1 xxxx-yy-zz v1.4.0 "Changes for next release"
2 vpxenc is changed to use VP9 by default.
3 Encoder controls added for 1 pass SVC.
5 2015-04-03 v1.4.0 "Indian Runner Duck"
6 This release includes significant improvements to the VP9 codec.
9 This release is ABI incompatible with 1.3.0. It drops the compatibility
10 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
14 Faster VP9 encoding and decoding
15 Multithreaded VP9 decoding (tile and frame-based)
16 Multithreaded VP9 encoding - on by default
17 YUV 4:2:2 and 4:4:4 support in VP9
18 10 and 12bit support in VP9
19 64bit ARM support by replacing ARM assembly with intrinsics
22 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
26 Frame Parallel decoding fails for segmented and non-420 files.
28 2013-11-15 v1.3.0 "Forest"
29 This release introduces the VP9 codec in a backward-compatible way.
30 All existing users of VP8 can continue to use the library without
31 modification. However, some VP8 options do not map to VP9 in the same manner.
33 The VP9 encoder in this release is not feature complete. Users interested in
34 the encoder are advised to use the git master branch and discuss issues on
38 This release is ABI and API compatible with Duclair (v1.0.0). Users
39 of older releases should refer to the Upgrading notes in this document
43 Get rid of bashisms in the main build scripts
44 Added usage info on command line options
45 Add lossless compression mode
47 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
48 Add option to disable documentation
49 configure: add --enable-external-build support
50 make: support V=1 as short form of verbose=yes
51 configure: support mingw-w64
52 configure: support hardfloat armv7 CHOSTS
53 configure: add support for android x86
54 Add estimated completion time to vpxenc
55 Don't exit on decode errors in vpxenc
56 vpxenc: support scaling prior to encoding
57 vpxdec: support scaling output
58 vpxenc: improve progress indicators with --skip
59 msvs: Don't link to winmm.lib
60 Add a new script for producing vcxproj files
61 Produce Visual Studio 10 and 11 project files
62 Produce Windows Phone project files
63 msvs-build: use msbuild for vs >= 2005
64 configure: default configure log to config.log
65 Add encoding option --static-thresh
68 Miscellaneous speed optimizations for VP8 and VP9.
71 In general, quality is consistent with the Eider release.
74 This release represents approximately a year of engineering effort,
75 and contains multiple bug fixes. Please refer to git history for details.
79 This release acts as a checkpoint for a large amount of internal refactoring
80 and testing. It also contains a number of small bugfixes, so all users are
81 encouraged to upgrade.
84 This release is ABI and API compatible with Duclair (v1.0.0). Users
85 of older releases should refer to the Upgrading notes in this
86 document for that release.
89 VP8 optimizations for MIPS dspr2
90 vpxenc: add -quiet option
93 Encoder and decoder speed is consistent with the Eider release.
96 In general, quality is consistent with the Eider release.
98 Minor tweaks to ARNR filtering
99 Minor improvements to real time encoding with multiple temporal layers
102 Fixes multithreaded encoder race condition in loopfilter
103 Fixes multi-resolution threaded encoding
104 Fix potential encoder dead-lock after picture resize
107 2012-05-09 v1.1.0 "Eider"
108 This introduces a number of enhancements, mostly focused on real-time
109 encoding. In addition, it fixes a decoder bug (first introduced in
110 Duclair) so all users of that release are encouraged to upgrade.
113 This release is ABI and API compatible with Duclair (v1.0.0). Users
114 of older releases should refer to the Upgrading notes in this
115 document for that release.
117 This release introduces a new temporal denoiser, controlled by the
118 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
119 currently take a strength parameter, so the control is effectively
120 a boolean - zero (off) or non-zero (on). For compatibility with
121 existing applications, the values accepted are the same as those
122 for the spatial denoiser (0-6). The temporal denoiser is enabled
123 by default, and the older spatial denoiser may be restored by
124 configuring with --disable-temporal-denoising. The temporal denoiser
125 is more computationally intensive than the spatial one.
127 This release removes support for a legacy, decode only API that was
128 supported, but deprecated, at the initial release of libvpx
129 (v0.9.0). This is not expected to have any impact. If you are
130 impacted, you can apply a reversion to commit 2bf8fb58 locally.
