1 2019-07-15 v1.8.1 "Orpington Duck"
2 This release collects incremental improvements to many aspects of the library.
5 VP8E_SET_CPUUSED now accepts values up to 9 for vp9.
6 VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT had a spelling fix (was VP8E).
7 The --sdk-path option has been removed. If you were using it to build for
8 Android please read build/make/Android.mk for alternatives.
9 All PPC optimizations have been disabled:
10 https://bugs.chromium.org/p/webm/issues/detail?id=1522.
13 Various changes to improve encoder rate control, quality and speed
14 for practically every use case.
17 vp9-rtc: Fix color artifacts for speed >= 8.
19 2019-01-31 v1.8.0 "Northern Shoveler Duck"
20 This release focused on encoding performance for realtime and VOD use cases.
23 This adds and improves several vp9 controls. Most are related to SVC:
24 VP9E_SET_SVC_FRAME_DROP_LAYER:
25 - Frame dropping in SVC.
26 VP9E_SET_SVC_INTER_LAYER_PRED:
27 - Inter-layer prediction in SVC.
28 VP9E_SET_SVC_GF_TEMPORAL_REF:
29 - Enable long term temporal reference in SVC.
30 VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG:
31 - Extend and improve this control for better flexibility in setting SVC
33 VP9E_SET_POSTENCODE_DROP:
34 - Allow for post-encode frame dropping (applies to non-SVC too).
35 VP9E_SET_SVC_SPATIAL_LAYER_SYNC:
36 - Enable spatial layer sync frames.
37 VP9E_SET_SVC_LAYER_ID:
38 - Extend api to specify temporal id for each spatial layers.
40 - Extend Region of Interest functionality to VP9.
43 2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6,
44 we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1,
45 the gains are approximately 4% for VBR and 5% for CQ.
47 For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at
48 screen sharing have improved when the content changes significantly (slide
49 sharing) or scrolls. There is a new speed 9 setting for mobile devices which
50 is about 10-20% faster than speed 8.
54 VP9 partition issue for 1080p.
55 VP9 rate control improvments.
56 Postprocessing Multi Frame Quality Enhancement (MFQE) issue.
57 VP8 multithread decoder issues.
58 A variety of fuzzing issues.
60 2018-01-04 v1.7.0 "Mandarin Duck"
61 This release focused on high bit depth performance (10/12 bit) and vp9
62 encoding improvements.
65 This release is ABI incompatible due to new vp9 encoder features.
67 Frame parallel decoding for vp9 has been removed.
70 vp9 encoding supports additional threads with --row-mt. This can be greater
71 than the number of tiles.
73 Two new vp9 encoder options have been added:
77 Additional tooling for respecting the vp9 "level" profiles has been added.
80 A variety of fuzzing issues.
81 vp8 threading fix for ARM.
82 Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
83 Reject invalid multi resolution configurations.
85 2017-01-09 v1.6.1 "Long Tailed Duck"
86 This release improves upon the VP9 encoder and speeds up the encoding and
90 This release is ABI compatible with 1.6.0.
93 Faster VP9 encoding and decoding.
94 High bit depth builds now provide similar speed for 8 bit encode and decode
95 for x86 targets. Other platforms and higher bit depth improvements are in
99 A variety of fuzzing issues.
101 2016-07-20 v1.6.0 "Khaki Campbell Duck"
102 This release improves upon the VP9 encoder and speeds up the encoding and
106 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
107 in vpx_image and some minor changes to the VP8_COMP structure.
109 The default key frame interval for VP9 has changed from 128 to 9999.
112 A core focus has been performance for low end Intel processors. SSSE3
113 instructions such as 'pshufb' have been avoided and instructions have been
114 reordered to better accommodate the more constrained pipelines.
116 As a result, devices based on Celeron processors have seen substantial
117 decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
118 decoding speed improved between 10 and 30%. Between Javan Whistling Duck
119 and Khaki Campbell Duck, it improved another 10 to 15%.
121 While Celeron benefited most, Core-i5 also improved 5% and 10% between the
124 Realtime performance for WebRTC for both speed and quality has received a
128 A number of fuzzing issues, found variously by Mozilla, Chromium and others,
129 have been fixed and we strongly recommend updating.
