Change-Id: I007d66a1cb1b44751dcceafbaa64649ed9a34562
Deb Mukherjee [Mon, 6 Jan 2014 20:01:30 +0000 (12:01 -0800)]
Merge "Corerctly sets frame type in the 2 pass case"
Yaowu Xu [Mon, 6 Jan 2014 19:51:18 +0000 (11:51 -0800)]
Added placeholder for real time mode
Change-Id: I203d10f76c7ca78d875eaae15557cd765c6240d1
Dmitry Kovalev [Mon, 6 Jan 2014 19:31:57 +0000 (11:31 -0800)]
Moving reset_segment_features() to encoder/vp9_segmentation.h.
Change-Id: I0db4b31cb2382d4f6249eae0a8f42d227ad0ac57
Dmitry Kovalev [Mon, 6 Jan 2014 18:50:16 +0000 (10:50 -0800)]
Merge "Moving get_scan() call out of decode_coeffs() function."
Deb Mukherjee [Fri, 3 Jan 2014 23:41:57 +0000 (15:41 -0800)]
Corerctly sets frame type in the 2 pass case
This patch sets frame types correctly in the new
vp9_get_second_pass_params() function called prior
to encode_frame_to_data_rate() function, so that the
latter function can just work with what is passed to
it. This will allow multiple vp9_get_second_pass_params()
to be created for various encode strategies without
messing with the core encode function.
There is no difference in derf and yt. stdhd/hd are pending.
Change-Id: I70dfb97e9f497e9cee04052e0e8e0c2892eab0c3
Jingning Han [Mon, 6 Jan 2014 17:38:19 +0000 (09:38 -0800)]
Merge "Tune IDCT8_1D macro function interface"
Paul Wilkins [Mon, 6 Jan 2014 11:26:02 +0000 (11:26 +0000)]
Clean up: unused function and variables
Remove adjust_maxq_qrange() and related variables.
Change-Id: I50d065f4619c499283e2680e31d0d99c22439dec
Paul Wilkins [Mon, 6 Jan 2014 10:04:55 +0000 (02:04 -0800)]
Merge "Modified Handling of min and max vbr rates."
Dmitry Kovalev [Sat, 4 Jan 2014 02:18:39 +0000 (18:18 -0800)]
Moving get_scan() call out of decode_coeffs() function.
Change-Id: I5db2e5a6c36e6c503dea2e07d9d2e0daf4ac9d89
Dmitry Kovalev [Sat, 4 Jan 2014 02:02:10 +0000 (18:02 -0800)]
Merge "Converting CHECK_POINT macro to inline function."
Dmitry Kovalev [Sat, 4 Jan 2014 01:38:28 +0000 (17:38 -0800)]
Merge "Replacing int_mv with MV."
Dmitry Kovalev [Sat, 4 Jan 2014 00:16:28 +0000 (16:16 -0800)]
Converting CHECK_POINT macro to inline function.
Change-Id: I08533d59a78346dac30a7dcbc12146f23ef88bbc
Dmitry Kovalev [Fri, 3 Jan 2014 23:57:36 +0000 (15:57 -0800)]
Merge "Replacing CHECK_BOUNDS macro with inline check_bounds function."
Dmitry Kovalev [Fri, 3 Jan 2014 23:56:16 +0000 (15:56 -0800)]
Merge "Moving hev mask calculation into filter4() function."
Jingning Han [Fri, 3 Jan 2014 23:05:25 +0000 (15:05 -0800)]
Tune IDCT8_1D macro function interface
This commit adds input/output ports for IDCT8_1D macro function to
provide more flexibility in variable use. It allows to skip several
buffer swap operations.
Change-Id: I21f3450509537322293043b3281bfd3949868677
Dmitry Kovalev [Sat, 28 Dec 2013 02:44:19 +0000 (18:44 -0800)]
Adding RefBuffer struct.
Adding RefBuffer to simplify reference buffer management. The struct has a
pointer to image data and scale factors relative to the current frame.
Change-Id: If38eb1491ff687cc11428aee339f3e052e2c5d9e
Dmitry Kovalev [Fri, 3 Jan 2014 23:13:22 +0000 (15:13 -0800)]
Merge "Cleaning up get_prediction_decay_rate() function."
Dmitry Kovalev [Fri, 3 Jan 2014 23:04:53 +0000 (15:04 -0800)]
Merge "Pre planes configuration cleanup."
Jingning Han [Fri, 3 Jan 2014 20:54:57 +0000 (12:54 -0800)]
Merge "Reduce num of buffer swap calls in idct8_1d_sse2"
Jingning Han [Fri, 3 Jan 2014 20:54:43 +0000 (12:54 -0800)]
Merge "Rework idct8x8_10 SSE2 implementation"
Dmitry Kovalev [Fri, 3 Jan 2014 20:50:57 +0000 (12:50 -0800)]
Pre planes configuration cleanup.
