1 2021-03-09 v1.10.0 "Ruddy Duck"
2 This maintenance release adds support for darwin20 and new codec controls, as
3 well as numerous bug fixes.
6 New codec control is added to disable loopfilter for VP9.
8 New encoder control is added to disable feature to increase Q on overshoot
11 Configure support for darwin20 is added.
13 New codec control is added for VP9 rate control. The control ID of this
14 interface is VP9E_SET_EXTERNAL_RATE_CONTROL. To make VP9 use a customized
15 external rate control model, users will have to implement each callback
16 function in vpx_rc_funcs_t and register them using libvpx API
17 vpx_codec_control_() with the control ID.
20 Use -std=gnu++11 instead of -std=c++11 for c++ files.
23 Override assembler with --as option of configure for MSVS.
24 Fix several compilation issues with gcc 4.8.5.
25 Fix to resetting rate control for temporal layers.
26 Fix to the rate control stats of SVC example encoder when number of spatial
28 Fix to reusing motion vectors from the base spatial layer in SVC.
29 2 pass related flags removed from SVC example encoder.
31 2020-07-29 v1.9.0 "Quacking Duck"
32 This release adds support for NV12, a separate library for rate control, as
33 well as incremental improvements.
36 NV12 support is added to this release.
37 A new interface is added for VP9 rate control. The new library libvp9rc.a
38 must be linked by applications.
39 Googletest is updated to v1.10.0.
40 simple_encode.cc is compiled into a new library libsimple_encode.a with
44 Various changes to improve VP9 SVC, rate control, quality and speed to real
48 Fix key frame update refresh simulcast flexible svc.
49 Fix to disable_16x16part speed feature for real time encoding.
50 Fix some signed integer overflows for VP9 rate control.
51 Fix initialization of delta_q_uv.
52 Fix condition in regulate_q for cyclic refresh.
53 Various fixes to dynamic resizing for VP9 SVC.
55 2019-12-09 v1.8.2 "Pekin Duck"
56 This release collects incremental improvements to many aspects of the library.
59 ARCH_* defines have been removed in favor of VPX_ARCH_*.
61 2019-07-15 v1.8.1 "Orpington Duck"
62 This release collects incremental improvements to many aspects of the library.
65 VP8E_SET_CPUUSED now accepts values up to 9 for vp9.
66 VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT had a spelling fix (was VP8E).
67 The --sdk-path option has been removed. If you were using it to build for
68 Android please read build/make/Android.mk for alternatives.
69 All PPC optimizations have been disabled:
70 https://bugs.chromium.org/p/webm/issues/detail?id=1522.
73 Various changes to improve encoder rate control, quality and speed
74 for practically every use case.
77 vp9-rtc: Fix color artifacts for speed >= 8.
79 2019-01-31 v1.8.0 "Northern Shoveler Duck"
80 This release focused on encoding performance for realtime and VOD use cases.
83 This adds and improves several vp9 controls. Most are related to SVC:
84 VP9E_SET_SVC_FRAME_DROP_LAYER:
85 - Frame dropping in SVC.
86 VP9E_SET_SVC_INTER_LAYER_PRED:
87 - Inter-layer prediction in SVC.
88 VP9E_SET_SVC_GF_TEMPORAL_REF:
89 - Enable long term temporal reference in SVC.
90 VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG:
91 - Extend and improve this control for better flexibility in setting SVC
93 VP9E_SET_POSTENCODE_DROP:
94 - Allow for post-encode frame dropping (applies to non-SVC too).
95 VP9E_SET_SVC_SPATIAL_LAYER_SYNC:
96 - Enable spatial layer sync frames.
97 VP9E_SET_SVC_LAYER_ID:
98 - Extend api to specify temporal id for each spatial layers.
100 - Extend Region of Interest functionality to VP9.
103 2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6,
104 we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1,
105 the gains are approximately 4% for VBR and 5% for CQ.
