| blob | 7510dc660b7f91eab4f5143a63a3f532c592e0e1 |
1 Next Release
2 - Incompatible changes:
3 The AV1 encoder's default keyframe interval changed to 128 from 9999.
4 Support for armv6 was removed.
6 2016-04-07 v0.1.0 "AOMedia Codec 1"
7 This release is the first Alliance for Open Media codec.
8 2015-11-09 v1.5.0 "Javan Whistling Duck"
9 This release improves upon the VP9 encoder and speeds up the encoding and
10 decoding processes.
12 - Upgrading:
13 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
14 controls and adds a variety of VP9 controls for testing.
16 The vpxenc utility now prefers VP9 by default.
18 - Enhancements:
19 Faster VP9 encoding and decoding
20 Smaller library size by combining functions used by VP8 and VP9
22 - Bug Fixes:
23 A variety of fuzzing issues
25 2015-04-03 v1.4.0 "Indian Runner Duck"
26 This release includes significant improvements to the VP9 codec.
28 - Upgrading:
29 This release is ABI incompatible with 1.3.0. It drops the compatibility
30 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
31 controls for VP9.
33 - Enhancements:
34 Faster VP9 encoding and decoding
35 Multithreaded VP9 decoding (tile and frame-based)
36 Multithreaded VP9 encoding - on by default
37 YUV 4:2:2 and 4:4:4 support in VP9
38 10 and 12bit support in VP9
39 64bit ARM support by replacing ARM assembly with intrinsics
41 - Bug Fixes:
42 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
43 files.
45 - Known Issues:
46 Frame Parallel decoding fails for segmented and non-420 files.
48 2013-11-15 v1.3.0 "Forest"
49 This release introduces the VP9 codec in a backward-compatible way.
50 All existing users of VP8 can continue to use the library without
51 modification. However, some VP8 options do not map to VP9 in the same manner.
53 The VP9 encoder in this release is not feature complete. Users interested in
54 the encoder are advised to use the git master branch and discuss issues on
55 libvpx mailing lists.
57 - Upgrading:
58 This release is ABI and API compatible with Duclair (v1.0.0). Users
59 of older releases should refer to the Upgrading notes in this document
60 for that release.
62 - Enhancements:
63 Get rid of bashisms in the main build scripts
64 Added usage info on command line options
65 Add lossless compression mode
66 Dll build of libvpx
67 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
68 Add option to disable documentation
69 configure: add --enable-external-build support
70 make: support V=1 as short form of verbose=yes
71 configure: support mingw-w64
72 configure: support hardfloat armv7 CHOSTS
73 configure: add support for android x86
74 Add estimated completion time to vpxenc
75 Don't exit on decode errors in vpxenc
76 vpxenc: support scaling prior to encoding
77 vpxdec: support scaling output
78 vpxenc: improve progress indicators with --skip
79 msvs: Don't link to winmm.lib
80 Add a new script for producing vcxproj files
81 Produce Visual Studio 10 and 11 project files
82 Produce Windows Phone project files
83 msvs-build: use msbuild for vs >= 2005
84 configure: default configure log to config.log
85 Add encoding option --static-thresh
87 - Speed:
88 Miscellaneous speed optimizations for VP8 and VP9.
90 - Quality:
91 In general, quality is consistent with the Eider release.
93 - Bug Fixes:
94 This release represents approximately a year of engineering effort,
95 and contains multiple bug fixes. Please refer to git history for details.
98 2012-12-21 v1.2.0
99 This release acts as a checkpoint for a large amount of internal refactoring
100 and testing. It also contains a number of small bugfixes, so all users are
101 encouraged to upgrade.
103 - Upgrading:
104 This release is ABI and API compatible with Duclair (v1.0.0). Users
105 of older releases should refer to the Upgrading notes in this
106 document for that release.
