1 2019-01-31 v1.8.0 "Northern Shoveler Duck"
2 This release focused on encoding performance for realtime and VOD use cases.
5 This adds and improves several vp9 controls. Most are related to SVC:
6 VP9E_SET_SVC_FRAME_DROP_LAYER:
7 - Frame dropping in SVC.
8 VP9E_SET_SVC_INTER_LAYER_PRED:
9 - Inter-layer prediction in SVC.
10 VP9E_SET_SVC_GF_TEMPORAL_REF:
11 - Enable long term temporal reference in SVC.
12 VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG:
13 - Extend and improve this control for better flexibility in setting SVC
15 VP9E_SET_POSTENCODE_DROP:
16 - Allow for post-encode frame dropping (applies to non-SVC too).
17 VP9E_SET_SVC_SPATIAL_LAYER_SYNC:
18 - Enable spatial layer sync frames.
19 VP9E_SET_SVC_LAYER_ID:
20 - Extend api to specify temporal id for each spatial layers.
22 - Extend Region of Interest functionality to VP9.
25 2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6,
26 we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1,
27 the gains are approximately 4% for VBR and 5% for CQ.
29 For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at
30 screen sharing have improved when the content changes significantly (slide
31 sharing) or scrolls. There is a new speed 9 setting for mobile devices which
32 is about 10-20% faster than speed 8.
36 VP9 partition issue for 1080p.
37 VP9 rate control improvments.
38 Postprocessing Multi Frame Quality Enhancement (MFQE) issue.
39 VP8 multithread decoder issues.
40 A variety of fuzzing issues.
42 2018-01-04 v1.7.0 "Mandarin Duck"
43 This release focused on high bit depth performance (10/12 bit) and vp9
44 encoding improvements.
47 This release is ABI incompatible due to new vp9 encoder features.
49 Frame parallel decoding for vp9 has been removed.
52 vp9 encoding supports additional threads with --row-mt. This can be greater
53 than the number of tiles.
55 Two new vp9 encoder options have been added:
59 Additional tooling for respecting the vp9 "level" profiles has been added.
62 A variety of fuzzing issues.
63 vp8 threading fix for ARM.
64 Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
65 Reject invalid multi resolution configurations.
67 2017-01-09 v1.6.1 "Long Tailed Duck"
68 This release improves upon the VP9 encoder and speeds up the encoding and
72 This release is ABI compatible with 1.6.0.
75 Faster VP9 encoding and decoding.
76 High bit depth builds now provide similar speed for 8 bit encode and decode
77 for x86 targets. Other platforms and higher bit depth improvements are in
81 A variety of fuzzing issues.
83 2016-07-20 v1.6.0 "Khaki Campbell Duck"
84 This release improves upon the VP9 encoder and speeds up the encoding and
88 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
89 in vpx_image and some minor changes to the VP8_COMP structure.
91 The default key frame interval for VP9 has changed from 128 to 9999.
94 A core focus has been performance for low end Intel processors. SSSE3
95 instructions such as 'pshufb' have been avoided and instructions have been
96 reordered to better accommodate the more constrained pipelines.
98 As a result, devices based on Celeron processors have seen substantial
99 decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
100 decoding speed improved between 10 and 30%. Between Javan Whistling Duck
101 and Khaki Campbell Duck, it improved another 10 to 15%.
103 While Celeron benefited most, Core-i5 also improved 5% and 10% between the
106 Realtime performance for WebRTC for both speed and quality has received a
110 A number of fuzzing issues, found variously by Mozilla, Chromium and others,
111 have been fixed and we strongly recommend updating.
113 2015-11-09 v1.5.0 "Javan Whistling Duck"
114 This release improves upon the VP9 encoder and speeds up the encoding and
118 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
119 controls and adds a variety of VP9 controls for testing.