Change-Id: I1d50f8701d9c9dedb84387a773a3e9b4daaad720
Dmitry Kovalev [Fri, 3 Jan 2014 20:12:53 +0000 (12:12 -0800)]
Merge "Merging best_ref_mv and second_best_ref_mv into best_ref_mv[2]."
Jingning Han [Fri, 3 Jan 2014 20:09:34 +0000 (12:09 -0800)]
Reduce num of buffer swap calls in idct8_1d_sse2
This commit merges the initial buffer swap operations in idct8_1d_sse2
into the array transpose step, hence reducing number of instructions
therein.
Change-Id: I219f6f50813390d2ec3ee37eecf2a4a2b44ae479
Dmitry Kovalev [Fri, 3 Jan 2014 20:11:36 +0000 (12:11 -0800)]
Cleaning up get_prediction_decay_rate() function.
Change-Id: Ie8fcee21f41f91f94b4fa02f2a55691dea1734e3
Jingning Han [Thu, 2 Jan 2014 23:33:38 +0000 (15:33 -0800)]
Rework idct8x8_10 SSE2 implementation
This commit optimizes the SSE2 implmentation of idct8x8_10. It exploits
the fact that only top-left 4x4 block contains non-zero coefficients,
and hence reduces the instructions needed.
The runtime of idct8x8_10_sse2 goes down from 216 to 198 CPU cycles,
estimated by averaging over 100000 runs. For pedestrian_area_1080p 300
frames coded at 4000kbps, the average decoding speed goes up from
79.3 fps to 79.7 fps.
Change-Id: I6d277bbaa3ec9e1562667906975bae06904cb180
Dmitry Kovalev [Fri, 3 Jan 2014 19:48:07 +0000 (11:48 -0800)]
Replacing int_mv with MV.
Change-Id: Ifd432fa3741ba47102d298e0b348eb00f5a9ce53
Dmitry Kovalev [Fri, 3 Jan 2014 19:47:57 +0000 (11:47 -0800)]
Merge changes Ic0a2427a,I3addbf6d
* changes:
Removing CONFIG_MD5.
Using VP9_FRAME_MARKER instead of raw number.
Dmitry Kovalev [Fri, 3 Jan 2014 19:31:00 +0000 (11:31 -0800)]
Merging best_ref_mv and second_best_ref_mv into best_ref_mv[2].
Change-Id: If04b57828847cee09a79c94e1098d1aa4990ea0d
Paul Wilkins [Thu, 2 Jan 2014 15:45:06 +0000 (15:45 +0000)]
Modified Handling of min and max vbr rates.
In two pass encodes bits are allocated to each frame
according to a modified error score for the frame as a
fraction of the modified error score for the clip or section.
Previously a minimum rate per frame was reserved and
subtracted from the bits allocatable by the two pass code.
The vbr max section rate was enforced by clipping the
actual number of bits allocated.
In this patch the min and max vbr rates are enforced
instead by clipping the modified error scores for each frame
rather than the number of bits allocated.
Small gains for all test sets (psnr and SSIM) ranging from
~ +0.05 for YT psnr up to ~ +0.25 for Std-hd SSIM.
Change-Id: Iae27d70bdd3944e3f0cceaf225bad2e8802833de
Dmitry Kovalev [Fri, 3 Jan 2014 02:29:56 +0000 (18:29 -0800)]
Reusing vp9_get_skip_context() function in encoder.
Change-Id: Ic0345622115941f49b6a568c7b8154ba892cbf0d
Christian Duvivier [Thu, 2 Jan 2014 23:03:58 +0000 (15:03 -0800)]
Merge "ARM NEON version of denoiser."
Christian Duvivier [Thu, 2 Jan 2014 02:46:47 +0000 (18:46 -0800)]
ARM NEON version of denoiser.
Change-Id: I951abd4ad0078f78949f3cb79453ac334fb82a7e
Yaowu Xu [Thu, 2 Jan 2014 17:27:28 +0000 (09:27 -0800)]
Merge "Fix show existing frame"
Dmitry Kovalev [Sat, 28 Dec 2013 00:10:18 +0000 (16:10 -0800)]
Removing CONFIG_MD5.
We don't need compile time md5 configuration because --md5 is a runtime
option.
Change-Id: Ic0a2427ae5de5a18f31e5ee60c3732481b377ca1
Jingning Han [Fri, 27 Dec 2013 23:04:16 +0000 (15:04 -0800)]
Merge "Adaptive motion control on ref and search range"
Dmitry Kovalev [Fri, 27 Dec 2013 22:50:27 +0000 (14:50 -0800)]
Using VP9_FRAME_MARKER instead of raw number.