107 For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at
108 screen sharing have improved when the content changes significantly (slide
109 sharing) or scrolls. There is a new speed 9 setting for mobile devices which
110 is about 10-20% faster than speed 8.
114 VP9 partition issue for 1080p.
115 VP9 rate control improvments.
116 Postprocessing Multi Frame Quality Enhancement (MFQE) issue.
117 VP8 multithread decoder issues.
118 A variety of fuzzing issues.
120 2018-01-04 v1.7.0 "Mandarin Duck"
121 This release focused on high bit depth performance (10/12 bit) and vp9
122 encoding improvements.
125 This release is ABI incompatible due to new vp9 encoder features.
127 Frame parallel decoding for vp9 has been removed.
130 vp9 encoding supports additional threads with --row-mt. This can be greater
131 than the number of tiles.
133 Two new vp9 encoder options have been added:
137 Additional tooling for respecting the vp9 "level" profiles has been added.
140 A variety of fuzzing issues.
141 vp8 threading fix for ARM.
142 Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
143 Reject invalid multi resolution configurations.
145 2017-01-09 v1.6.1 "Long Tailed Duck"
146 This release improves upon the VP9 encoder and speeds up the encoding and
150 This release is ABI compatible with 1.6.0.
153 Faster VP9 encoding and decoding.
154 High bit depth builds now provide similar speed for 8 bit encode and decode
155 for x86 targets. Other platforms and higher bit depth improvements are in
159 A variety of fuzzing issues.
161 2016-07-20 v1.6.0 "Khaki Campbell Duck"
162 This release improves upon the VP9 encoder and speeds up the encoding and
166 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
167 in vpx_image and some minor changes to the VP8_COMP structure.
169 The default key frame interval for VP9 has changed from 128 to 9999.
172 A core focus has been performance for low end Intel processors. SSSE3
173 instructions such as 'pshufb' have been avoided and instructions have been
174 reordered to better accommodate the more constrained pipelines.
176 As a result, devices based on Celeron processors have seen substantial
177 decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
178 decoding speed improved between 10 and 30%. Between Javan Whistling Duck
179 and Khaki Campbell Duck, it improved another 10 to 15%.
181 While Celeron benefited most, Core-i5 also improved 5% and 10% between the
184 Realtime performance for WebRTC for both speed and quality has received a
188 A number of fuzzing issues, found variously by Mozilla, Chromium and others,
189 have been fixed and we strongly recommend updating.
191 2015-11-09 v1.5.0 "Javan Whistling Duck"
192 This release improves upon the VP9 encoder and speeds up the encoding and
196 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
197 controls and adds a variety of VP9 controls for testing.
199 The vpxenc utility now prefers VP9 by default.
202 Faster VP9 encoding and decoding
203 Smaller library size by combining functions used by VP8 and VP9
206 A variety of fuzzing issues
208 2015-04-03 v1.4.0 "Indian Runner Duck"
209 This release includes significant improvements to the VP9 codec.
212 This release is ABI incompatible with 1.3.0. It drops the compatibility
213 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
217 Faster VP9 encoding and decoding
218 Multithreaded VP9 decoding (tile and frame-based)
219 Multithreaded VP9 encoding - on by default
220 YUV 4:2:2 and 4:4:4 support in VP9
221 10 and 12bit support in VP9
222 64bit ARM support by replacing ARM assembly with intrinsics
225 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
229 Frame Parallel decoding fails for segmented and non-420 files.
231 2013-11-15 v1.3.0 "Forest"
232 This release introduces the VP9 codec in a backward-compatible way.
233 All existing users of VP8 can continue to use the library without
234 modification. However, some VP8 options do not map to VP9 in the same manner.
236 The VP9 encoder in this release is not feature complete. Users interested in
237 the encoder are advised to use the git master branch and discuss issues on
238 libvpx mailing lists.