121 The vpxenc utility now prefers VP9 by default.
124 Faster VP9 encoding and decoding
125 Smaller library size by combining functions used by VP8 and VP9
128 A variety of fuzzing issues
130 2015-04-03 v1.4.0 "Indian Runner Duck"
131 This release includes significant improvements to the VP9 codec.
134 This release is ABI incompatible with 1.3.0. It drops the compatibility
135 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
139 Faster VP9 encoding and decoding
140 Multithreaded VP9 decoding (tile and frame-based)
141 Multithreaded VP9 encoding - on by default
142 YUV 4:2:2 and 4:4:4 support in VP9
143 10 and 12bit support in VP9
144 64bit ARM support by replacing ARM assembly with intrinsics
147 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
151 Frame Parallel decoding fails for segmented and non-420 files.
153 2013-11-15 v1.3.0 "Forest"
154 This release introduces the VP9 codec in a backward-compatible way.
155 All existing users of VP8 can continue to use the library without
156 modification. However, some VP8 options do not map to VP9 in the same manner.
158 The VP9 encoder in this release is not feature complete. Users interested in
159 the encoder are advised to use the git master branch and discuss issues on
160 libvpx mailing lists.
163 This release is ABI and API compatible with Duclair (v1.0.0). Users
164 of older releases should refer to the Upgrading notes in this document
168 Get rid of bashisms in the main build scripts
169 Added usage info on command line options
170 Add lossless compression mode
172 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
173 Add option to disable documentation
174 configure: add --enable-external-build support
175 make: support V=1 as short form of verbose=yes
176 configure: support mingw-w64
177 configure: support hardfloat armv7 CHOSTS
178 configure: add support for android x86
179 Add estimated completion time to vpxenc
180 Don't exit on decode errors in vpxenc
181 vpxenc: support scaling prior to encoding
182 vpxdec: support scaling output
183 vpxenc: improve progress indicators with --skip
184 msvs: Don't link to winmm.lib
185 Add a new script for producing vcxproj files
186 Produce Visual Studio 10 and 11 project files
187 Produce Windows Phone project files
188 msvs-build: use msbuild for vs >= 2005
189 configure: default configure log to config.log
190 Add encoding option --static-thresh
193 Miscellaneous speed optimizations for VP8 and VP9.
196 In general, quality is consistent with the Eider release.
199 This release represents approximately a year of engineering effort,
200 and contains multiple bug fixes. Please refer to git history for details.
204 This release acts as a checkpoint for a large amount of internal refactoring
205 and testing. It also contains a number of small bugfixes, so all users are
206 encouraged to upgrade.
209 This release is ABI and API compatible with Duclair (v1.0.0). Users
210 of older releases should refer to the Upgrading notes in this
211 document for that release.
214 VP8 optimizations for MIPS dspr2
215 vpxenc: add -quiet option
218 Encoder and decoder speed is consistent with the Eider release.
221 In general, quality is consistent with the Eider release.
223 Minor tweaks to ARNR filtering
224 Minor improvements to real time encoding with multiple temporal layers
227 Fixes multithreaded encoder race condition in loopfilter
228 Fixes multi-resolution threaded encoding
229 Fix potential encoder dead-lock after picture resize
232 2012-05-09 v1.1.0 "Eider"
233 This introduces a number of enhancements, mostly focused on real-time
234 encoding. In addition, it fixes a decoder bug (first introduced in
235 Duclair) so all users of that release are encouraged to upgrade.
238 This release is ABI and API compatible with Duclair (v1.0.0). Users
239 of older releases should refer to the Upgrading notes in this
240 document for that release.
242 This release introduces a new temporal denoiser, controlled by the
243 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
244 currently take a strength parameter, so the control is effectively
245 a boolean - zero (off) or non-zero (on). For compatibility with
246 existing applications, the values accepted are the same as those
247 for the spatial denoiser (0-6). The temporal denoiser is enabled
248 by default, and the older spatial denoiser may be restored by
249 configuring with --disable-temporal-denoising. The temporal denoiser
250 is more computationally intensive than the spatial one.
252 This release removes support for a legacy, decode only API that was
253 supported, but deprecated, at the initial release of libvpx
254 (v0.9.0). This is not expected to have any impact. If you are
255 impacted, you can apply a reversion to commit 2bf8fb58 locally.
