1 2019-12-09 v1.8.2 "Pekin Duck"
2 This release collects incremental improvements to many aspects of the library.
5 ARCH_* defines have been removed in favor of VPX_ARCH_*.
7 2019-07-15 v1.8.1 "Orpington Duck"
8 This release collects incremental improvements to many aspects of the library.
11 VP8E_SET_CPUUSED now accepts values up to 9 for vp9.
12 VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT had a spelling fix (was VP8E).
13 The --sdk-path option has been removed. If you were using it to build for
14 Android please read build/make/Android.mk for alternatives.
15 All PPC optimizations have been disabled:
16 https://bugs.chromium.org/p/webm/issues/detail?id=1522.
19 Various changes to improve encoder rate control, quality and speed
20 for practically every use case.
23 vp9-rtc: Fix color artifacts for speed >= 8.
25 2019-01-31 v1.8.0 "Northern Shoveler Duck"
26 This release focused on encoding performance for realtime and VOD use cases.
29 This adds and improves several vp9 controls. Most are related to SVC:
30 VP9E_SET_SVC_FRAME_DROP_LAYER:
31 - Frame dropping in SVC.
32 VP9E_SET_SVC_INTER_LAYER_PRED:
33 - Inter-layer prediction in SVC.
34 VP9E_SET_SVC_GF_TEMPORAL_REF:
35 - Enable long term temporal reference in SVC.
36 VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG:
37 - Extend and improve this control for better flexibility in setting SVC
39 VP9E_SET_POSTENCODE_DROP:
40 - Allow for post-encode frame dropping (applies to non-SVC too).
41 VP9E_SET_SVC_SPATIAL_LAYER_SYNC:
42 - Enable spatial layer sync frames.
43 VP9E_SET_SVC_LAYER_ID:
44 - Extend api to specify temporal id for each spatial layers.
46 - Extend Region of Interest functionality to VP9.
49 2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6,
50 we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1,
51 the gains are approximately 4% for VBR and 5% for CQ.
53 For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at
54 screen sharing have improved when the content changes significantly (slide
55 sharing) or scrolls. There is a new speed 9 setting for mobile devices which
56 is about 10-20% faster than speed 8.
60 VP9 partition issue for 1080p.
61 VP9 rate control improvments.
62 Postprocessing Multi Frame Quality Enhancement (MFQE) issue.
63 VP8 multithread decoder issues.
64 A variety of fuzzing issues.
66 2018-01-04 v1.7.0 "Mandarin Duck"
67 This release focused on high bit depth performance (10/12 bit) and vp9
68 encoding improvements.
71 This release is ABI incompatible due to new vp9 encoder features.
73 Frame parallel decoding for vp9 has been removed.
76 vp9 encoding supports additional threads with --row-mt. This can be greater
77 than the number of tiles.
79 Two new vp9 encoder options have been added:
83 Additional tooling for respecting the vp9 "level" profiles has been added.
86 A variety of fuzzing issues.
87 vp8 threading fix for ARM.
88 Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
89 Reject invalid multi resolution configurations.
91 2017-01-09 v1.6.1 "Long Tailed Duck"
92 This release improves upon the VP9 encoder and speeds up the encoding and
96 This release is ABI compatible with 1.6.0.
99 Faster VP9 encoding and decoding.
100 High bit depth builds now provide similar speed for 8 bit encode and decode
101 for x86 targets. Other platforms and higher bit depth improvements are in
105 A variety of fuzzing issues.
107 2016-07-20 v1.6.0 "Khaki Campbell Duck"
108 This release improves upon the VP9 encoder and speeds up the encoding and
112 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
113 in vpx_image and some minor changes to the VP8_COMP structure.
115 The default key frame interval for VP9 has changed from 128 to 9999.
118 A core focus has been performance for low end Intel processors. SSSE3
119 instructions such as 'pshufb' have been avoided and instructions have been
120 reordered to better accommodate the more constrained pipelines.
122 As a result, devices based on Celeron processors have seen substantial
123 decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
124 decoding speed improved between 10 and 30%. Between Javan Whistling Duck
125 and Khaki Campbell Duck, it improved another 10 to 15%.
