| blob | dad0ea1d29c59ee9706ccf6c5b2ba0ac5b549e71 |
1 Next Release
2 - Incompatible changes:
3 The AV1 encoder's default keyframe interval changed to 128 from 9999.
5 2016-04-07 v0.1.0 "AOMedia Codec 1"
6 This release is the first Alliance for Open Media codec.
7 2015-11-09 v1.5.0 "Javan Whistling Duck"
8 This release improves upon the VP9 encoder and speeds up the encoding and
9 decoding processes.
11 - Upgrading:
12 This release is ABI incompatible with 1.4.0. It drops deprecated VP8
13 controls and adds a variety of VP9 controls for testing.
15 The vpxenc utility now prefers VP9 by default.
17 - Enhancements:
18 Faster VP9 encoding and decoding
19 Smaller library size by combining functions used by VP8 and VP9
21 - Bug Fixes:
22 A variety of fuzzing issues
24 2015-04-03 v1.4.0 "Indian Runner Duck"
25 This release includes significant improvements to the VP9 codec.
27 - Upgrading:
28 This release is ABI incompatible with 1.3.0. It drops the compatibility
29 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
30 controls for VP9.
32 - Enhancements:
33 Faster VP9 encoding and decoding
34 Multithreaded VP9 decoding (tile and frame-based)
35 Multithreaded VP9 encoding - on by default
36 YUV 4:2:2 and 4:4:4 support in VP9
37 10 and 12bit support in VP9
38 64bit ARM support by replacing ARM assembly with intrinsics
40 - Bug Fixes:
41 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
42 files.
44 - Known Issues:
45 Frame Parallel decoding fails for segmented and non-420 files.
47 2013-11-15 v1.3.0 "Forest"
48 This release introduces the VP9 codec in a backward-compatible way.
49 All existing users of VP8 can continue to use the library without
50 modification. However, some VP8 options do not map to VP9 in the same manner.
52 The VP9 encoder in this release is not feature complete. Users interested in
53 the encoder are advised to use the git master branch and discuss issues on
54 libvpx mailing lists.
56 - Upgrading:
57 This release is ABI and API compatible with Duclair (v1.0.0). Users
58 of older releases should refer to the Upgrading notes in this document
59 for that release.
61 - Enhancements:
62 Get rid of bashisms in the main build scripts
63 Added usage info on command line options
64 Add lossless compression mode
65 Dll build of libvpx
66 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
67 Add option to disable documentation
68 configure: add --enable-external-build support
69 make: support V=1 as short form of verbose=yes
70 configure: support mingw-w64
71 configure: support hardfloat armv7 CHOSTS
72 configure: add support for android x86
73 Add estimated completion time to vpxenc
74 Don't exit on decode errors in vpxenc
75 vpxenc: support scaling prior to encoding
76 vpxdec: support scaling output
77 vpxenc: improve progress indicators with --skip
78 msvs: Don't link to winmm.lib
79 Add a new script for producing vcxproj files
80 Produce Visual Studio 10 and 11 project files
81 Produce Windows Phone project files
82 msvs-build: use msbuild for vs >= 2005
83 configure: default configure log to config.log
84 Add encoding option --static-thresh
86 - Speed:
87 Miscellaneous speed optimizations for VP8 and VP9.
89 - Quality:
90 In general, quality is consistent with the Eider release.
92 - Bug Fixes:
93 This release represents approximately a year of engineering effort,
94 and contains multiple bug fixes. Please refer to git history for details.
97 2012-12-21 v1.2.0
98 This release acts as a checkpoint for a large amount of internal refactoring
99 and testing. It also contains a number of small bugfixes, so all users are
100 encouraged to upgrade.
102 - Upgrading:
103 This release is ABI and API compatible with Duclair (v1.0.0). Users
104 of older releases should refer to the Upgrading notes in this
105 document for that release.