108 - Enhancements:
109 VP8 optimizations for MIPS dspr2
110 vpxenc: add -quiet option
112 - Speed:
113 Encoder and decoder speed is consistent with the Eider release.
115 - Quality:
116 In general, quality is consistent with the Eider release.
118 Minor tweaks to ARNR filtering
119 Minor improvements to real time encoding with multiple temporal layers
121 - Bug Fixes:
122 Fixes multithreaded encoder race condition in loopfilter
123 Fixes multi-resolution threaded encoding
124 Fix potential encoder dead-lock after picture resize
127 2012-05-09 v1.1.0 "Eider"
128 This introduces a number of enhancements, mostly focused on real-time
129 encoding. In addition, it fixes a decoder bug (first introduced in
130 Duclair) so all users of that release are encouraged to upgrade.
132 - Upgrading:
133 This release is ABI and API compatible with Duclair (v1.0.0). Users
134 of older releases should refer to the Upgrading notes in this
135 document for that release.
137 This release introduces a new temporal denoiser, controlled by the
138 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
139 currently take a strength parameter, so the control is effectively
140 a boolean - zero (off) or non-zero (on). For compatibility with
141 existing applications, the values accepted are the same as those
142 for the spatial denoiser (0-6). The temporal denoiser is enabled
143 by default, and the older spatial denoiser may be restored by
144 configuring with --disable-temporal-denoising. The temporal denoiser
145 is more computationally intensive than the spatial one.
147 This release removes support for a legacy, decode only API that was
148 supported, but deprecated, at the initial release of libvpx
149 (v0.9.0). This is not expected to have any impact. If you are
150 impacted, you can apply a reversion to commit 2bf8fb58 locally.
151 Please update to the latest libvpx API if you are affected.
153 - Enhancements:
154 Adds a motion compensated temporal denoiser to the encoder, which
155 gives higher quality than the older spatial denoiser. (See above
156 for notes on upgrading).
158 In addition, support for new compilers and platforms were added,
159 including:
160 improved support for XCode
161 Android x86 NDK build
162 OS/2 support
163 SunCC support
165 Changing resolution with vpx_codec_enc_config_set() is now
166 supported. Previously, reinitializing the codec was required to
167 change the input resolution.
169 The vpxenc application has initial support for producing multiple
170 encodes from the same input in one call. Resizing is not yet
171 supported, but varying other codec parameters is. Use -- to
172 delineate output streams. Options persist from one stream to the
173 next.
175 Also, the vpxenc application will now use a keyframe interval of
176 5 seconds by default. Use the --kf-max-dist option to override.
178 - Speed:
179 Decoder performance improved 2.5% versus Duclair. Encoder speed is
180 consistent with Duclair for most material. Two pass encoding of
181 slideshow-like material will see significant improvements.
183 Large realtime encoding speed gains at a small quality expense are
184 possible by configuring the on-the-fly bitpacking experiment with
185 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
186 faster (ARM) depending on the number of threads and bitrate
187 settings. This technique sees constant gain over the 5-16 speed
188 range. For VC style input the loss seen is up to 0.2dB. See commit
189 52cf4dca for further details.
191 - Quality:
192 On the whole, quality is consistent with the Duclair release. Some
193 tweaks:
195 Reduced blockiness in easy sections by applying a penalty to
196 intra modes.
198 Improved quality of static sections (like slideshows) with
199 two pass encoding.
201 Improved keyframe sizing with multiple temporal layers
203 - Bug Fixes:
204 Corrected alt-ref contribution to frame rate for visible updates
205 to the alt-ref buffer. This affected applications making manual
206 usage of the frame reference flags, or temporal layers.
208 Additional constraints were added to disable multi-frame quality
209 enhancement (MFQE) in sections of the frame where there is motion.
210 (#392)
212 Fixed corruption issues when vpx_codec_enc_config_set() was called
213 with spatial resampling enabled.