131 Please update to the latest libvpx API if you are affected.
134 Adds a motion compensated temporal denoiser to the encoder, which
135 gives higher quality than the older spatial denoiser. (See above
136 for notes on upgrading).
138 In addition, support for new compilers and platforms were added,
140 improved support for XCode
141 Android x86 NDK build
145 Changing resolution with vpx_codec_enc_config_set() is now
146 supported. Previously, reinitializing the codec was required to
147 change the input resolution.
149 The vpxenc application has initial support for producing multiple
150 encodes from the same input in one call. Resizing is not yet
151 supported, but varying other codec parameters is. Use -- to
152 delineate output streams. Options persist from one stream to the
155 Also, the vpxenc application will now use a keyframe interval of
156 5 seconds by default. Use the --kf-max-dist option to override.
159 Decoder performance improved 2.5% versus Duclair. Encoder speed is
160 consistent with Duclair for most material. Two pass encoding of
161 slideshow-like material will see significant improvements.
163 Large realtime encoding speed gains at a small quality expense are
164 possible by configuring the on-the-fly bitpacking experiment with
165 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
166 faster (ARM) depending on the number of threads and bitrate
167 settings. This technique sees constant gain over the 5-16 speed
168 range. For VC style input the loss seen is up to 0.2dB. See commit
169 52cf4dca for further details.
172 On the whole, quality is consistent with the Duclair release. Some
175 Reduced blockiness in easy sections by applying a penalty to
178 Improved quality of static sections (like slideshows) with
181 Improved keyframe sizing with multiple temporal layers
184 Corrected alt-ref contribution to frame rate for visible updates
185 to the alt-ref buffer. This affected applications making manual
186 usage of the frame reference flags, or temporal layers.
188 Additional constraints were added to disable multi-frame quality
189 enhancement (MFQE) in sections of the frame where there is motion.
192 Fixed corruption issues when vpx_codec_enc_config_set() was called
193 with spatial resampling enabled.
195 Fixed a decoder error introduced in Duclair where the segmentation
196 map was not being reinitialized on keyframes (#378)
199 2012-01-27 v1.0.0 "Duclair"
200 Our fourth named release, focused on performance and features related to
201 real-time encoding. It also fixes a decoder crash bug introduced in
202 v0.9.7, so all users of that release are encouraged to upgrade.
205 This release is ABI incompatible with prior releases of libvpx, so the
206 "major" version number has been bumped to 1. You must recompile your
207 applications against the latest version of the libvpx headers. The
208 API remains compatible, and this should not require code changes in most
212 This release introduces several substantial new features to the encoder,
213 of particular interest to real time streaming applications.
215 Temporal scalability allows the encoder to produce a stream that can
216 be decimated to different frame rates, with independent rate targetting
219 Multiframe quality enhancement postprocessing can make visual quality
220 more consistent in the presence of frames that are substantially
221 different quality than the surrounding frames, as in the temporal
222 scalability case and in some forced keyframe scenarios.
224 Multiple-resolution encoding support allows the encoding of the
225 same content at different resolutions faster than encoding them
229 Optimization targets for this release included the decoder and the real-
230 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
231 this release. Encoder improvements followed a curve where speeds 1-3
232 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
233 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
237 Encoder quality in the single stream case is consistent with the Cayuga
241 This release fixes an OOB read decoder crash bug present in v0.9.7
242 related to the clamping of motion vectors in SPLITMV blocks. This
243 behavior could be triggered by corrupt input or by starting
244 decoding from a P-frame.
247 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
248 This is an incremental bugfix release against Cayuga. All users of that
249 release are strongly encouraged to upgrade.
251 - Fix potential OOB reads (cdae03a)
253 An unbounded out of bounds read was discovered when the
254 decoder was requested to perform error concealment (new in
255 Cayuga) given a frame with corrupt partition sizes.
257 A bounded out of bounds read was discovered affecting all
258 versions of libvpx. Given an multipartition input frame that
259 is truncated between the mode/mv partition and the first
260 residiual paritition (in the block of partition offsets), up
261 to 3 extra bytes could have been read from the source buffer.