127 While Celeron benefited most, Core-i5 also improved 5% and 10% between the
130 Realtime performance for WebRTC for both speed and quality has received a
134 A number of fuzzing issues, found variously by Mozilla, Chromium and others,
135 have been fixed and we strongly recommend updating.
137 2015-11-09 v1.5.0 "Javan Whistling Duck"
138 This release improves upon the VP9 encoder and speeds up the encoding and
142 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
143 controls and adds a variety of VP9 controls for testing.
145 The vpxenc utility now prefers VP9 by default.
148 Faster VP9 encoding and decoding
149 Smaller library size by combining functions used by VP8 and VP9
152 A variety of fuzzing issues
154 2015-04-03 v1.4.0 "Indian Runner Duck"
155 This release includes significant improvements to the VP9 codec.
158 This release is ABI incompatible with 1.3.0. It drops the compatibility
159 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
163 Faster VP9 encoding and decoding
164 Multithreaded VP9 decoding (tile and frame-based)
165 Multithreaded VP9 encoding - on by default
166 YUV 4:2:2 and 4:4:4 support in VP9
167 10 and 12bit support in VP9
168 64bit ARM support by replacing ARM assembly with intrinsics
171 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
175 Frame Parallel decoding fails for segmented and non-420 files.
177 2013-11-15 v1.3.0 "Forest"
178 This release introduces the VP9 codec in a backward-compatible way.
179 All existing users of VP8 can continue to use the library without
180 modification. However, some VP8 options do not map to VP9 in the same manner.
182 The VP9 encoder in this release is not feature complete. Users interested in
183 the encoder are advised to use the git master branch and discuss issues on
184 libvpx mailing lists.
187 This release is ABI and API compatible with Duclair (v1.0.0). Users
188 of older releases should refer to the Upgrading notes in this document
192 Get rid of bashisms in the main build scripts
193 Added usage info on command line options
194 Add lossless compression mode
196 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
197 Add option to disable documentation
198 configure: add --enable-external-build support
199 make: support V=1 as short form of verbose=yes
200 configure: support mingw-w64
201 configure: support hardfloat armv7 CHOSTS
202 configure: add support for android x86
203 Add estimated completion time to vpxenc
204 Don't exit on decode errors in vpxenc
205 vpxenc: support scaling prior to encoding
206 vpxdec: support scaling output
207 vpxenc: improve progress indicators with --skip
208 msvs: Don't link to winmm.lib
209 Add a new script for producing vcxproj files
210 Produce Visual Studio 10 and 11 project files
211 Produce Windows Phone project files
212 msvs-build: use msbuild for vs >= 2005
213 configure: default configure log to config.log
214 Add encoding option --static-thresh
217 Miscellaneous speed optimizations for VP8 and VP9.
220 In general, quality is consistent with the Eider release.
223 This release represents approximately a year of engineering effort,
224 and contains multiple bug fixes. Please refer to git history for details.
228 This release acts as a checkpoint for a large amount of internal refactoring
229 and testing. It also contains a number of small bugfixes, so all users are
230 encouraged to upgrade.
233 This release is ABI and API compatible with Duclair (v1.0.0). Users
234 of older releases should refer to the Upgrading notes in this
235 document for that release.
238 VP8 optimizations for MIPS dspr2
239 vpxenc: add -quiet option
242 Encoder and decoder speed is consistent with the Eider release.
245 In general, quality is consistent with the Eider release.
247 Minor tweaks to ARNR filtering
248 Minor improvements to real time encoding with multiple temporal layers
251 Fixes multithreaded encoder race condition in loopfilter
252 Fixes multi-resolution threaded encoding
253 Fix potential encoder dead-lock after picture resize
256 2012-05-09 v1.1.0 "Eider"
257 This introduces a number of enhancements, mostly focused on real-time
258 encoding. In addition, it fixes a decoder bug (first introduced in
259 Duclair) so all users of that release are encouraged to upgrade.
262 This release is ABI and API compatible with Duclair (v1.0.0). Users
263 of older releases should refer to the Upgrading notes in this
264 document for that release.
266 This release introduces a new temporal denoiser, controlled by the
267 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
268 currently take a strength parameter, so the control is effectively
269 a boolean - zero (off) or non-zero (on). For compatibility with
270 existing applications, the values accepted are the same as those
271 for the spatial denoiser (0-6). The temporal denoiser is enabled
272 by default, and the older spatial denoiser may be restored by
273 configuring with --disable-temporal-denoising. The temporal denoiser
274 is more computationally intensive than the spatial one.