215 Fixed a decoder error introduced in Duclair where the segmentation
216 map was not being reinitialized on keyframes (#378)
219 2012-01-27 v1.0.0 "Duclair"
220 Our fourth named release, focused on performance and features related to
221 real-time encoding. It also fixes a decoder crash bug introduced in
222 v0.9.7, so all users of that release are encouraged to upgrade.
224 - Upgrading:
225 This release is ABI incompatible with prior releases of libvpx, so the
226 "major" version number has been bumped to 1. You must recompile your
227 applications against the latest version of the libvpx headers. The
228 API remains compatible, and this should not require code changes in most
229 applications.
231 - Enhancements:
232 This release introduces several substantial new features to the encoder,
233 of particular interest to real time streaming applications.
235 Temporal scalability allows the encoder to produce a stream that can
236 be decimated to different frame rates, with independent rate targetting
237 for each substream.
239 Multiframe quality enhancement postprocessing can make visual quality
240 more consistent in the presence of frames that are substantially
241 different quality than the surrounding frames, as in the temporal
242 scalability case and in some forced keyframe scenarios.
244 Multiple-resolution encoding support allows the encoding of the
245 same content at different resolutions faster than encoding them
246 separately.
248 - Speed:
249 Optimization targets for this release included the decoder and the real-
250 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
251 this release. Encoder improvements followed a curve where speeds 1-3
252 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
253 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
254 Cayuga release.
256 - Quality:
257 Encoder quality in the single stream case is consistent with the Cayuga
258 release.
260 - Bug Fixes:
261 This release fixes an OOB read decoder crash bug present in v0.9.7
262 related to the clamping of motion vectors in SPLITMV blocks. This
263 behavior could be triggered by corrupt input or by starting
264 decoding from a P-frame.
267 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
268 This is an incremental bugfix release against Cayuga. All users of that
269 release are strongly encouraged to upgrade.
271 - Fix potential OOB reads (cdae03a)
273 An unbounded out of bounds read was discovered when the
274 decoder was requested to perform error concealment (new in
275 Cayuga) given a frame with corrupt partition sizes.
277 A bounded out of bounds read was discovered affecting all
278 versions of libvpx. Given an multipartition input frame that
279 is truncated between the mode/mv partition and the first
280 residiual paritition (in the block of partition offsets), up
281 to 3 extra bytes could have been read from the source buffer.
282 The code will not take any action regardless of the contents
283 of these undefined bytes, as the truncated buffer is detected
284 immediately following the read based on the calculated
285 starting position of the coefficient partition.
287 - Fix potential error concealment crash when the very first frame
288 is missing or corrupt (a609be5)
290 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
292 - Revert 1-pass CBR rate control changes (e961317)
293 Further testing showed this change produced undesirable visual
294 artifacts, rolling back for now.
297 2011-08-02 v0.9.7 "Cayuga"
298 Our third named release, focused on a faster, higher quality, encoder.
300 - Upgrading:
301 This release is backwards compatible with Aylesbury (v0.9.5) and
302 Bali (v0.9.6). Users of older releases should refer to the Upgrading
303 notes in this document for that release.
305 - Enhancements:
306 Stereo 3D format support for vpxenc
307 Runtime detection of available processor cores.
308 Allow specifying --end-usage by enum name
309 vpxdec: test for frame corruption
310 vpxenc: add quantizer histogram display
311 vpxenc: add rate histogram display
312 Set VPX_FRAME_IS_DROPPABLE
313 update configure for ios sdk 4.3
314 Avoid text relocations in ARM vp8 decoder
315 Generate a vpx.pc file for pkg-config.
316 New ways of passing encoded data between encoder and decoder.
318 - Speed:
319 This release includes across-the-board speed improvements to the
320 encoder. On x86, these measure at approximately 11.5% in Best mode,
321 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
322 On ARM Cortex A9 with Neon extensions, real-time encoding of video
323 telephony content is 35% faster than Bali on single core and 48%
324 faster on multi-core. On the NVidia Tegra2 platform, real time
325 encoding is 40% faster than Bali.