107 - Enhancements:
108 VP8 optimizations for MIPS dspr2
109 vpxenc: add -quiet option
111 - Speed:
112 Encoder and decoder speed is consistent with the Eider release.
114 - Quality:
115 In general, quality is consistent with the Eider release.
117 Minor tweaks to ARNR filtering
118 Minor improvements to real time encoding with multiple temporal layers
120 - Bug Fixes:
121 Fixes multithreaded encoder race condition in loopfilter
122 Fixes multi-resolution threaded encoding
123 Fix potential encoder dead-lock after picture resize
126 2012-05-09 v1.1.0 "Eider"
127 This introduces a number of enhancements, mostly focused on real-time
128 encoding. In addition, it fixes a decoder bug (first introduced in
129 Duclair) so all users of that release are encouraged to upgrade.
131 - Upgrading:
132 This release is ABI and API compatible with Duclair (v1.0.0). Users
133 of older releases should refer to the Upgrading notes in this
134 document for that release.
136 This release introduces a new temporal denoiser, controlled by the
137 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
138 currently take a strength parameter, so the control is effectively
139 a boolean - zero (off) or non-zero (on). For compatibility with
140 existing applications, the values accepted are the same as those
141 for the spatial denoiser (0-6). The temporal denoiser is enabled
142 by default, and the older spatial denoiser may be restored by
143 configuring with --disable-temporal-denoising. The temporal denoiser
144 is more computationally intensive than the spatial one.
146 This release removes support for a legacy, decode only API that was
147 supported, but deprecated, at the initial release of libvpx
148 (v0.9.0). This is not expected to have any impact. If you are
149 impacted, you can apply a reversion to commit 2bf8fb58 locally.
150 Please update to the latest libvpx API if you are affected.
152 - Enhancements:
153 Adds a motion compensated temporal denoiser to the encoder, which
154 gives higher quality than the older spatial denoiser. (See above
155 for notes on upgrading).
157 In addition, support for new compilers and platforms were added,
158 including:
159 improved support for XCode
160 Android x86 NDK build
161 OS/2 support
162 SunCC support
164 Changing resolution with vpx_codec_enc_config_set() is now
165 supported. Previously, reinitializing the codec was required to
166 change the input resolution.
168 The vpxenc application has initial support for producing multiple
169 encodes from the same input in one call. Resizing is not yet
170 supported, but varying other codec parameters is. Use -- to
171 delineate output streams. Options persist from one stream to the
172 next.
174 Also, the vpxenc application will now use a keyframe interval of
175 5 seconds by default. Use the --kf-max-dist option to override.
177 - Speed:
178 Decoder performance improved 2.5% versus Duclair. Encoder speed is
179 consistent with Duclair for most material. Two pass encoding of
180 slideshow-like material will see significant improvements.
182 Large realtime encoding speed gains at a small quality expense are
183 possible by configuring the on-the-fly bitpacking experiment with
184 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
185 faster (ARM) depending on the number of threads and bitrate
186 settings. This technique sees constant gain over the 5-16 speed
187 range. For VC style input the loss seen is up to 0.2dB. See commit
188 52cf4dca for further details.
190 - Quality:
191 On the whole, quality is consistent with the Duclair release. Some
192 tweaks:
194 Reduced blockiness in easy sections by applying a penalty to
195 intra modes.
197 Improved quality of static sections (like slideshows) with
198 two pass encoding.
200 Improved keyframe sizing with multiple temporal layers
202 - Bug Fixes:
203 Corrected alt-ref contribution to frame rate for visible updates
204 to the alt-ref buffer. This affected applications making manual
205 usage of the frame reference flags, or temporal layers.
207 Additional constraints were added to disable multi-frame quality
208 enhancement (MFQE) in sections of the frame where there is motion.
209 (#392)
211 Fixed corruption issues when vpx_codec_enc_config_set() was called
212 with spatial resampling enabled.