256 Please update to the latest libvpx API if you are affected.
259 Adds a motion compensated temporal denoiser to the encoder, which
260 gives higher quality than the older spatial denoiser. (See above
261 for notes on upgrading).
263 In addition, support for new compilers and platforms were added,
265 improved support for XCode
266 Android x86 NDK build
270 Changing resolution with vpx_codec_enc_config_set() is now
271 supported. Previously, reinitializing the codec was required to
272 change the input resolution.
274 The vpxenc application has initial support for producing multiple
275 encodes from the same input in one call. Resizing is not yet
276 supported, but varying other codec parameters is. Use -- to
277 delineate output streams. Options persist from one stream to the
280 Also, the vpxenc application will now use a keyframe interval of
281 5 seconds by default. Use the --kf-max-dist option to override.
284 Decoder performance improved 2.5% versus Duclair. Encoder speed is
285 consistent with Duclair for most material. Two pass encoding of
286 slideshow-like material will see significant improvements.
288 Large realtime encoding speed gains at a small quality expense are
289 possible by configuring the on-the-fly bitpacking experiment with
290 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
291 faster (ARM) depending on the number of threads and bitrate
292 settings. This technique sees constant gain over the 5-16 speed
293 range. For VC style input the loss seen is up to 0.2dB. See commit
294 52cf4dca for further details.
297 On the whole, quality is consistent with the Duclair release. Some
300 Reduced blockiness in easy sections by applying a penalty to
303 Improved quality of static sections (like slideshows) with
306 Improved keyframe sizing with multiple temporal layers
309 Corrected alt-ref contribution to frame rate for visible updates
310 to the alt-ref buffer. This affected applications making manual
311 usage of the frame reference flags, or temporal layers.
313 Additional constraints were added to disable multi-frame quality
314 enhancement (MFQE) in sections of the frame where there is motion.
317 Fixed corruption issues when vpx_codec_enc_config_set() was called
318 with spatial resampling enabled.
320 Fixed a decoder error introduced in Duclair where the segmentation
321 map was not being reinitialized on keyframes (#378)
324 2012-01-27 v1.0.0 "Duclair"
325 Our fourth named release, focused on performance and features related to
326 real-time encoding. It also fixes a decoder crash bug introduced in
327 v0.9.7, so all users of that release are encouraged to upgrade.
330 This release is ABI incompatible with prior releases of libvpx, so the
331 "major" version number has been bumped to 1. You must recompile your
332 applications against the latest version of the libvpx headers. The
333 API remains compatible, and this should not require code changes in most
337 This release introduces several substantial new features to the encoder,
338 of particular interest to real time streaming applications.
340 Temporal scalability allows the encoder to produce a stream that can
341 be decimated to different frame rates, with independent rate targetting
344 Multiframe quality enhancement postprocessing can make visual quality
345 more consistent in the presence of frames that are substantially
346 different quality than the surrounding frames, as in the temporal
347 scalability case and in some forced keyframe scenarios.
349 Multiple-resolution encoding support allows the encoding of the
350 same content at different resolutions faster than encoding them
354 Optimization targets for this release included the decoder and the real-
355 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
356 this release. Encoder improvements followed a curve where speeds 1-3
357 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
358 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
362 Encoder quality in the single stream case is consistent with the Cayuga
366 This release fixes an OOB read decoder crash bug present in v0.9.7
367 related to the clamping of motion vectors in SPLITMV blocks. This
368 behavior could be triggered by corrupt input or by starting
369 decoding from a P-frame.
372 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
373 This is an incremental bugfix release against Cayuga. All users of that
374 release are strongly encouraged to upgrade.
376 - Fix potential OOB reads (cdae03a)
378 An unbounded out of bounds read was discovered when the
379 decoder was requested to perform error concealment (new in
380 Cayuga) given a frame with corrupt partition sizes.
382 A bounded out of bounds read was discovered affecting all
383 versions of libvpx. Given an multipartition input frame that
384 is truncated between the mode/mv partition and the first
385 residiual paritition (in the block of partition offsets), up
386 to 3 extra bytes could have been read from the source buffer.
387 The code will not take any action regardless of the contents
388 of these undefined bytes, as the truncated buffer is detected
389 immediately following the read based on the calculated
390 starting position of the coefficient partition.