131 2015-11-09 v1.5.0 "Javan Whistling Duck"
132 This release improves upon the VP9 encoder and speeds up the encoding and
136 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
137 controls and adds a variety of VP9 controls for testing.
139 The vpxenc utility now prefers VP9 by default.
142 Faster VP9 encoding and decoding
143 Smaller library size by combining functions used by VP8 and VP9
146 A variety of fuzzing issues
148 2015-04-03 v1.4.0 "Indian Runner Duck"
149 This release includes significant improvements to the VP9 codec.
152 This release is ABI incompatible with 1.3.0. It drops the compatibility
153 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
157 Faster VP9 encoding and decoding
158 Multithreaded VP9 decoding (tile and frame-based)
159 Multithreaded VP9 encoding - on by default
160 YUV 4:2:2 and 4:4:4 support in VP9
161 10 and 12bit support in VP9
162 64bit ARM support by replacing ARM assembly with intrinsics
165 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
169 Frame Parallel decoding fails for segmented and non-420 files.
171 2013-11-15 v1.3.0 "Forest"
172 This release introduces the VP9 codec in a backward-compatible way.
173 All existing users of VP8 can continue to use the library without
174 modification. However, some VP8 options do not map to VP9 in the same manner.
176 The VP9 encoder in this release is not feature complete. Users interested in
177 the encoder are advised to use the git master branch and discuss issues on
178 libvpx mailing lists.
181 This release is ABI and API compatible with Duclair (v1.0.0). Users
182 of older releases should refer to the Upgrading notes in this document
186 Get rid of bashisms in the main build scripts
187 Added usage info on command line options
188 Add lossless compression mode
190 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
191 Add option to disable documentation
192 configure: add --enable-external-build support
193 make: support V=1 as short form of verbose=yes
194 configure: support mingw-w64
195 configure: support hardfloat armv7 CHOSTS
196 configure: add support for android x86
197 Add estimated completion time to vpxenc
198 Don't exit on decode errors in vpxenc
199 vpxenc: support scaling prior to encoding
200 vpxdec: support scaling output
201 vpxenc: improve progress indicators with --skip
202 msvs: Don't link to winmm.lib
203 Add a new script for producing vcxproj files
204 Produce Visual Studio 10 and 11 project files
205 Produce Windows Phone project files
206 msvs-build: use msbuild for vs >= 2005
207 configure: default configure log to config.log
208 Add encoding option --static-thresh
211 Miscellaneous speed optimizations for VP8 and VP9.
214 In general, quality is consistent with the Eider release.
217 This release represents approximately a year of engineering effort,
218 and contains multiple bug fixes. Please refer to git history for details.
222 This release acts as a checkpoint for a large amount of internal refactoring
223 and testing. It also contains a number of small bugfixes, so all users are
224 encouraged to upgrade.
227 This release is ABI and API compatible with Duclair (v1.0.0). Users
228 of older releases should refer to the Upgrading notes in this
229 document for that release.
232 VP8 optimizations for MIPS dspr2
233 vpxenc: add -quiet option
236 Encoder and decoder speed is consistent with the Eider release.
239 In general, quality is consistent with the Eider release.
241 Minor tweaks to ARNR filtering
242 Minor improvements to real time encoding with multiple temporal layers
245 Fixes multithreaded encoder race condition in loopfilter
246 Fixes multi-resolution threaded encoding
247 Fix potential encoder dead-lock after picture resize
250 2012-05-09 v1.1.0 "Eider"
251 This introduces a number of enhancements, mostly focused on real-time
252 encoding. In addition, it fixes a decoder bug (first introduced in
253 Duclair) so all users of that release are encouraged to upgrade.
256 This release is ABI and API compatible with Duclair (v1.0.0). Users
257 of older releases should refer to the Upgrading notes in this
258 document for that release.
260 This release introduces a new temporal denoiser, controlled by the
261 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
262 currently take a strength parameter, so the control is effectively
263 a boolean - zero (off) or non-zero (on). For compatibility with
264 existing applications, the values accepted are the same as those
265 for the spatial denoiser (0-6). The temporal denoiser is enabled
266 by default, and the older spatial denoiser may be restored by
267 configuring with --disable-temporal-denoising. The temporal denoiser
268 is more computationally intensive than the spatial one.