214 Fixed a decoder error introduced in Duclair where the segmentation
215 map was not being reinitialized on keyframes (#378)
218 2012-01-27 v1.0.0 "Duclair"
219 Our fourth named release, focused on performance and features related to
220 real-time encoding. It also fixes a decoder crash bug introduced in
221 v0.9.7, so all users of that release are encouraged to upgrade.
223 - Upgrading:
224 This release is ABI incompatible with prior releases of libvpx, so the
225 "major" version number has been bumped to 1. You must recompile your
226 applications against the latest version of the libvpx headers. The
227 API remains compatible, and this should not require code changes in most
228 applications.
230 - Enhancements:
231 This release introduces several substantial new features to the encoder,
232 of particular interest to real time streaming applications.
234 Temporal scalability allows the encoder to produce a stream that can
235 be decimated to different frame rates, with independent rate targetting
236 for each substream.
238 Multiframe quality enhancement postprocessing can make visual quality
239 more consistent in the presence of frames that are substantially
240 different quality than the surrounding frames, as in the temporal
241 scalability case and in some forced keyframe scenarios.
243 Multiple-resolution encoding support allows the encoding of the
244 same content at different resolutions faster than encoding them
245 separately.
247 - Speed:
248 Optimization targets for this release included the decoder and the real-
249 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
250 this release. Encoder improvements followed a curve where speeds 1-3
251 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
252 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
253 Cayuga release.
255 - Quality:
256 Encoder quality in the single stream case is consistent with the Cayuga
257 release.
259 - Bug Fixes:
260 This release fixes an OOB read decoder crash bug present in v0.9.7
261 related to the clamping of motion vectors in SPLITMV blocks. This
262 behavior could be triggered by corrupt input or by starting
263 decoding from a P-frame.
266 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
267 This is an incremental bugfix release against Cayuga. All users of that
268 release are strongly encouraged to upgrade.
270 - Fix potential OOB reads (cdae03a)
272 An unbounded out of bounds read was discovered when the
273 decoder was requested to perform error concealment (new in
274 Cayuga) given a frame with corrupt partition sizes.
276 A bounded out of bounds read was discovered affecting all
277 versions of libvpx. Given an multipartition input frame that
278 is truncated between the mode/mv partition and the first
279 residiual paritition (in the block of partition offsets), up
280 to 3 extra bytes could have been read from the source buffer.
281 The code will not take any action regardless of the contents
282 of these undefined bytes, as the truncated buffer is detected
283 immediately following the read based on the calculated
284 starting position of the coefficient partition.
286 - Fix potential error concealment crash when the very first frame
287 is missing or corrupt (a609be5)
289 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
291 - Revert 1-pass CBR rate control changes (e961317)
292 Further testing showed this change produced undesirable visual
293 artifacts, rolling back for now.
296 2011-08-02 v0.9.7 "Cayuga"
297 Our third named release, focused on a faster, higher quality, encoder.
299 - Upgrading:
300 This release is backwards compatible with Aylesbury (v0.9.5) and
301 Bali (v0.9.6). Users of older releases should refer to the Upgrading
302 notes in this document for that release.
304 - Enhancements:
305 Stereo 3D format support for vpxenc
306 Runtime detection of available processor cores.
307 Allow specifying --end-usage by enum name
308 vpxdec: test for frame corruption
309 vpxenc: add quantizer histogram display
310 vpxenc: add rate histogram display
311 Set VPX_FRAME_IS_DROPPABLE
312 update configure for ios sdk 4.3
313 Avoid text relocations in ARM vp8 decoder
314 Generate a vpx.pc file for pkg-config.
315 New ways of passing encoded data between encoder and decoder.
317 - Speed:
318 This release includes across-the-board speed improvements to the
319 encoder. On x86, these measure at approximately 11.5% in Best mode,
320 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
321 On ARM Cortex A9 with Neon extensions, real-time encoding of video
322 telephony content is 35% faster than Bali on single core and 48%
323 faster on multi-core. On the NVidia Tegra2 platform, real time
324 encoding is 40% faster than Bali.