276 This release removes support for a legacy, decode only API that was
277 supported, but deprecated, at the initial release of libvpx
278 (v0.9.0). This is not expected to have any impact. If you are
279 impacted, you can apply a reversion to commit 2bf8fb58 locally.
280 Please update to the latest libvpx API if you are affected.
283 Adds a motion compensated temporal denoiser to the encoder, which
284 gives higher quality than the older spatial denoiser. (See above
285 for notes on upgrading).
287 In addition, support for new compilers and platforms were added,
289 improved support for XCode
290 Android x86 NDK build
294 Changing resolution with vpx_codec_enc_config_set() is now
295 supported. Previously, reinitializing the codec was required to
296 change the input resolution.
298 The vpxenc application has initial support for producing multiple
299 encodes from the same input in one call. Resizing is not yet
300 supported, but varying other codec parameters is. Use -- to
301 delineate output streams. Options persist from one stream to the
304 Also, the vpxenc application will now use a keyframe interval of
305 5 seconds by default. Use the --kf-max-dist option to override.
308 Decoder performance improved 2.5% versus Duclair. Encoder speed is
309 consistent with Duclair for most material. Two pass encoding of
310 slideshow-like material will see significant improvements.
312 Large realtime encoding speed gains at a small quality expense are
313 possible by configuring the on-the-fly bitpacking experiment with
314 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
315 faster (ARM) depending on the number of threads and bitrate
316 settings. This technique sees constant gain over the 5-16 speed
317 range. For VC style input the loss seen is up to 0.2dB. See commit
318 52cf4dca for further details.
321 On the whole, quality is consistent with the Duclair release. Some
324 Reduced blockiness in easy sections by applying a penalty to
327 Improved quality of static sections (like slideshows) with
330 Improved keyframe sizing with multiple temporal layers
333 Corrected alt-ref contribution to frame rate for visible updates
334 to the alt-ref buffer. This affected applications making manual
335 usage of the frame reference flags, or temporal layers.
337 Additional constraints were added to disable multi-frame quality
338 enhancement (MFQE) in sections of the frame where there is motion.
341 Fixed corruption issues when vpx_codec_enc_config_set() was called
342 with spatial resampling enabled.
344 Fixed a decoder error introduced in Duclair where the segmentation
345 map was not being reinitialized on keyframes (#378)
348 2012-01-27 v1.0.0 "Duclair"
349 Our fourth named release, focused on performance and features related to
350 real-time encoding. It also fixes a decoder crash bug introduced in
351 v0.9.7, so all users of that release are encouraged to upgrade.
354 This release is ABI incompatible with prior releases of libvpx, so the
355 "major" version number has been bumped to 1. You must recompile your
356 applications against the latest version of the libvpx headers. The
357 API remains compatible, and this should not require code changes in most
361 This release introduces several substantial new features to the encoder,
362 of particular interest to real time streaming applications.
364 Temporal scalability allows the encoder to produce a stream that can
365 be decimated to different frame rates, with independent rate targeting
368 Multiframe quality enhancement postprocessing can make visual quality
369 more consistent in the presence of frames that are substantially
370 different quality than the surrounding frames, as in the temporal
371 scalability case and in some forced keyframe scenarios.
373 Multiple-resolution encoding support allows the encoding of the
374 same content at different resolutions faster than encoding them
378 Optimization targets for this release included the decoder and the real-
379 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
380 this release. Encoder improvements followed a curve where speeds 1-3
381 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
382 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
386 Encoder quality in the single stream case is consistent with the Cayuga
390 This release fixes an OOB read decoder crash bug present in v0.9.7
391 related to the clamping of motion vectors in SPLITMV blocks. This
392 behavior could be triggered by corrupt input or by starting
393 decoding from a P-frame.
396 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
397 This is an incremental bugfix release against Cayuga. All users of that
398 release are strongly encouraged to upgrade.