Change-Id: I3addbf6d89a86a707c8df1a463da3e9e367910df
Dmitry Kovalev [Fri, 27 Dec 2013 22:47:57 +0000 (14:47 -0800)]
Merge "Removing vpx_codec_vp9x_cx and internal experimental flag."
Yunqing Wang [Fri, 27 Dec 2013 22:05:38 +0000 (14:05 -0800)]
Merge "Remove a unused sub-pixel search"
Dmitry Kovalev [Fri, 27 Dec 2013 22:01:12 +0000 (14:01 -0800)]
Removing vpx_codec_vp9x_cx and internal experimental flag.
vpx_codec_vp9x_cx is not used internally. Experimental flag from
vp9_extracfg is also not really used. YUV 4:4:4 just works after these
changes (you have to specify --profile=1 for the encoder).
Change-Id: Ib1c8461d0d19d159827e005efe868f891eea0140
Jingning Han [Fri, 20 Dec 2013 23:24:22 +0000 (15:24 -0800)]
Adaptive motion control on ref and search range
This commit takes a preliminary attempt to refine the motion search
control. It detects the SAD associated with mv predictor per reference
frame, and based on which to determine whether the encoder wants to
reduce the motion search range (if the predicted mv provides fairly
small SAD), or to skip the current reference frame (if there exists
another ref frame that gives much smaller SAD cost).
This feature is turned on in the settings of speed 1 and above.
In speed 1, compression performance changed
derf -0.018%
yt -0.043%
hd -0.045%
stdhd -0.281%
speed-up
pedestrian_area_1080p at 4000 kbps 100 frames
199651ms -> 188846ms (5.5% speed-up)
blue_sky_1080p at 6000 kbps
443531ms -> 415239ms (6.3% speed-up)
In speed 2, compression performance changed
derf -0.026%
yt -0.090%
hd -0.055%
stdhd -0.210%
speed-up
pedstrian 113949ms -> 108855ms (4.5% speed-up)
blue_sky 271057ms -> 257322ms (5% speed-up)
Change-Id: I1b74ea28278c94fea329d971d706d573983d810d
Dmitry Kovalev [Thu, 26 Dec 2013 21:41:02 +0000 (13:41 -0800)]
Merge "Calculating has_second_ref only once for single_ref context."
Dmitry Kovalev [Thu, 26 Dec 2013 21:38:38 +0000 (13:38 -0800)]
Merge "Removing vp9_findnearmv.{h, c} files."
Dmitry Kovalev [Fri, 20 Dec 2013 23:56:59 +0000 (15:56 -0800)]
Moving MAX_PROB constant to vp9_prob.h.
Change-Id: I07470ad1b7a0344d088911428ffab8ba9a0d8708
Dmitry Kovalev [Fri, 20 Dec 2013 19:22:25 +0000 (11:22 -0800)]
Renaming vp9_dboolhuff.{h, c} to vp9_reader.{h, c}.
Change-Id: I50c009ff8108bda1c57427f23d63a79c04f7e776
Dmitry Kovalev [Fri, 20 Dec 2013 20:51:37 +0000 (12:51 -0800)]
Merge "Renaming vp9_boolcoder.{h, c} to vp9_writer.{h, c}."
Dmitry Kovalev [Fri, 20 Dec 2013 19:22:02 +0000 (11:22 -0800)]
Merge "Using single struct to represent scale factors."
Dmitry Kovalev [Fri, 20 Dec 2013 19:10:24 +0000 (11:10 -0800)]
Renaming vp9_boolcoder.{h, c} to vp9_writer.{h, c}.
Change-Id: I9b9a5fcce8530284df0f270706ee060a0edc1517
Dmitry Kovalev [Fri, 20 Dec 2013 18:56:31 +0000 (10:56 -0800)]
Merge "Reusing FRAME_COUNTS in the encoder."
Jingning Han [Fri, 20 Dec 2013 18:37:20 +0000 (10:37 -0800)]
Merge "Store the SSE of prediction residuals"
Marco Paniconi [Fri, 20 Dec 2013 18:28:09 +0000 (10:28 -0800)]
Merge "Initialize avg_frame_qindex to worst_allowed for 1 pass."
Yaowu Xu [Fri, 20 Dec 2013 17:20:50 +0000 (09:20 -0800)]
Merge "Fix a bug"
Yaowu Xu [Fri, 20 Dec 2013 16:52:35 +0000 (08:52 -0800)]
Fix a bug
The line was accidently removed in
4dbad63a7.
Change-Id: Ic1e18f209cead95cecc684f952ae667271b58a97
Yunqing Wang [Fri, 20 Dec 2013 16:46:11 +0000 (08:46 -0800)]
Merge "Code clean up"
Paul Wilkins [Fri, 20 Dec 2013 12:21:56 +0000 (04:21 -0800)]
Merge "Adjust gf_group_error_left for arf groups."