270 This release removes support for a legacy, decode only API that was
271 supported, but deprecated, at the initial release of libvpx
272 (v0.9.0). This is not expected to have any impact. If you are
273 impacted, you can apply a reversion to commit 2bf8fb58 locally.
274 Please update to the latest libvpx API if you are affected.
277 Adds a motion compensated temporal denoiser to the encoder, which
278 gives higher quality than the older spatial denoiser. (See above
279 for notes on upgrading).
281 In addition, support for new compilers and platforms were added,
283 improved support for XCode
284 Android x86 NDK build
288 Changing resolution with vpx_codec_enc_config_set() is now
289 supported. Previously, reinitializing the codec was required to
290 change the input resolution.
292 The vpxenc application has initial support for producing multiple
293 encodes from the same input in one call. Resizing is not yet
294 supported, but varying other codec parameters is. Use -- to
295 delineate output streams. Options persist from one stream to the
298 Also, the vpxenc application will now use a keyframe interval of
299 5 seconds by default. Use the --kf-max-dist option to override.
302 Decoder performance improved 2.5% versus Duclair. Encoder speed is
303 consistent with Duclair for most material. Two pass encoding of
304 slideshow-like material will see significant improvements.
306 Large realtime encoding speed gains at a small quality expense are
307 possible by configuring the on-the-fly bitpacking experiment with
308 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
309 faster (ARM) depending on the number of threads and bitrate
310 settings. This technique sees constant gain over the 5-16 speed
311 range. For VC style input the loss seen is up to 0.2dB. See commit
312 52cf4dca for further details.
315 On the whole, quality is consistent with the Duclair release. Some
318 Reduced blockiness in easy sections by applying a penalty to
321 Improved quality of static sections (like slideshows) with
324 Improved keyframe sizing with multiple temporal layers
327 Corrected alt-ref contribution to frame rate for visible updates
328 to the alt-ref buffer. This affected applications making manual
329 usage of the frame reference flags, or temporal layers.
331 Additional constraints were added to disable multi-frame quality
332 enhancement (MFQE) in sections of the frame where there is motion.
335 Fixed corruption issues when vpx_codec_enc_config_set() was called
336 with spatial resampling enabled.
338 Fixed a decoder error introduced in Duclair where the segmentation
339 map was not being reinitialized on keyframes (#378)
342 2012-01-27 v1.0.0 "Duclair"
343 Our fourth named release, focused on performance and features related to
344 real-time encoding. It also fixes a decoder crash bug introduced in
345 v0.9.7, so all users of that release are encouraged to upgrade.
348 This release is ABI incompatible with prior releases of libvpx, so the
349 "major" version number has been bumped to 1. You must recompile your
350 applications against the latest version of the libvpx headers. The
351 API remains compatible, and this should not require code changes in most
355 This release introduces several substantial new features to the encoder,
356 of particular interest to real time streaming applications.
358 Temporal scalability allows the encoder to produce a stream that can
359 be decimated to different frame rates, with independent rate targetting
362 Multiframe quality enhancement postprocessing can make visual quality
363 more consistent in the presence of frames that are substantially
364 different quality than the surrounding frames, as in the temporal
365 scalability case and in some forced keyframe scenarios.
367 Multiple-resolution encoding support allows the encoding of the
368 same content at different resolutions faster than encoding them
372 Optimization targets for this release included the decoder and the real-
373 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
374 this release. Encoder improvements followed a curve where speeds 1-3
375 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
376 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
380 Encoder quality in the single stream case is consistent with the Cayuga
384 This release fixes an OOB read decoder crash bug present in v0.9.7
385 related to the clamping of motion vectors in SPLITMV blocks. This
386 behavior could be triggered by corrupt input or by starting
387 decoding from a P-frame.
390 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
391 This is an incremental bugfix release against Cayuga. All users of that
392 release are strongly encouraged to upgrade.
394 - Fix potential OOB reads (cdae03a)
396 An unbounded out of bounds read was discovered when the
397 decoder was requested to perform error concealment (new in
398 Cayuga) given a frame with corrupt partition sizes.
400 A bounded out of bounds read was discovered affecting all
401 versions of libvpx. Given an multipartition input frame that
402 is truncated between the mode/mv partition and the first
403 residiual paritition (in the block of partition offsets), up
404 to 3 extra bytes could have been read from the source buffer.