241 This release is ABI and API compatible with Duclair (v1.0.0). Users
242 of older releases should refer to the Upgrading notes in this document
246 Get rid of bashisms in the main build scripts
247 Added usage info on command line options
248 Add lossless compression mode
250 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
251 Add option to disable documentation
252 configure: add --enable-external-build support
253 make: support V=1 as short form of verbose=yes
254 configure: support mingw-w64
255 configure: support hardfloat armv7 CHOSTS
256 configure: add support for android x86
257 Add estimated completion time to vpxenc
258 Don't exit on decode errors in vpxenc
259 vpxenc: support scaling prior to encoding
260 vpxdec: support scaling output
261 vpxenc: improve progress indicators with --skip
262 msvs: Don't link to winmm.lib
263 Add a new script for producing vcxproj files
264 Produce Visual Studio 10 and 11 project files
265 Produce Windows Phone project files
266 msvs-build: use msbuild for vs >= 2005
267 configure: default configure log to config.log
268 Add encoding option --static-thresh
271 Miscellaneous speed optimizations for VP8 and VP9.
274 In general, quality is consistent with the Eider release.
277 This release represents approximately a year of engineering effort,
278 and contains multiple bug fixes. Please refer to git history for details.
282 This release acts as a checkpoint for a large amount of internal refactoring
283 and testing. It also contains a number of small bugfixes, so all users are
284 encouraged to upgrade.
287 This release is ABI and API compatible with Duclair (v1.0.0). Users
288 of older releases should refer to the Upgrading notes in this
289 document for that release.
292 VP8 optimizations for MIPS dspr2
293 vpxenc: add -quiet option
296 Encoder and decoder speed is consistent with the Eider release.
299 In general, quality is consistent with the Eider release.
301 Minor tweaks to ARNR filtering
302 Minor improvements to real time encoding with multiple temporal layers
305 Fixes multithreaded encoder race condition in loopfilter
306 Fixes multi-resolution threaded encoding
307 Fix potential encoder dead-lock after picture resize
310 2012-05-09 v1.1.0 "Eider"
311 This introduces a number of enhancements, mostly focused on real-time
312 encoding. In addition, it fixes a decoder bug (first introduced in
313 Duclair) so all users of that release are encouraged to upgrade.
316 This release is ABI and API compatible with Duclair (v1.0.0). Users
317 of older releases should refer to the Upgrading notes in this
318 document for that release.
320 This release introduces a new temporal denoiser, controlled by the
321 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
322 currently take a strength parameter, so the control is effectively
323 a boolean - zero (off) or non-zero (on). For compatibility with
324 existing applications, the values accepted are the same as those
325 for the spatial denoiser (0-6). The temporal denoiser is enabled
326 by default, and the older spatial denoiser may be restored by
327 configuring with --disable-temporal-denoising. The temporal denoiser
328 is more computationally intensive than the spatial one.
330 This release removes support for a legacy, decode only API that was
331 supported, but deprecated, at the initial release of libvpx
332 (v0.9.0). This is not expected to have any impact. If you are
333 impacted, you can apply a reversion to commit 2bf8fb58 locally.
334 Please update to the latest libvpx API if you are affected.
337 Adds a motion compensated temporal denoiser to the encoder, which
338 gives higher quality than the older spatial denoiser. (See above
339 for notes on upgrading).
341 In addition, support for new compilers and platforms were added,
343 improved support for XCode
344 Android x86 NDK build
348 Changing resolution with vpx_codec_enc_config_set() is now
349 supported. Previously, reinitializing the codec was required to
350 change the input resolution.
352 The vpxenc application has initial support for producing multiple
353 encodes from the same input in one call. Resizing is not yet
354 supported, but varying other codec parameters is. Use -- to
355 delineate output streams. Options persist from one stream to the
358 Also, the vpxenc application will now use a keyframe interval of
359 5 seconds by default. Use the --kf-max-dist option to override.