392 - Fix potential error concealment crash when the very first frame
393 is missing or corrupt (a609be5)
395 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
397 - Revert 1-pass CBR rate control changes (e961317)
398 Further testing showed this change produced undesirable visual
399 artifacts, rolling back for now.
402 2011-08-02 v0.9.7 "Cayuga"
403 Our third named release, focused on a faster, higher quality, encoder.
406 This release is backwards compatible with Aylesbury (v0.9.5) and
407 Bali (v0.9.6). Users of older releases should refer to the Upgrading
408 notes in this document for that release.
411 Stereo 3D format support for vpxenc
412 Runtime detection of available processor cores.
413 Allow specifying --end-usage by enum name
414 vpxdec: test for frame corruption
415 vpxenc: add quantizer histogram display
416 vpxenc: add rate histogram display
417 Set VPX_FRAME_IS_DROPPABLE
418 update configure for ios sdk 4.3
419 Avoid text relocations in ARM vp8 decoder
420 Generate a vpx.pc file for pkg-config.
421 New ways of passing encoded data between encoder and decoder.
424 This release includes across-the-board speed improvements to the
425 encoder. On x86, these measure at approximately 11.5% in Best mode,
426 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
427 On ARM Cortex A9 with Neon extensions, real-time encoding of video
428 telephony content is 35% faster than Bali on single core and 48%
429 faster on multi-core. On the NVidia Tegra2 platform, real time
430 encoding is 40% faster than Bali.
432 Decoder speed was not a priority for this release, but improved
433 approximately 8.4% on x86.
435 Reduce motion vector search on alt-ref frame.
436 Encoder loopfilter running in its own thread
437 Reworked loopfilter to precalculate more parameters
438 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
439 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
440 Removed redundant checks
441 Reduced structure sizes
442 utilize preload in ARMv6 MC/LPF/Copy routines
443 ARM optimized quantization, dfct, variance, subtract
444 Increase chrow row alignment to 16 bytes.
445 disable trellis optimization for first pass
446 Write SSSE3 sub-pixel filter function
447 Improve SSE2 half-pixel filter funtions
448 Add vp8_sub_pixel_variance16x8_ssse3 function
449 Reduce unnecessary distortion computation
450 Use diamond search to replace full search
451 Preload reference area in sub-pixel motion search (real-time mode)
454 This release focused primarily on one-pass use cases, including
455 video conferencing. Low latency data rate control was significantly
456 improved, improving streamability over bandwidth constrained links.
457 Added support for error concealment, allowing frames to maintain
458 visual quality in the presence of substantial packet loss.
460 Add rc_max_intra_bitrate_pct control
461 Limit size of initial keyframe in one-pass.
462 Improve framerate adaptation
463 Improved 1-pass CBR rate control
464 Improved KF insertion after fades to still.
465 Improved key frame detection.
466 Improved activity masking (lower PSNR impact for same SSIM boost)
467 Improved interaction between GF and ARFs
468 Adding error-concealment to the decoder.
469 Adding support for independent partitions
470 Adjusted rate-distortion constants
474 Removed firstpass motion map
475 Fix parallel make install
476 Fix multithreaded encoding for 1 MB wide frame
477 Fixed iwalsh_neon build problems with RVDS4.1
478 Fix semaphore emulation, spin-wait intrinsics on Windows
479 Fix build with xcode4 and simplify GLOBAL.
480 Mark ARM asm objects as allowing a non-executable stack.
481 Fix vpxenc encoding incorrect webm file header on big endian
484 2011-03-07 v0.9.6 "Bali"
485 Our second named release, focused on a faster, higher quality, encoder.
488 This release is backwards compatible with Aylesbury (v0.9.5). Users
489 of older releases should refer to the Upgrading notes in this
490 document for that release.
493 vpxenc --psnr shows a summary when encode completes
494 --tune=ssim option to enable activity masking
495 improved postproc visualizations for development
496 updated support for Apple iOS to SDK 4.2
497 query decoder to determine which reference frames were updated
498 implemented error tracking in the decoder
499 fix pipe support on windows
502 Primary focus was on good quality mode, speed 0. Average improvement
503 on x86 about 40%, up to 100% on user-generated content at that speed.