405 The code will not take any action regardless of the contents
406 of these undefined bytes, as the truncated buffer is detected
407 immediately following the read based on the calculated
408 starting position of the coefficient partition.
410 - Fix potential error concealment crash when the very first frame
411 is missing or corrupt (a609be5)
413 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
415 - Revert 1-pass CBR rate control changes (e961317)
416 Further testing showed this change produced undesirable visual
417 artifacts, rolling back for now.
420 2011-08-02 v0.9.7 "Cayuga"
421 Our third named release, focused on a faster, higher quality, encoder.
424 This release is backwards compatible with Aylesbury (v0.9.5) and
425 Bali (v0.9.6). Users of older releases should refer to the Upgrading
426 notes in this document for that release.
429 Stereo 3D format support for vpxenc
430 Runtime detection of available processor cores.
431 Allow specifying --end-usage by enum name
432 vpxdec: test for frame corruption
433 vpxenc: add quantizer histogram display
434 vpxenc: add rate histogram display
435 Set VPX_FRAME_IS_DROPPABLE
436 update configure for ios sdk 4.3
437 Avoid text relocations in ARM vp8 decoder
438 Generate a vpx.pc file for pkg-config.
439 New ways of passing encoded data between encoder and decoder.
442 This release includes across-the-board speed improvements to the
443 encoder. On x86, these measure at approximately 11.5% in Best mode,
444 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
445 On ARM Cortex A9 with Neon extensions, real-time encoding of video
446 telephony content is 35% faster than Bali on single core and 48%
447 faster on multi-core. On the NVidia Tegra2 platform, real time
448 encoding is 40% faster than Bali.
450 Decoder speed was not a priority for this release, but improved
451 approximately 8.4% on x86.
453 Reduce motion vector search on alt-ref frame.
454 Encoder loopfilter running in its own thread
455 Reworked loopfilter to precalculate more parameters
456 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
457 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
458 Removed redundant checks
459 Reduced structure sizes
460 utilize preload in ARMv6 MC/LPF/Copy routines
461 ARM optimized quantization, dfct, variance, subtract
462 Increase chrow row alignment to 16 bytes.
463 disable trellis optimization for first pass
464 Write SSSE3 sub-pixel filter function
465 Improve SSE2 half-pixel filter funtions
466 Add vp8_sub_pixel_variance16x8_ssse3 function
467 Reduce unnecessary distortion computation
468 Use diamond search to replace full search
469 Preload reference area in sub-pixel motion search (real-time mode)
472 This release focused primarily on one-pass use cases, including
473 video conferencing. Low latency data rate control was significantly
474 improved, improving streamability over bandwidth constrained links.
475 Added support for error concealment, allowing frames to maintain
476 visual quality in the presence of substantial packet loss.
478 Add rc_max_intra_bitrate_pct control
479 Limit size of initial keyframe in one-pass.
480 Improve framerate adaptation
481 Improved 1-pass CBR rate control
482 Improved KF insertion after fades to still.
483 Improved key frame detection.
484 Improved activity masking (lower PSNR impact for same SSIM boost)
485 Improved interaction between GF and ARFs
486 Adding error-concealment to the decoder.
487 Adding support for independent partitions
488 Adjusted rate-distortion constants
492 Removed firstpass motion map
493 Fix parallel make install
494 Fix multithreaded encoding for 1 MB wide frame
495 Fixed iwalsh_neon build problems with RVDS4.1
496 Fix semaphore emulation, spin-wait intrinsics on Windows
497 Fix build with xcode4 and simplify GLOBAL.
498 Mark ARM asm objects as allowing a non-executable stack.
499 Fix vpxenc encoding incorrect webm file header on big endian
502 2011-03-07 v0.9.6 "Bali"
503 Our second named release, focused on a faster, higher quality, encoder.
506 This release is backwards compatible with Aylesbury (v0.9.5). Users
507 of older releases should refer to the Upgrading notes in this
508 document for that release.