262 The code will not take any action regardless of the contents
263 of these undefined bytes, as the truncated buffer is detected
264 immediately following the read based on the calculated
265 starting position of the coefficient partition.
267 - Fix potential error concealment crash when the very first frame
268 is missing or corrupt (a609be5)
270 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
272 - Revert 1-pass CBR rate control changes (e961317)
273 Further testing showed this change produced undesirable visual
274 artifacts, rolling back for now.
277 2011-08-02 v0.9.7 "Cayuga"
278 Our third named release, focused on a faster, higher quality, encoder.
281 This release is backwards compatible with Aylesbury (v0.9.5) and
282 Bali (v0.9.6). Users of older releases should refer to the Upgrading
283 notes in this document for that release.
286 Stereo 3D format support for vpxenc
287 Runtime detection of available processor cores.
288 Allow specifying --end-usage by enum name
289 vpxdec: test for frame corruption
290 vpxenc: add quantizer histogram display
291 vpxenc: add rate histogram display
292 Set VPX_FRAME_IS_DROPPABLE
293 update configure for ios sdk 4.3
294 Avoid text relocations in ARM vp8 decoder
295 Generate a vpx.pc file for pkg-config.
296 New ways of passing encoded data between encoder and decoder.
299 This release includes across-the-board speed improvements to the
300 encoder. On x86, these measure at approximately 11.5% in Best mode,
301 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
302 On ARM Cortex A9 with Neon extensions, real-time encoding of video
303 telephony content is 35% faster than Bali on single core and 48%
304 faster on multi-core. On the NVidia Tegra2 platform, real time
305 encoding is 40% faster than Bali.
307 Decoder speed was not a priority for this release, but improved
308 approximately 8.4% on x86.
310 Reduce motion vector search on alt-ref frame.
311 Encoder loopfilter running in its own thread
312 Reworked loopfilter to precalculate more parameters
313 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
314 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
315 Removed redundant checks
316 Reduced structure sizes
317 utilize preload in ARMv6 MC/LPF/Copy routines
318 ARM optimized quantization, dfct, variance, subtract
319 Increase chrow row alignment to 16 bytes.
320 disable trellis optimization for first pass
321 Write SSSE3 sub-pixel filter function
322 Improve SSE2 half-pixel filter funtions
323 Add vp8_sub_pixel_variance16x8_ssse3 function
324 Reduce unnecessary distortion computation
325 Use diamond search to replace full search
326 Preload reference area in sub-pixel motion search (real-time mode)
329 This release focused primarily on one-pass use cases, including
330 video conferencing. Low latency data rate control was significantly
331 improved, improving streamability over bandwidth constrained links.
332 Added support for error concealment, allowing frames to maintain
333 visual quality in the presence of substantial packet loss.
335 Add rc_max_intra_bitrate_pct control
336 Limit size of initial keyframe in one-pass.
337 Improve framerate adaptation
338 Improved 1-pass CBR rate control
339 Improved KF insertion after fades to still.
340 Improved key frame detection.
341 Improved activity masking (lower PSNR impact for same SSIM boost)
342 Improved interaction between GF and ARFs
343 Adding error-concealment to the decoder.
344 Adding support for independent partitions
345 Adjusted rate-distortion constants
349 Removed firstpass motion map
350 Fix parallel make install
351 Fix multithreaded encoding for 1 MB wide frame
352 Fixed iwalsh_neon build problems with RVDS4.1
353 Fix semaphore emulation, spin-wait intrinsics on Windows
354 Fix build with xcode4 and simplify GLOBAL.
355 Mark ARM asm objects as allowing a non-executable stack.
356 Fix vpxenc encoding incorrect webm file header on big endian
359 2011-03-07 v0.9.6 "Bali"
360 Our second named release, focused on a faster, higher quality, encoder.
363 This release is backwards compatible with Aylesbury (v0.9.5). Users
364 of older releases should refer to the Upgrading notes in this
365 document for that release.