362 Decoder performance improved 2.5% versus Duclair. Encoder speed is
363 consistent with Duclair for most material. Two pass encoding of
364 slideshow-like material will see significant improvements.
366 Large realtime encoding speed gains at a small quality expense are
367 possible by configuring the on-the-fly bitpacking experiment with
368 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
369 faster (ARM) depending on the number of threads and bitrate
370 settings. This technique sees constant gain over the 5-16 speed
371 range. For VC style input the loss seen is up to 0.2dB. See commit
372 52cf4dca for further details.
375 On the whole, quality is consistent with the Duclair release. Some
378 Reduced blockiness in easy sections by applying a penalty to
381 Improved quality of static sections (like slideshows) with
384 Improved keyframe sizing with multiple temporal layers
387 Corrected alt-ref contribution to frame rate for visible updates
388 to the alt-ref buffer. This affected applications making manual
389 usage of the frame reference flags, or temporal layers.
391 Additional constraints were added to disable multi-frame quality
392 enhancement (MFQE) in sections of the frame where there is motion.
395 Fixed corruption issues when vpx_codec_enc_config_set() was called
396 with spatial resampling enabled.
398 Fixed a decoder error introduced in Duclair where the segmentation
399 map was not being reinitialized on keyframes (#378)
402 2012-01-27 v1.0.0 "Duclair"
403 Our fourth named release, focused on performance and features related to
404 real-time encoding. It also fixes a decoder crash bug introduced in
405 v0.9.7, so all users of that release are encouraged to upgrade.
408 This release is ABI incompatible with prior releases of libvpx, so the
409 "major" version number has been bumped to 1. You must recompile your
410 applications against the latest version of the libvpx headers. The
411 API remains compatible, and this should not require code changes in most
415 This release introduces several substantial new features to the encoder,
416 of particular interest to real time streaming applications.
418 Temporal scalability allows the encoder to produce a stream that can
419 be decimated to different frame rates, with independent rate targeting
422 Multiframe quality enhancement postprocessing can make visual quality
423 more consistent in the presence of frames that are substantially
424 different quality than the surrounding frames, as in the temporal
425 scalability case and in some forced keyframe scenarios.
427 Multiple-resolution encoding support allows the encoding of the
428 same content at different resolutions faster than encoding them
432 Optimization targets for this release included the decoder and the real-
433 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
434 this release. Encoder improvements followed a curve where speeds 1-3
435 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
436 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
440 Encoder quality in the single stream case is consistent with the Cayuga
444 This release fixes an OOB read decoder crash bug present in v0.9.7
445 related to the clamping of motion vectors in SPLITMV blocks. This
446 behavior could be triggered by corrupt input or by starting
447 decoding from a P-frame.
450 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
451 This is an incremental bugfix release against Cayuga. All users of that
452 release are strongly encouraged to upgrade.
454 - Fix potential OOB reads (cdae03a)
456 An unbounded out of bounds read was discovered when the
457 decoder was requested to perform error concealment (new in
458 Cayuga) given a frame with corrupt partition sizes.
460 A bounded out of bounds read was discovered affecting all
461 versions of libvpx. Given an multipartition input frame that
462 is truncated between the mode/mv partition and the first
463 residiual paritition (in the block of partition offsets), up
464 to 3 extra bytes could have been read from the source buffer.
465 The code will not take any action regardless of the contents
466 of these undefined bytes, as the truncated buffer is detected
467 immediately following the read based on the calculated
468 starting position of the coefficient partition.
470 - Fix potential error concealment crash when the very first frame
471 is missing or corrupt (a609be5)
473 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
475 - Revert 1-pass CBR rate control changes (e961317)
476 Further testing showed this change produced undesirable visual
477 artifacts, rolling back for now.
480 2011-08-02 v0.9.7 "Cayuga"
481 Our third named release, focused on a faster, higher quality, encoder.
484 This release is backwards compatible with Aylesbury (v0.9.5) and
485 Bali (v0.9.6). Users of older releases should refer to the Upgrading
486 notes in this document for that release.