511 vpxenc --psnr shows a summary when encode completes
512 --tune=ssim option to enable activity masking
513 improved postproc visualizations for development
514 updated support for Apple iOS to SDK 4.2
515 query decoder to determine which reference frames were updated
516 implemented error tracking in the decoder
517 fix pipe support on windows
520 Primary focus was on good quality mode, speed 0. Average improvement
521 on x86 about 40%, up to 100% on user-generated content at that speed.
522 Best quality mode speed improved 35%, and realtime speed 10-20%. This
523 release also saw significant improvement in realtime encoding speed
526 Improved encoder threading
527 Dont pick encoder filter level when loopfilter is disabled.
528 Avoid double copying of key frames into alt and golden buffer
530 x86 sse2 temporal filter
531 SSSE3 version of fast quantizer
532 vp8_rd_pick_best_mbsegmentation code restructure
533 Adjusted breakout RD for SPLITMV
534 Changed segmentation check order
535 Improved rd_pick_intra4x4block
536 Adds armv6 optimized variance calculation
537 ARMv6 optimized sad16x16
538 ARMv6 optimized half pixel variance calculations
539 Full search SAD function optimization in SSE4.1
540 Improve MV prediction accuracy to achieve performance gain
541 Improve MV prediction in vp8_pick_inter_mode() for speed>3
544 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
545 also includes support for "activity masking," which greatly improves
546 SSIM at the expense of PSNR. For now, this feature is available with
547 the --tune=ssim option. Further experimentation in this area
548 is ongoing. This release also introduces a new rate control mode
549 called "CQ," which changes the allocation of bits within a clip to
550 the sections where they will have the most visual impact.
552 Tuning for the more exact quantizer.
553 Relax rate control for last few frames
555 Limit key frame quantizer for forced key frames.
557 Add simple version of activity masking.
558 make rdmult adaptive for intra in quantizer RDO
559 cap the best quantizer for 2nd order DC
560 change the threshold of DC check for encode breakout
563 Fix crash on Sparc Solaris.
564 Fix counter of fixed keyframe distance
565 ARNR filter pointer update bug fix
566 Fixed use of motion percentage in KF/GF group calc
567 Changed condition for using RD in Intra Mode
568 Fix encoder real-time only configuration.
569 Fix ARM encoder crash with multiple token partitions
570 Fixed bug first cluster timecode of webm file is wrong.
571 Fixed various encoder bugs with odd-sized images
572 vp8e_get_preview fixed when spatial resampling enabled
573 quantizer: fix assertion in fast quantizer path
574 Allocate source buffers to be multiples of 16
575 Fix for manual Golden frame frequency
576 Fix drastic undershoot in long form content
579 2010-10-28 v0.9.5 "Aylesbury"
580 Our first named release, focused on a faster decoder, and a better encoder.
583 This release incorporates backwards-incompatible changes to the
584 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
587 * the -q (quiet) option has been removed, and replaced with
588 -v (verbose). the output is quiet by default. Use -v to see
589 the version number of the binary.
591 * The default behavior is now to write output to a single file
592 instead of individual frames. The -y option has been removed.
593 Y4M output is the default.
595 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
596 options must be specified.
598 $ ivfdec -o OUTPUT INPUT
599 $ vpxdec --i420 -o OUTPUT INPUT
601 * If an output file is not specified, the default is to write
602 Y4M to stdout. This makes piping more natural.
604 $ ivfdec -y -o - INPUT | ...
607 * The output file has additional flexibility for formatting the
608 filename. It supports escape characters for constructing a
609 filename from the width, height, and sequence number. This
610 replaces the -p option. To get the equivalent:
612 $ ivfdec -p frame INPUT
613 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
616 * The output file must be specified with -o, rather than as the
619 $ ivfenc <options> INPUT OUTPUT
620 $ vpxenc <options> -o OUTPUT INPUT
622 * The output defaults to webm. To get IVF output, use the --ivf
625 $ ivfenc <options> INPUT OUTPUT.ivf
626 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
630 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
631 vpxdec supports .webm input
632 vpxdec writes .y4m by default
633 vpxenc writes .webm output by default
634 vpxenc --psnr now shows the average/overall PSNR at the end
635 ARM platforms now support runtime cpu detection
636 vpxdec visualizations added for motion vectors, block modes, references
637 vpxdec now silent by default
638 vpxdec --progress shows frame-by-frame timing information
639 vpxenc supports the distinction between --fps and --timebase
640 NASM is now a supported assembler
641 configure: enable PIC for shared libs by default
642 configure: add --enable-small
643 configure: support for ppc32-linux-gcc
644 configure: support for sparc-solaris-gcc
647 Improve handling of invalid frames
648 Fix valgrind errors in the NEON loop filters.
649 Fix loopfilter delta zero transitions
650 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
651 Build fixes for darwin-icc
654 20-40% (average 28%) improvement in libvpx decoder speed,
656 Rewrite vp8_short_walsh4x4_sse2()
657 Optimizations on the loopfilters.
658 Miscellaneous improvements for Atom
659 Add 4-tap version of 2nd-pass ARMv6 MC filter.
660 Improved multithread utilization
661 Better instruction choices on x86
662 reorder data to use wider instructions
663 Update NEON wide idcts
664 Make block access to frame buffer sequential
665 Improved subset block search
666 Bilinear subpixel optimizations for ssse3.
667 Decrease memory footprint
669 Encoder speed improvements (percentage gain not measured):
670 Skip unnecessary search of identical frames
671 Add SSE2 subtract functions
672 Improve bounds checking in vp8_diamond_search_sadx4()
673 Added vp8_fast_quantize_b_sse2
676 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
677 encoding mode, and up to 60% improvement on very noisy, still
678 or slow moving source video
680 Motion compensated temporal filter for Alt-Ref Noise Reduction
681 Improved use of trellis quantization on 2nd order Y blocks
682 Tune effect of motion on KF/GF boost in two pass
683 Allow coefficient optimization for good quality speed 0.
684 Improved control of active min quantizer for two pass.
685 Enable ARFs for non-lagged compress
689 Disable frame dropping by default
690 Improved multithreaded performance
691 Improved Force Key Frame Behaviour
692 Increased rate control buffer level precision
693 Fix bug in 1st pass motion compensation
694 ivfenc: correct fixed kf interval, --disable-kf
696 Changed above and left context data layout
697 Rework idct calling structure.
698 Removed unnecessary MB_MODE_INFO copies
699 x86: SSSE3 sixtap prediction
700 Reworked IDCT to include reconstruction (add) step
701 Swap alt/gold/new/last frame buffer ptrs instead of copying.
702 Improve SSE2 loopfilter functions
703 Change bitreader to use a larger window.
704 Avoid loopfilter reinitialization when possible
706 Normalize quantizer's zero bin and rounding factors
707 Add trellis quantization.
708 Make the quantizer exact.
709 Updates to ARNR filtering algorithm
710 Fix breakout thresh computation for golden & AltRef frames
711 Redo the forward 4x4 dct
712 Improve the accuracy of forward walsh-hadamard transform
713 Further adjustment of RD behaviour with Q and Zbin.
715 Allow linking of libs built with MinGW to MSVC
716 Fix target auto-detection on mingw32
717 Allow --cpu= to work for x86.
718 configure: pass original arguments through to make dist
719 Fix builds without runtime CPU detection
720 msvs: fix install of codec sources
721 msvs: Change devenv.com command line for better msys support
722 msvs: Add vs9 targets.
723 Add x86_64-linux-icc target
725 Potential crashes on older MinGW builds
726 Fix two-pass framrate for Y4M input.
727 Fixed simple loop filter, other crashes on ARM v6
728 arm: fix missing dependency with --enable-shared
729 configure: support directories containing .o
730 Replace pinsrw (SSE) with MMX instructions
731 apple: include proper mach primatives
732 Fixed rate control bug with long key frame interval.
733 Fix DSO link errors on x86-64 when not using a version script
734 Fixed buffer selection for UV in AltRef filtering
739 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
740 * Speed optimizations
743 * Prevent out-of-bounds accesses on invalid data
744 - Build system updates:
745 * Detect toolchain to be used automatically for native builds
746 * Support building shared libraries
747 * Better autotools emulation (--prefix, --libdir, DESTDIR)
749 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
753 - Initial open source release. Welcome to WebM and VP8!