327 Decoder speed was not a priority for this release, but improved
328 approximately 8.4% on x86.
330 Reduce motion vector search on alt-ref frame.
331 Encoder loopfilter running in its own thread
332 Reworked loopfilter to precalculate more parameters
333 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
334 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
335 Removed redundant checks
336 Reduced structure sizes
337 utilize preload in ARMv6 MC/LPF/Copy routines
338 ARM optimized quantization, dfct, variance, subtract
339 Increase chrow row alignment to 16 bytes.
340 disable trellis optimization for first pass
341 Write SSSE3 sub-pixel filter function
342 Improve SSE2 half-pixel filter funtions
343 Add vp8_sub_pixel_variance16x8_ssse3 function
344 Reduce unnecessary distortion computation
345 Use diamond search to replace full search
346 Preload reference area in sub-pixel motion search (real-time mode)
348 - Quality:
349 This release focused primarily on one-pass use cases, including
350 video conferencing. Low latency data rate control was significantly
351 improved, improving streamability over bandwidth constrained links.
352 Added support for error concealment, allowing frames to maintain
353 visual quality in the presence of substantial packet loss.
355 Add rc_max_intra_bitrate_pct control
356 Limit size of initial keyframe in one-pass.
357 Improve framerate adaptation
358 Improved 1-pass CBR rate control
359 Improved KF insertion after fades to still.
360 Improved key frame detection.
361 Improved activity masking (lower PSNR impact for same SSIM boost)
362 Improved interaction between GF and ARFs
363 Adding error-concealment to the decoder.
364 Adding support for independent partitions
365 Adjusted rate-distortion constants
368 - Bug Fixes:
369 Removed firstpass motion map
370 Fix parallel make install
371 Fix multithreaded encoding for 1 MB wide frame
372 Fixed iwalsh_neon build problems with RVDS4.1
373 Fix semaphore emulation, spin-wait intrinsics on Windows
374 Fix build with xcode4 and simplify GLOBAL.
375 Mark ARM asm objects as allowing a non-executable stack.
376 Fix vpxenc encoding incorrect webm file header on big endian
379 2011-03-07 v0.9.6 "Bali"
380 Our second named release, focused on a faster, higher quality, encoder.
382 - Upgrading:
383 This release is backwards compatible with Aylesbury (v0.9.5). Users
384 of older releases should refer to the Upgrading notes in this
385 document for that release.
387 - Enhancements:
388 vpxenc --psnr shows a summary when encode completes
389 --tune=ssim option to enable activity masking
390 improved postproc visualizations for development
391 updated support for Apple iOS to SDK 4.2
392 query decoder to determine which reference frames were updated
393 implemented error tracking in the decoder
394 fix pipe support on windows
396 - Speed:
397 Primary focus was on good quality mode, speed 0. Average improvement
398 on x86 about 40%, up to 100% on user-generated content at that speed.
399 Best quality mode speed improved 35%, and realtime speed 10-20%. This
400 release also saw significant improvement in realtime encoding speed
401 on ARM platforms.
403 Improved encoder threading
404 Dont pick encoder filter level when loopfilter is disabled.
405 Avoid double copying of key frames into alt and golden buffer
406 FDCT optimizations.
407 x86 sse2 temporal filter
408 SSSE3 version of fast quantizer
409 vp8_rd_pick_best_mbsegmentation code restructure
410 Adjusted breakout RD for SPLITMV
411 Changed segmentation check order
412 Improved rd_pick_intra4x4block
413 Adds armv6 optimized variance calculation
414 ARMv6 optimized sad16x16
415 ARMv6 optimized half pixel variance calculations
416 Full search SAD function optimization in SSE4.1
417 Improve MV prediction accuracy to achieve performance gain
418 Improve MV prediction in vp8_pick_inter_mode() for speed>3
420 - Quality:
421 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
422 also includes support for "activity masking," which greatly improves
423 SSIM at the expense of PSNR. For now, this feature is available with
424 the --tune=ssim option. Further experimentation in this area
425 is ongoing. This release also introduces a new rate control mode
426 called "CQ," which changes the allocation of bits within a clip to
427 the sections where they will have the most visual impact.