504 Best quality mode speed improved 35%, and realtime speed 10-20%. This
505 release also saw significant improvement in realtime encoding speed
508 Improved encoder threading
509 Dont pick encoder filter level when loopfilter is disabled.
510 Avoid double copying of key frames into alt and golden buffer
512 x86 sse2 temporal filter
513 SSSE3 version of fast quantizer
514 vp8_rd_pick_best_mbsegmentation code restructure
515 Adjusted breakout RD for SPLITMV
516 Changed segmentation check order
517 Improved rd_pick_intra4x4block
518 Adds armv6 optimized variance calculation
519 ARMv6 optimized sad16x16
520 ARMv6 optimized half pixel variance calculations
521 Full search SAD function optimization in SSE4.1
522 Improve MV prediction accuracy to achieve performance gain
523 Improve MV prediction in vp8_pick_inter_mode() for speed>3
526 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
527 also includes support for "activity masking," which greatly improves
528 SSIM at the expense of PSNR. For now, this feature is available with
529 the --tune=ssim option. Further experimentation in this area
530 is ongoing. This release also introduces a new rate control mode
531 called "CQ," which changes the allocation of bits within a clip to
532 the sections where they will have the most visual impact.
534 Tuning for the more exact quantizer.
535 Relax rate control for last few frames
537 Limit key frame quantizer for forced key frames.
539 Add simple version of activity masking.
540 make rdmult adaptive for intra in quantizer RDO
541 cap the best quantizer for 2nd order DC
542 change the threshold of DC check for encode breakout
545 Fix crash on Sparc Solaris.
546 Fix counter of fixed keyframe distance
547 ARNR filter pointer update bug fix
548 Fixed use of motion percentage in KF/GF group calc
549 Changed condition for using RD in Intra Mode
550 Fix encoder real-time only configuration.
551 Fix ARM encoder crash with multiple token partitions
552 Fixed bug first cluster timecode of webm file is wrong.
553 Fixed various encoder bugs with odd-sized images
554 vp8e_get_preview fixed when spatial resampling enabled
555 quantizer: fix assertion in fast quantizer path
556 Allocate source buffers to be multiples of 16
557 Fix for manual Golden frame frequency
558 Fix drastic undershoot in long form content
561 2010-10-28 v0.9.5 "Aylesbury"
562 Our first named release, focused on a faster decoder, and a better encoder.
565 This release incorporates backwards-incompatible changes to the
566 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
569 * the -q (quiet) option has been removed, and replaced with
570 -v (verbose). the output is quiet by default. Use -v to see
571 the version number of the binary.
573 * The default behavior is now to write output to a single file
574 instead of individual frames. The -y option has been removed.
575 Y4M output is the default.
577 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
578 options must be specified.
580 $ ivfdec -o OUTPUT INPUT
581 $ vpxdec --i420 -o OUTPUT INPUT
583 * If an output file is not specified, the default is to write
584 Y4M to stdout. This makes piping more natural.
586 $ ivfdec -y -o - INPUT | ...
589 * The output file has additional flexibility for formatting the
590 filename. It supports escape characters for constructing a
591 filename from the width, height, and sequence number. This
592 replaces the -p option. To get the equivalent:
594 $ ivfdec -p frame INPUT
595 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
598 * The output file must be specified with -o, rather than as the
601 $ ivfenc <options> INPUT OUTPUT
602 $ vpxenc <options> -o OUTPUT INPUT
604 * The output defaults to webm. To get IVF output, use the --ivf
607 $ ivfenc <options> INPUT OUTPUT.ivf
608 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
612 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
613 vpxdec supports .webm input
614 vpxdec writes .y4m by default
615 vpxenc writes .webm output by default
616 vpxenc --psnr now shows the average/overall PSNR at the end
617 ARM platforms now support runtime cpu detection
618 vpxdec visualizations added for motion vectors, block modes, references
619 vpxdec now silent by default
620 vpxdec --progress shows frame-by-frame timing information
621 vpxenc supports the distinction between --fps and --timebase
622 NASM is now a supported assembler
623 configure: enable PIC for shared libs by default
624 configure: add --enable-small
625 configure: support for ppc32-linux-gcc
626 configure: support for sparc-solaris-gcc
629 Improve handling of invalid frames
630 Fix valgrind errors in the NEON loop filters.
631 Fix loopfilter delta zero transitions
632 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
633 Build fixes for darwin-icc
636 20-40% (average 28%) improvement in libvpx decoder speed,
638 Rewrite vp8_short_walsh4x4_sse2()
639 Optimizations on the loopfilters.
640 Miscellaneous improvements for Atom
641 Add 4-tap version of 2nd-pass ARMv6 MC filter.
642 Improved multithread utilization
643 Better instruction choices on x86
644 reorder data to use wider instructions
645 Update NEON wide idcts
646 Make block access to frame buffer sequential
647 Improved subset block search
648 Bilinear subpixel optimizations for ssse3.
649 Decrease memory footprint
651 Encoder speed improvements (percentage gain not measured):
652 Skip unnecessary search of identical frames
653 Add SSE2 subtract functions
654 Improve bounds checking in vp8_diamond_search_sadx4()
655 Added vp8_fast_quantize_b_sse2
658 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
659 encoding mode, and up to 60% improvement on very noisy, still
660 or slow moving source video
662 Motion compensated temporal filter for Alt-Ref Noise Reduction
663 Improved use of trellis quantization on 2nd order Y blocks
664 Tune effect of motion on KF/GF boost in two pass
665 Allow coefficient optimization for good quality speed 0.
666 Improved control of active min quantizer for two pass.
667 Enable ARFs for non-lagged compress
671 Disable frame dropping by default
672 Improved multithreaded performance
673 Improved Force Key Frame Behaviour
674 Increased rate control buffer level precision
675 Fix bug in 1st pass motion compensation
676 ivfenc: correct fixed kf interval, --disable-kf
678 Changed above and left context data layout
679 Rework idct calling structure.
680 Removed unnecessary MB_MODE_INFO copies
681 x86: SSSE3 sixtap prediction
682 Reworked IDCT to include reconstruction (add) step
683 Swap alt/gold/new/last frame buffer ptrs instead of copying.
684 Improve SSE2 loopfilter functions
685 Change bitreader to use a larger window.
686 Avoid loopfilter reinitialization when possible
688 Normalize quantizer's zero bin and rounding factors
689 Add trellis quantization.
690 Make the quantizer exact.
691 Updates to ARNR filtering algorithm
692 Fix breakout thresh computation for golden & AltRef frames
693 Redo the forward 4x4 dct
694 Improve the accuracy of forward walsh-hadamard transform
695 Further adjustment of RD behaviour with Q and Zbin.
697 Allow linking of libs built with MinGW to MSVC
698 Fix target auto-detection on mingw32
699 Allow --cpu= to work for x86.
700 configure: pass original arguments through to make dist
701 Fix builds without runtime CPU detection
702 msvs: fix install of codec sources
703 msvs: Change devenv.com command line for better msys support
704 msvs: Add vs9 targets.
705 Add x86_64-linux-icc target
707 Potential crashes on older MinGW builds
708 Fix two-pass framrate for Y4M input.
709 Fixed simple loop filter, other crashes on ARM v6
710 arm: fix missing dependency with --enable-shared
711 configure: support directories containing .o
712 Replace pinsrw (SSE) with MMX instructions
713 apple: include proper mach primatives
714 Fixed rate control bug with long key frame interval.
715 Fix DSO link errors on x86-64 when not using a version script
716 Fixed buffer selection for UV in AltRef filtering
721 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
722 * Speed optimizations
725 * Prevent out-of-bounds accesses on invalid data
726 - Build system updates:
727 * Detect toolchain to be used automatically for native builds
728 * Support building shared libraries
729 * Better autotools emulation (--prefix, --libdir, DESTDIR)
731 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
735 - Initial open source release. Welcome to WebM and VP8!