400 - Fix potential OOB reads (cdae03a)
402 An unbounded out of bounds read was discovered when the
403 decoder was requested to perform error concealment (new in
404 Cayuga) given a frame with corrupt partition sizes.
406 A bounded out of bounds read was discovered affecting all
407 versions of libvpx. Given an multipartition input frame that
408 is truncated between the mode/mv partition and the first
409 residiual paritition (in the block of partition offsets), up
410 to 3 extra bytes could have been read from the source buffer.
411 The code will not take any action regardless of the contents
412 of these undefined bytes, as the truncated buffer is detected
413 immediately following the read based on the calculated
414 starting position of the coefficient partition.
416 - Fix potential error concealment crash when the very first frame
417 is missing or corrupt (a609be5)
419 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
421 - Revert 1-pass CBR rate control changes (e961317)
422 Further testing showed this change produced undesirable visual
423 artifacts, rolling back for now.
426 2011-08-02 v0.9.7 "Cayuga"
427 Our third named release, focused on a faster, higher quality, encoder.
430 This release is backwards compatible with Aylesbury (v0.9.5) and
431 Bali (v0.9.6). Users of older releases should refer to the Upgrading
432 notes in this document for that release.
435 Stereo 3D format support for vpxenc
436 Runtime detection of available processor cores.
437 Allow specifying --end-usage by enum name
438 vpxdec: test for frame corruption
439 vpxenc: add quantizer histogram display
440 vpxenc: add rate histogram display
441 Set VPX_FRAME_IS_DROPPABLE
442 update configure for ios sdk 4.3
443 Avoid text relocations in ARM vp8 decoder
444 Generate a vpx.pc file for pkg-config.
445 New ways of passing encoded data between encoder and decoder.
448 This release includes across-the-board speed improvements to the
449 encoder. On x86, these measure at approximately 11.5% in Best mode,
450 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
451 On ARM Cortex A9 with Neon extensions, real-time encoding of video
452 telephony content is 35% faster than Bali on single core and 48%
453 faster on multi-core. On the NVidia Tegra2 platform, real time
454 encoding is 40% faster than Bali.
456 Decoder speed was not a priority for this release, but improved
457 approximately 8.4% on x86.
459 Reduce motion vector search on alt-ref frame.
460 Encoder loopfilter running in its own thread
461 Reworked loopfilter to precalculate more parameters
462 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
463 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
464 Removed redundant checks
465 Reduced structure sizes
466 utilize preload in ARMv6 MC/LPF/Copy routines
467 ARM optimized quantization, dfct, variance, subtract
468 Increase chrow row alignment to 16 bytes.
469 disable trellis optimization for first pass
470 Write SSSE3 sub-pixel filter function
471 Improve SSE2 half-pixel filter funtions
472 Add vp8_sub_pixel_variance16x8_ssse3 function
473 Reduce unnecessary distortion computation
474 Use diamond search to replace full search
475 Preload reference area in sub-pixel motion search (real-time mode)
478 This release focused primarily on one-pass use cases, including
479 video conferencing. Low latency data rate control was significantly
480 improved, improving streamability over bandwidth constrained links.
481 Added support for error concealment, allowing frames to maintain
482 visual quality in the presence of substantial packet loss.
484 Add rc_max_intra_bitrate_pct control
485 Limit size of initial keyframe in one-pass.
486 Improve framerate adaptation
487 Improved 1-pass CBR rate control
488 Improved KF insertion after fades to still.
489 Improved key frame detection.
490 Improved activity masking (lower PSNR impact for same SSIM boost)
491 Improved interaction between GF and ARFs
492 Adding error-concealment to the decoder.
493 Adding support for independent partitions
494 Adjusted rate-distortion constants
498 Removed firstpass motion map
499 Fix parallel make install
500 Fix multithreaded encoding for 1 MB wide frame
501 Fixed iwalsh_neon build problems with RVDS4.1
502 Fix semaphore emulation, spin-wait intrinsics on Windows
503 Fix build with xcode4 and simplify GLOBAL.
504 Mark ARM asm objects as allowing a non-executable stack.
505 Fix vpxenc encoding incorrect webm file header on big endian
508 2011-03-07 v0.9.6 "Bali"
509 Our second named release, focused on a faster, higher quality, encoder.
512 This release is backwards compatible with Aylesbury (v0.9.5). Users
513 of older releases should refer to the Upgrading notes in this
514 document for that release.