429 Tuning for the more exact quantizer.
430 Relax rate control for last few frames
431 CQ Mode
432 Limit key frame quantizer for forced key frames.
433 KF/GF Pulsing
434 Add simple version of activity masking.
435 make rdmult adaptive for intra in quantizer RDO
436 cap the best quantizer for 2nd order DC
437 change the threshold of DC check for encode breakout
439 - Bug Fixes:
440 Fix crash on Sparc Solaris.
441 Fix counter of fixed keyframe distance
442 ARNR filter pointer update bug fix
443 Fixed use of motion percentage in KF/GF group calc
444 Changed condition for using RD in Intra Mode
445 Fix encoder real-time only configuration.
446 Fix ARM encoder crash with multiple token partitions
447 Fixed bug first cluster timecode of webm file is wrong.
448 Fixed various encoder bugs with odd-sized images
449 vp8e_get_preview fixed when spatial resampling enabled
450 quantizer: fix assertion in fast quantizer path
451 Allocate source buffers to be multiples of 16
452 Fix for manual Golden frame frequency
453 Fix drastic undershoot in long form content
456 2010-10-28 v0.9.5 "Aylesbury"
457 Our first named release, focused on a faster decoder, and a better encoder.
459 - Upgrading:
460 This release incorporates backwards-incompatible changes to the
461 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
463 vpxdec
464 * the -q (quiet) option has been removed, and replaced with
465 -v (verbose). the output is quiet by default. Use -v to see
466 the version number of the binary.
468 * The default behavior is now to write output to a single file
469 instead of individual frames. The -y option has been removed.
470 Y4M output is the default.
472 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
473 options must be specified.
475 $ ivfdec -o OUTPUT INPUT
476 $ vpxdec --i420 -o OUTPUT INPUT
478 * If an output file is not specified, the default is to write
479 Y4M to stdout. This makes piping more natural.
481 $ ivfdec -y -o - INPUT | ...
482 $ vpxdec INPUT | ...
484 * The output file has additional flexibility for formatting the
485 filename. It supports escape characters for constructing a
486 filename from the width, height, and sequence number. This
487 replaces the -p option. To get the equivalent:
489 $ ivfdec -p frame INPUT
490 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
492 vpxenc
493 * The output file must be specified with -o, rather than as the
494 last argument.
496 $ ivfenc <options> INPUT OUTPUT
497 $ vpxenc <options> -o OUTPUT INPUT
499 * The output defaults to webm. To get IVF output, use the --ivf
500 option.
502 $ ivfenc <options> INPUT OUTPUT.ivf
503 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
506 - Enhancements:
507 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
508 vpxdec supports .webm input
509 vpxdec writes .y4m by default
510 vpxenc writes .webm output by default
511 vpxenc --psnr now shows the average/overall PSNR at the end
512 ARM platforms now support runtime cpu detection
513 vpxdec visualizations added for motion vectors, block modes, references
514 vpxdec now silent by default
515 vpxdec --progress shows frame-by-frame timing information
516 vpxenc supports the distinction between --fps and --timebase
517 NASM is now a supported assembler
518 configure: enable PIC for shared libs by default
519 configure: add --enable-small
520 configure: support for ppc32-linux-gcc
521 configure: support for sparc-solaris-gcc
523 - Bugs:
524 Improve handling of invalid frames
525 Fix valgrind errors in the NEON loop filters.
526 Fix loopfilter delta zero transitions
527 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
528 Build fixes for darwin-icc
530 - Speed:
531 20-40% (average 28%) improvement in libvpx decoder speed,
532 including:
533 Rewrite vp8_short_walsh4x4_sse2()
534 Optimizations on the loopfilters.
535 Miscellaneous improvements for Atom
536 Add 4-tap version of 2nd-pass ARMv6 MC filter.
537 Improved multithread utilization
538 Better instruction choices on x86
539 reorder data to use wider instructions
540 Update NEON wide idcts
541 Make block access to frame buffer sequential
542 Improved subset block search
543 Bilinear subpixel optimizations for ssse3.
544 Decrease memory footprint
546 Encoder speed improvements (percentage gain not measured):
547 Skip unnecessary search of identical frames
548 Add SSE2 subtract functions
549 Improve bounds checking in vp8_diamond_search_sadx4()
550 Added vp8_fast_quantize_b_sse2
552 - Quality:
553 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
554 encoding mode, and up to 60% improvement on very noisy, still
555 or slow moving source video
557 Motion compensated temporal filter for Alt-Ref Noise Reduction
558 Improved use of trellis quantization on 2nd order Y blocks
559 Tune effect of motion on KF/GF boost in two pass
560 Allow coefficient optimization for good quality speed 0.
561 Improved control of active min quantizer for two pass.
562 Enable ARFs for non-lagged compress
564 2010-09-02 v0.9.2
565 - Enhancements:
566 Disable frame dropping by default
567 Improved multithreaded performance
568 Improved Force Key Frame Behaviour
569 Increased rate control buffer level precision
570 Fix bug in 1st pass motion compensation
571 ivfenc: correct fixed kf interval, --disable-kf
572 - Speed:
573 Changed above and left context data layout
574 Rework idct calling structure.
575 Removed unnecessary MB_MODE_INFO copies
576 x86: SSSE3 sixtap prediction
577 Reworked IDCT to include reconstruction (add) step
578 Swap alt/gold/new/last frame buffer ptrs instead of copying.
579 Improve SSE2 loopfilter functions
580 Change bitreader to use a larger window.
581 Avoid loopfilter reinitialization when possible
582 - Quality:
583 Normalize quantizer's zero bin and rounding factors
584 Add trellis quantization.
585 Make the quantizer exact.
586 Updates to ARNR filtering algorithm
587 Fix breakout thresh computation for golden & AltRef frames
588 Redo the forward 4x4 dct
589 Improve the accuracy of forward walsh-hadamard transform
590 Further adjustment of RD behaviour with Q and Zbin.
591 - Build System:
592 Allow linking of libs built with MinGW to MSVC
593 Fix target auto-detection on mingw32
594 Allow --cpu= to work for x86.
595 configure: pass original arguments through to make dist
596 Fix builds without runtime CPU detection
597 msvs: fix install of codec sources
598 msvs: Change devenv.com command line for better msys support
599 msvs: Add vs9 targets.
600 Add x86_64-linux-icc target
601 - Bugs:
602 Potential crashes on older MinGW builds
603 Fix two-pass framrate for Y4M input.
604 Fixed simple loop filter, other crashes on ARM v6
605 arm: fix missing dependency with --enable-shared
606 configure: support directories containing .o
607 Replace pinsrw (SSE) with MMX instructions
608 apple: include proper mach primatives
609 Fixed rate control bug with long key frame interval.
610 Fix DSO link errors on x86-64 when not using a version script
611 Fixed buffer selection for UV in AltRef filtering
614 2010-06-17 v0.9.1
615 - Enhancements:
616 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
617 * Speed optimizations
618 - Bugfixes:
619 * Rate control
620 * Prevent out-of-bounds accesses on invalid data
621 - Build system updates:
622 * Detect toolchain to be used automatically for native builds
623 * Support building shared libraries
624 * Better autotools emulation (--prefix, --libdir, DESTDIR)
625 - Updated LICENSE
626 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
629 2010-05-18 v0.9.0
630 - Initial open source release. Welcome to WebM and VP8!