489 Stereo 3D format support for vpxenc
490 Runtime detection of available processor cores.
491 Allow specifying --end-usage by enum name
492 vpxdec: test for frame corruption
493 vpxenc: add quantizer histogram display
494 vpxenc: add rate histogram display
495 Set VPX_FRAME_IS_DROPPABLE
496 update configure for ios sdk 4.3
497 Avoid text relocations in ARM vp8 decoder
498 Generate a vpx.pc file for pkg-config.
499 New ways of passing encoded data between encoder and decoder.
502 This release includes across-the-board speed improvements to the
503 encoder. On x86, these measure at approximately 11.5% in Best mode,
504 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
505 On ARM Cortex A9 with Neon extensions, real-time encoding of video
506 telephony content is 35% faster than Bali on single core and 48%
507 faster on multi-core. On the NVidia Tegra2 platform, real time
508 encoding is 40% faster than Bali.
510 Decoder speed was not a priority for this release, but improved
511 approximately 8.4% on x86.
513 Reduce motion vector search on alt-ref frame.
514 Encoder loopfilter running in its own thread
515 Reworked loopfilter to precalculate more parameters
516 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
517 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
518 Removed redundant checks
519 Reduced structure sizes
520 utilize preload in ARMv6 MC/LPF/Copy routines
521 ARM optimized quantization, dfct, variance, subtract
522 Increase chrow row alignment to 16 bytes.
523 disable trellis optimization for first pass
524 Write SSSE3 sub-pixel filter function
525 Improve SSE2 half-pixel filter funtions
526 Add vp8_sub_pixel_variance16x8_ssse3 function
527 Reduce unnecessary distortion computation
528 Use diamond search to replace full search
529 Preload reference area in sub-pixel motion search (real-time mode)
532 This release focused primarily on one-pass use cases, including
533 video conferencing. Low latency data rate control was significantly
534 improved, improving streamability over bandwidth constrained links.
535 Added support for error concealment, allowing frames to maintain
536 visual quality in the presence of substantial packet loss.
538 Add rc_max_intra_bitrate_pct control
539 Limit size of initial keyframe in one-pass.
540 Improve framerate adaptation
541 Improved 1-pass CBR rate control
542 Improved KF insertion after fades to still.
543 Improved key frame detection.
544 Improved activity masking (lower PSNR impact for same SSIM boost)
545 Improved interaction between GF and ARFs
546 Adding error-concealment to the decoder.
547 Adding support for independent partitions
548 Adjusted rate-distortion constants
552 Removed firstpass motion map
553 Fix parallel make install
554 Fix multithreaded encoding for 1 MB wide frame
555 Fixed iwalsh_neon build problems with RVDS4.1
556 Fix semaphore emulation, spin-wait intrinsics on Windows
557 Fix build with xcode4 and simplify GLOBAL.
558 Mark ARM asm objects as allowing a non-executable stack.
559 Fix vpxenc encoding incorrect webm file header on big endian
562 2011-03-07 v0.9.6 "Bali"
563 Our second named release, focused on a faster, higher quality, encoder.
566 This release is backwards compatible with Aylesbury (v0.9.5). Users
567 of older releases should refer to the Upgrading notes in this
568 document for that release.
571 vpxenc --psnr shows a summary when encode completes
572 --tune=ssim option to enable activity masking
573 improved postproc visualizations for development
574 updated support for Apple iOS to SDK 4.2
575 query decoder to determine which reference frames were updated
576 implemented error tracking in the decoder
577 fix pipe support on windows
580 Primary focus was on good quality mode, speed 0. Average improvement
581 on x86 about 40%, up to 100% on user-generated content at that speed.