326 Decoder speed was not a priority for this release, but improved
327 approximately 8.4% on x86.
329 Reduce motion vector search on alt-ref frame.
330 Encoder loopfilter running in its own thread
331 Reworked loopfilter to precalculate more parameters
332 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
333 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
334 Removed redundant checks
335 Reduced structure sizes
336 utilize preload in ARMv6 MC/LPF/Copy routines
337 ARM optimized quantization, dfct, variance, subtract
338 Increase chrow row alignment to 16 bytes.
339 disable trellis optimization for first pass
340 Write SSSE3 sub-pixel filter function
341 Improve SSE2 half-pixel filter funtions
342 Add vp8_sub_pixel_variance16x8_ssse3 function
343 Reduce unnecessary distortion computation
344 Use diamond search to replace full search
345 Preload reference area in sub-pixel motion search (real-time mode)
347 - Quality:
348 This release focused primarily on one-pass use cases, including
349 video conferencing. Low latency data rate control was significantly
350 improved, improving streamability over bandwidth constrained links.
351 Added support for error concealment, allowing frames to maintain
352 visual quality in the presence of substantial packet loss.
354 Add rc_max_intra_bitrate_pct control
355 Limit size of initial keyframe in one-pass.
356 Improve framerate adaptation
357 Improved 1-pass CBR rate control
358 Improved KF insertion after fades to still.
359 Improved key frame detection.
360 Improved activity masking (lower PSNR impact for same SSIM boost)
361 Improved interaction between GF and ARFs
362 Adding error-concealment to the decoder.
363 Adding support for independent partitions
364 Adjusted rate-distortion constants
367 - Bug Fixes:
368 Removed firstpass motion map
369 Fix parallel make install
370 Fix multithreaded encoding for 1 MB wide frame
371 Fixed iwalsh_neon build problems with RVDS4.1
372 Fix semaphore emulation, spin-wait intrinsics on Windows
373 Fix build with xcode4 and simplify GLOBAL.
374 Mark ARM asm objects as allowing a non-executable stack.
375 Fix vpxenc encoding incorrect webm file header on big endian
378 2011-03-07 v0.9.6 "Bali"
379 Our second named release, focused on a faster, higher quality, encoder.
381 - Upgrading:
382 This release is backwards compatible with Aylesbury (v0.9.5). Users
383 of older releases should refer to the Upgrading notes in this
384 document for that release.
386 - Enhancements:
387 vpxenc --psnr shows a summary when encode completes
388 --tune=ssim option to enable activity masking
389 improved postproc visualizations for development
390 updated support for Apple iOS to SDK 4.2
391 query decoder to determine which reference frames were updated
392 implemented error tracking in the decoder
393 fix pipe support on windows
395 - Speed:
396 Primary focus was on good quality mode, speed 0. Average improvement
397 on x86 about 40%, up to 100% on user-generated content at that speed.
398 Best quality mode speed improved 35%, and realtime speed 10-20%. This
399 release also saw significant improvement in realtime encoding speed
400 on ARM platforms.
402 Improved encoder threading
403 Dont pick encoder filter level when loopfilter is disabled.
404 Avoid double copying of key frames into alt and golden buffer
405 FDCT optimizations.
406 x86 sse2 temporal filter
407 SSSE3 version of fast quantizer
408 vp8_rd_pick_best_mbsegmentation code restructure
409 Adjusted breakout RD for SPLITMV
410 Changed segmentation check order
411 Improved rd_pick_intra4x4block
412 Adds armv6 optimized variance calculation
413 ARMv6 optimized sad16x16
414 ARMv6 optimized half pixel variance calculations
415 Full search SAD function optimization in SSE4.1
416 Improve MV prediction accuracy to achieve performance gain
417 Improve MV prediction in vp8_pick_inter_mode() for speed>3
419 - Quality:
420 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
421 also includes support for "activity masking," which greatly improves
422 SSIM at the expense of PSNR. For now, this feature is available with
423 the --tune=ssim option. Further experimentation in this area
424 is ongoing. This release also introduces a new rate control mode
425 called "CQ," which changes the allocation of bits within a clip to
426 the sections where they will have the most visual impact.