517 vpxenc --psnr shows a summary when encode completes
518 --tune=ssim option to enable activity masking
519 improved postproc visualizations for development
520 updated support for Apple iOS to SDK 4.2
521 query decoder to determine which reference frames were updated
522 implemented error tracking in the decoder
523 fix pipe support on windows
526 Primary focus was on good quality mode, speed 0. Average improvement
527 on x86 about 40%, up to 100% on user-generated content at that speed.
528 Best quality mode speed improved 35%, and realtime speed 10-20%. This
529 release also saw significant improvement in realtime encoding speed
532 Improved encoder threading
533 Dont pick encoder filter level when loopfilter is disabled.
534 Avoid double copying of key frames into alt and golden buffer
536 x86 sse2 temporal filter
537 SSSE3 version of fast quantizer
538 vp8_rd_pick_best_mbsegmentation code restructure
539 Adjusted breakout RD for SPLITMV
540 Changed segmentation check order
541 Improved rd_pick_intra4x4block
542 Adds armv6 optimized variance calculation
543 ARMv6 optimized sad16x16
544 ARMv6 optimized half pixel variance calculations
545 Full search SAD function optimization in SSE4.1
546 Improve MV prediction accuracy to achieve performance gain
547 Improve MV prediction in vp8_pick_inter_mode() for speed>3
550 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
551 also includes support for "activity masking," which greatly improves
552 SSIM at the expense of PSNR. For now, this feature is available with
553 the --tune=ssim option. Further experimentation in this area
554 is ongoing. This release also introduces a new rate control mode
555 called "CQ," which changes the allocation of bits within a clip to
556 the sections where they will have the most visual impact.
558 Tuning for the more exact quantizer.
559 Relax rate control for last few frames
561 Limit key frame quantizer for forced key frames.
563 Add simple version of activity masking.
564 make rdmult adaptive for intra in quantizer RDO
565 cap the best quantizer for 2nd order DC
566 change the threshold of DC check for encode breakout
569 Fix crash on Sparc Solaris.
570 Fix counter of fixed keyframe distance
571 ARNR filter pointer update bug fix
572 Fixed use of motion percentage in KF/GF group calc
573 Changed condition for using RD in Intra Mode
574 Fix encoder real-time only configuration.
575 Fix ARM encoder crash with multiple token partitions
576 Fixed bug first cluster timecode of webm file is wrong.
577 Fixed various encoder bugs with odd-sized images
578 vp8e_get_preview fixed when spatial resampling enabled
579 quantizer: fix assertion in fast quantizer path
580 Allocate source buffers to be multiples of 16
581 Fix for manual Golden frame frequency
582 Fix drastic undershoot in long form content
585 2010-10-28 v0.9.5 "Aylesbury"
586 Our first named release, focused on a faster decoder, and a better encoder.
589 This release incorporates backwards-incompatible changes to the
590 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
593 * the -q (quiet) option has been removed, and replaced with
594 -v (verbose). the output is quiet by default. Use -v to see
595 the version number of the binary.
597 * The default behavior is now to write output to a single file
598 instead of individual frames. The -y option has been removed.
599 Y4M output is the default.
601 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
602 options must be specified.
604 $ ivfdec -o OUTPUT INPUT
605 $ vpxdec --i420 -o OUTPUT INPUT
607 * If an output file is not specified, the default is to write
608 Y4M to stdout. This makes piping more natural.
610 $ ivfdec -y -o - INPUT | ...
613 * The output file has additional flexibility for formatting the
614 filename. It supports escape characters for constructing a
615 filename from the width, height, and sequence number. This
616 replaces the -p option. To get the equivalent:
618 $ ivfdec -p frame INPUT
619 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
622 * The output file must be specified with -o, rather than as the
625 $ ivfenc <options> INPUT OUTPUT
626 $ vpxenc <options> -o OUTPUT INPUT
628 * The output defaults to webm. To get IVF output, use the --ivf
631 $ ivfenc <options> INPUT OUTPUT.ivf
632 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
636 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
637 vpxdec supports .webm input
638 vpxdec writes .y4m by default
639 vpxenc writes .webm output by default
640 vpxenc --psnr now shows the average/overall PSNR at the end
641 ARM platforms now support runtime cpu detection
642 vpxdec visualizations added for motion vectors, block modes, references
643 vpxdec now silent by default
644 vpxdec --progress shows frame-by-frame timing information
645 vpxenc supports the distinction between --fps and --timebase
646 NASM is now a supported assembler
647 configure: enable PIC for shared libs by default
648 configure: add --enable-small
649 configure: support for ppc32-linux-gcc
650 configure: support for sparc-solaris-gcc
653 Improve handling of invalid frames
654 Fix valgrind errors in the NEON loop filters.
655 Fix loopfilter delta zero transitions
656 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
657 Build fixes for darwin-icc
660 20-40% (average 28%) improvement in libvpx decoder speed,
662 Rewrite vp8_short_walsh4x4_sse2()
663 Optimizations on the loopfilters.
664 Miscellaneous improvements for Atom
665 Add 4-tap version of 2nd-pass ARMv6 MC filter.
666 Improved multithread utilization
667 Better instruction choices on x86
668 reorder data to use wider instructions
669 Update NEON wide idcts
670 Make block access to frame buffer sequential
671 Improved subset block search
672 Bilinear subpixel optimizations for ssse3.
673 Decrease memory footprint
675 Encoder speed improvements (percentage gain not measured):
676 Skip unnecessary search of identical frames
677 Add SSE2 subtract functions
678 Improve bounds checking in vp8_diamond_search_sadx4()
679 Added vp8_fast_quantize_b_sse2
682 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
683 encoding mode, and up to 60% improvement on very noisy, still
684 or slow moving source video
686 Motion compensated temporal filter for Alt-Ref Noise Reduction
687 Improved use of trellis quantization on 2nd order Y blocks
688 Tune effect of motion on KF/GF boost in two pass
689 Allow coefficient optimization for good quality speed 0.
690 Improved control of active min quantizer for two pass.
691 Enable ARFs for non-lagged compress
695 Disable frame dropping by default
696 Improved multithreaded performance
697 Improved Force Key Frame Behaviour
698 Increased rate control buffer level precision
699 Fix bug in 1st pass motion compensation
700 ivfenc: correct fixed kf interval, --disable-kf
702 Changed above and left context data layout
703 Rework idct calling structure.
704 Removed unnecessary MB_MODE_INFO copies
705 x86: SSSE3 sixtap prediction
706 Reworked IDCT to include reconstruction (add) step
707 Swap alt/gold/new/last frame buffer ptrs instead of copying.
708 Improve SSE2 loopfilter functions
709 Change bitreader to use a larger window.
710 Avoid loopfilter reinitialization when possible
712 Normalize quantizer's zero bin and rounding factors
713 Add trellis quantization.
714 Make the quantizer exact.
715 Updates to ARNR filtering algorithm
716 Fix breakout thresh computation for golden & AltRef frames
717 Redo the forward 4x4 dct
718 Improve the accuracy of forward walsh-hadamard transform
719 Further adjustment of RD behaviour with Q and Zbin.
721 Allow linking of libs built with MinGW to MSVC
722 Fix target auto-detection on mingw32
723 Allow --cpu= to work for x86.
724 configure: pass original arguments through to make dist
725 Fix builds without runtime CPU detection
726 msvs: fix install of codec sources
727 msvs: Change devenv.com command line for better msys support
728 msvs: Add vs9 targets.
729 Add x86_64-linux-icc target
731 Potential crashes on older MinGW builds
732 Fix two-pass framrate for Y4M input.
733 Fixed simple loop filter, other crashes on ARM v6
734 arm: fix missing dependency with --enable-shared
735 configure: support directories containing .o
736 Replace pinsrw (SSE) with MMX instructions
737 apple: include proper mach primatives
738 Fixed rate control bug with long key frame interval.
739 Fix DSO link errors on x86-64 when not using a version script
740 Fixed buffer selection for UV in AltRef filtering
745 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
746 * Speed optimizations
749 * Prevent out-of-bounds accesses on invalid data
750 - Build system updates:
751 * Detect toolchain to be used automatically for native builds
752 * Support building shared libraries
753 * Better autotools emulation (--prefix, --libdir, DESTDIR)
755 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
759 - Initial open source release. Welcome to WebM and VP8!