582 Best quality mode speed improved 35%, and realtime speed 10-20%. This
583 release also saw significant improvement in realtime encoding speed
586 Improved encoder threading
587 Dont pick encoder filter level when loopfilter is disabled.
588 Avoid double copying of key frames into alt and golden buffer
590 x86 sse2 temporal filter
591 SSSE3 version of fast quantizer
592 vp8_rd_pick_best_mbsegmentation code restructure
593 Adjusted breakout RD for SPLITMV
594 Changed segmentation check order
595 Improved rd_pick_intra4x4block
596 Adds armv6 optimized variance calculation
597 ARMv6 optimized sad16x16
598 ARMv6 optimized half pixel variance calculations
599 Full search SAD function optimization in SSE4.1
600 Improve MV prediction accuracy to achieve performance gain
601 Improve MV prediction in vp8_pick_inter_mode() for speed>3
604 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
605 also includes support for "activity masking," which greatly improves
606 SSIM at the expense of PSNR. For now, this feature is available with
607 the --tune=ssim option. Further experimentation in this area
608 is ongoing. This release also introduces a new rate control mode
609 called "CQ," which changes the allocation of bits within a clip to
610 the sections where they will have the most visual impact.
612 Tuning for the more exact quantizer.
613 Relax rate control for last few frames
615 Limit key frame quantizer for forced key frames.
617 Add simple version of activity masking.
618 make rdmult adaptive for intra in quantizer RDO
619 cap the best quantizer for 2nd order DC
620 change the threshold of DC check for encode breakout
623 Fix crash on Sparc Solaris.
624 Fix counter of fixed keyframe distance
625 ARNR filter pointer update bug fix
626 Fixed use of motion percentage in KF/GF group calc
627 Changed condition for using RD in Intra Mode
628 Fix encoder real-time only configuration.
629 Fix ARM encoder crash with multiple token partitions
630 Fixed bug first cluster timecode of webm file is wrong.
631 Fixed various encoder bugs with odd-sized images
632 vp8e_get_preview fixed when spatial resampling enabled
633 quantizer: fix assertion in fast quantizer path
634 Allocate source buffers to be multiples of 16
635 Fix for manual Golden frame frequency
636 Fix drastic undershoot in long form content
639 2010-10-28 v0.9.5 "Aylesbury"
640 Our first named release, focused on a faster decoder, and a better encoder.
643 This release incorporates backwards-incompatible changes to the
644 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
647 * the -q (quiet) option has been removed, and replaced with
648 -v (verbose). the output is quiet by default. Use -v to see
649 the version number of the binary.
651 * The default behavior is now to write output to a single file
652 instead of individual frames. The -y option has been removed.
653 Y4M output is the default.
655 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
656 options must be specified.
658 $ ivfdec -o OUTPUT INPUT
659 $ vpxdec --i420 -o OUTPUT INPUT
661 * If an output file is not specified, the default is to write
662 Y4M to stdout. This makes piping more natural.
664 $ ivfdec -y -o - INPUT | ...
667 * The output file has additional flexibility for formatting the
668 filename. It supports escape characters for constructing a
669 filename from the width, height, and sequence number. This
670 replaces the -p option. To get the equivalent:
672 $ ivfdec -p frame INPUT
673 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
676 * The output file must be specified with -o, rather than as the
679 $ ivfenc <options> INPUT OUTPUT
680 $ vpxenc <options> -o OUTPUT INPUT
682 * The output defaults to webm. To get IVF output, use the --ivf
685 $ ivfenc <options> INPUT OUTPUT.ivf
686 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
690 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
691 vpxdec supports .webm input
692 vpxdec writes .y4m by default
693 vpxenc writes .webm output by default
694 vpxenc --psnr now shows the average/overall PSNR at the end
695 ARM platforms now support runtime cpu detection
696 vpxdec visualizations added for motion vectors, block modes, references
697 vpxdec now silent by default
698 vpxdec --progress shows frame-by-frame timing information
699 vpxenc supports the distinction between --fps and --timebase
700 NASM is now a supported assembler
701 configure: enable PIC for shared libs by default
702 configure: add --enable-small
703 configure: support for ppc32-linux-gcc
704 configure: support for sparc-solaris-gcc
707 Improve handling of invalid frames
708 Fix valgrind errors in the NEON loop filters.
709 Fix loopfilter delta zero transitions
710 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
711 Build fixes for darwin-icc
714 20-40% (average 28%) improvement in libvpx decoder speed,
716 Rewrite vp8_short_walsh4x4_sse2()
717 Optimizations on the loopfilters.
718 Miscellaneous improvements for Atom
719 Add 4-tap version of 2nd-pass ARMv6 MC filter.
720 Improved multithread utilization
721 Better instruction choices on x86
722 reorder data to use wider instructions
723 Update NEON wide idcts
724 Make block access to frame buffer sequential
725 Improved subset block search
726 Bilinear subpixel optimizations for ssse3.
727 Decrease memory footprint
729 Encoder speed improvements (percentage gain not measured):
730 Skip unnecessary search of identical frames
731 Add SSE2 subtract functions
732 Improve bounds checking in vp8_diamond_search_sadx4()
733 Added vp8_fast_quantize_b_sse2
736 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
737 encoding mode, and up to 60% improvement on very noisy, still
738 or slow moving source video
740 Motion compensated temporal filter for Alt-Ref Noise Reduction
741 Improved use of trellis quantization on 2nd order Y blocks
742 Tune effect of motion on KF/GF boost in two pass
743 Allow coefficient optimization for good quality speed 0.
744 Improved control of active min quantizer for two pass.
745 Enable ARFs for non-lagged compress
749 Disable frame dropping by default
750 Improved multithreaded performance
751 Improved Force Key Frame Behaviour
752 Increased rate control buffer level precision
753 Fix bug in 1st pass motion compensation
754 ivfenc: correct fixed kf interval, --disable-kf
756 Changed above and left context data layout
757 Rework idct calling structure.
758 Removed unnecessary MB_MODE_INFO copies
759 x86: SSSE3 sixtap prediction
760 Reworked IDCT to include reconstruction (add) step
761 Swap alt/gold/new/last frame buffer ptrs instead of copying.
762 Improve SSE2 loopfilter functions
763 Change bitreader to use a larger window.
764 Avoid loopfilter reinitialization when possible
766 Normalize quantizer's zero bin and rounding factors
767 Add trellis quantization.
768 Make the quantizer exact.
769 Updates to ARNR filtering algorithm
770 Fix breakout thresh computation for golden & AltRef frames
771 Redo the forward 4x4 dct
772 Improve the accuracy of forward walsh-hadamard transform
773 Further adjustment of RD behaviour with Q and Zbin.
775 Allow linking of libs built with MinGW to MSVC
776 Fix target auto-detection on mingw32
777 Allow --cpu= to work for x86.
778 configure: pass original arguments through to make dist
779 Fix builds without runtime CPU detection
780 msvs: fix install of codec sources
781 msvs: Change devenv.com command line for better msys support
782 msvs: Add vs9 targets.
783 Add x86_64-linux-icc target
785 Potential crashes on older MinGW builds
786 Fix two-pass framrate for Y4M input.
787 Fixed simple loop filter, other crashes on ARM v6
788 arm: fix missing dependency with --enable-shared
789 configure: support directories containing .o
790 Replace pinsrw (SSE) with MMX instructions
791 apple: include proper mach primatives
792 Fixed rate control bug with long key frame interval.
793 Fix DSO link errors on x86-64 when not using a version script
794 Fixed buffer selection for UV in AltRef filtering
799 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
800 * Speed optimizations
803 * Prevent out-of-bounds accesses on invalid data
804 - Build system updates:
805 * Detect toolchain to be used automatically for native builds
806 * Support building shared libraries
807 * Better autotools emulation (--prefix, --libdir, DESTDIR)
809 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
813 - Initial open source release. Welcome to WebM and VP8!