428 Tuning for the more exact quantizer.
429 Relax rate control for last few frames
430 CQ Mode
431 Limit key frame quantizer for forced key frames.
432 KF/GF Pulsing
433 Add simple version of activity masking.
434 make rdmult adaptive for intra in quantizer RDO
435 cap the best quantizer for 2nd order DC
436 change the threshold of DC check for encode breakout
438 - Bug Fixes:
439 Fix crash on Sparc Solaris.
440 Fix counter of fixed keyframe distance
441 ARNR filter pointer update bug fix
442 Fixed use of motion percentage in KF/GF group calc
443 Changed condition for using RD in Intra Mode
444 Fix encoder real-time only configuration.
445 Fix ARM encoder crash with multiple token partitions
446 Fixed bug first cluster timecode of webm file is wrong.
447 Fixed various encoder bugs with odd-sized images
448 vp8e_get_preview fixed when spatial resampling enabled
449 quantizer: fix assertion in fast quantizer path
450 Allocate source buffers to be multiples of 16
451 Fix for manual Golden frame frequency
452 Fix drastic undershoot in long form content
455 2010-10-28 v0.9.5 "Aylesbury"
456 Our first named release, focused on a faster decoder, and a better encoder.
458 - Upgrading:
459 This release incorporates backwards-incompatible changes to the
460 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
462 vpxdec
463 * the -q (quiet) option has been removed, and replaced with
464 -v (verbose). the output is quiet by default. Use -v to see
465 the version number of the binary.
467 * The default behavior is now to write output to a single file
468 instead of individual frames. The -y option has been removed.
469 Y4M output is the default.
471 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
472 options must be specified.
474 $ ivfdec -o OUTPUT INPUT
475 $ vpxdec --i420 -o OUTPUT INPUT
477 * If an output file is not specified, the default is to write
478 Y4M to stdout. This makes piping more natural.
480 $ ivfdec -y -o - INPUT | ...
481 $ vpxdec INPUT | ...
483 * The output file has additional flexibility for formatting the
484 filename. It supports escape characters for constructing a
485 filename from the width, height, and sequence number. This
486 replaces the -p option. To get the equivalent:
488 $ ivfdec -p frame INPUT
489 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
491 vpxenc
492 * The output file must be specified with -o, rather than as the
493 last argument.
495 $ ivfenc <options> INPUT OUTPUT
496 $ vpxenc <options> -o OUTPUT INPUT
498 * The output defaults to webm. To get IVF output, use the --ivf
499 option.
501 $ ivfenc <options> INPUT OUTPUT.ivf
502 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
505 - Enhancements:
506 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
507 vpxdec supports .webm input
508 vpxdec writes .y4m by default
509 vpxenc writes .webm output by default
510 vpxenc --psnr now shows the average/overall PSNR at the end
511 ARM platforms now support runtime cpu detection
512 vpxdec visualizations added for motion vectors, block modes, references
513 vpxdec now silent by default
514 vpxdec --progress shows frame-by-frame timing information
515 vpxenc supports the distinction between --fps and --timebase
516 NASM is now a supported assembler
517 configure: enable PIC for shared libs by default
518 configure: add --enable-small
519 configure: support for ppc32-linux-gcc
520 configure: support for sparc-solaris-gcc
522 - Bugs:
523 Improve handling of invalid frames
524 Fix valgrind errors in the NEON loop filters.
525 Fix loopfilter delta zero transitions
526 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
527 Build fixes for darwin-icc
529 - Speed:
530 20-40% (average 28%) improvement in libvpx decoder speed,
531 including:
532 Rewrite vp8_short_walsh4x4_sse2()
533 Optimizations on the loopfilters.
534 Miscellaneous improvements for Atom
535 Add 4-tap version of 2nd-pass ARMv6 MC filter.
536 Improved multithread utilization
537 Better instruction choices on x86
538 reorder data to use wider instructions
539 Update NEON wide idcts
540 Make block access to frame buffer sequential
541 Improved subset block search
542 Bilinear subpixel optimizations for ssse3.
543 Decrease memory footprint
545 Encoder speed improvements (percentage gain not measured):
546 Skip unnecessary search of identical frames
547 Add SSE2 subtract functions
548 Improve bounds checking in vp8_diamond_search_sadx4()
549 Added vp8_fast_quantize_b_sse2
551 - Quality:
552 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
553 encoding mode, and up to 60% improvement on very noisy, still
554 or slow moving source video
556 Motion compensated temporal filter for Alt-Ref Noise Reduction
557 Improved use of trellis quantization on 2nd order Y blocks
558 Tune effect of motion on KF/GF boost in two pass
559 Allow coefficient optimization for good quality speed 0.
560 Improved control of active min quantizer for two pass.
561 Enable ARFs for non-lagged compress
563 2010-09-02 v0.9.2
564 - Enhancements:
565 Disable frame dropping by default
566 Improved multithreaded performance
567 Improved Force Key Frame Behaviour
568 Increased rate control buffer level precision
569 Fix bug in 1st pass motion compensation
570 ivfenc: correct fixed kf interval, --disable-kf
571 - Speed:
572 Changed above and left context data layout
573 Rework idct calling structure.
574 Removed unnecessary MB_MODE_INFO copies
575 x86: SSSE3 sixtap prediction
576 Reworked IDCT to include reconstruction (add) step
577 Swap alt/gold/new/last frame buffer ptrs instead of copying.
578 Improve SSE2 loopfilter functions
579 Change bitreader to use a larger window.
580 Avoid loopfilter reinitialization when possible
581 - Quality:
582 Normalize quantizer's zero bin and rounding factors
583 Add trellis quantization.
584 Make the quantizer exact.
585 Updates to ARNR filtering algorithm
586 Fix breakout thresh computation for golden & AltRef frames
587 Redo the forward 4x4 dct
588 Improve the accuracy of forward walsh-hadamard transform
589 Further adjustment of RD behaviour with Q and Zbin.
590 - Build System:
591 Allow linking of libs built with MinGW to MSVC
592 Fix target auto-detection on mingw32
593 Allow --cpu= to work for x86.
594 configure: pass original arguments through to make dist
595 Fix builds without runtime CPU detection
596 msvs: fix install of codec sources
597 msvs: Change devenv.com command line for better msys support
598 msvs: Add vs9 targets.
599 Add x86_64-linux-icc target
600 - Bugs:
601 Potential crashes on older MinGW builds
602 Fix two-pass framrate for Y4M input.
603 Fixed simple loop filter, other crashes on ARM v6
604 arm: fix missing dependency with --enable-shared
605 configure: support directories containing .o
606 Replace pinsrw (SSE) with MMX instructions
607 apple: include proper mach primatives
608 Fixed rate control bug with long key frame interval.
609 Fix DSO link errors on x86-64 when not using a version script
610 Fixed buffer selection for UV in AltRef filtering
613 2010-06-17 v0.9.1
614 - Enhancements:
615 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
616 * Speed optimizations
617 - Bugfixes:
618 * Rate control
619 * Prevent out-of-bounds accesses on invalid data
620 - Build system updates:
621 * Detect toolchain to be used automatically for native builds
622 * Support building shared libraries
623 * Better autotools emulation (--prefix, --libdir, DESTDIR)
624 - Updated LICENSE
625 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
628 2010-05-18 v0.9.0
629 - Initial open source release. Welcome to WebM and VP8!