368 vpxenc --psnr shows a summary when encode completes
369 --tune=ssim option to enable activity masking
370 improved postproc visualizations for development
371 updated support for Apple iOS to SDK 4.2
372 query decoder to determine which reference frames were updated
373 implemented error tracking in the decoder
374 fix pipe support on windows
377 Primary focus was on good quality mode, speed 0. Average improvement
378 on x86 about 40%, up to 100% on user-generated content at that speed.
379 Best quality mode speed improved 35%, and realtime speed 10-20%. This
380 release also saw significant improvement in realtime encoding speed
383 Improved encoder threading
384 Dont pick encoder filter level when loopfilter is disabled.
385 Avoid double copying of key frames into alt and golden buffer
387 x86 sse2 temporal filter
388 SSSE3 version of fast quantizer
389 vp8_rd_pick_best_mbsegmentation code restructure
390 Adjusted breakout RD for SPLITMV
391 Changed segmentation check order
392 Improved rd_pick_intra4x4block
393 Adds armv6 optimized variance calculation
394 ARMv6 optimized sad16x16
395 ARMv6 optimized half pixel variance calculations
396 Full search SAD function optimization in SSE4.1
397 Improve MV prediction accuracy to achieve performance gain
398 Improve MV prediction in vp8_pick_inter_mode() for speed>3
401 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
402 also includes support for "activity masking," which greatly improves
403 SSIM at the expense of PSNR. For now, this feature is available with
404 the --tune=ssim option. Further experimentation in this area
405 is ongoing. This release also introduces a new rate control mode
406 called "CQ," which changes the allocation of bits within a clip to
407 the sections where they will have the most visual impact.
409 Tuning for the more exact quantizer.
410 Relax rate control for last few frames
412 Limit key frame quantizer for forced key frames.
414 Add simple version of activity masking.
415 make rdmult adaptive for intra in quantizer RDO
416 cap the best quantizer for 2nd order DC
417 change the threshold of DC check for encode breakout
420 Fix crash on Sparc Solaris.
421 Fix counter of fixed keyframe distance
422 ARNR filter pointer update bug fix
423 Fixed use of motion percentage in KF/GF group calc
424 Changed condition for using RD in Intra Mode
425 Fix encoder real-time only configuration.
426 Fix ARM encoder crash with multiple token partitions
427 Fixed bug first cluster timecode of webm file is wrong.
428 Fixed various encoder bugs with odd-sized images
429 vp8e_get_preview fixed when spatial resampling enabled
430 quantizer: fix assertion in fast quantizer path
431 Allocate source buffers to be multiples of 16
432 Fix for manual Golden frame frequency
433 Fix drastic undershoot in long form content
436 2010-10-28 v0.9.5 "Aylesbury"
437 Our first named release, focused on a faster decoder, and a better encoder.
440 This release incorporates backwards-incompatible changes to the
441 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
444 * the -q (quiet) option has been removed, and replaced with
445 -v (verbose). the output is quiet by default. Use -v to see
446 the version number of the binary.
448 * The default behavior is now to write output to a single file
449 instead of individual frames. The -y option has been removed.
450 Y4M output is the default.
452 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
453 options must be specified.
455 $ ivfdec -o OUTPUT INPUT
456 $ vpxdec --i420 -o OUTPUT INPUT
458 * If an output file is not specified, the default is to write
459 Y4M to stdout. This makes piping more natural.
461 $ ivfdec -y -o - INPUT | ...
464 * The output file has additional flexibility for formatting the
465 filename. It supports escape characters for constructing a
466 filename from the width, height, and sequence number. This
467 replaces the -p option. To get the equivalent:
469 $ ivfdec -p frame INPUT
470 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
473 * The output file must be specified with -o, rather than as the
476 $ ivfenc <options> INPUT OUTPUT
477 $ vpxenc <options> -o OUTPUT INPUT
479 * The output defaults to webm. To get IVF output, use the --ivf
482 $ ivfenc <options> INPUT OUTPUT.ivf
483 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
487 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
488 vpxdec supports .webm input
489 vpxdec writes .y4m by default
490 vpxenc writes .webm output by default
491 vpxenc --psnr now shows the average/overall PSNR at the end
492 ARM platforms now support runtime cpu detection
493 vpxdec visualizations added for motion vectors, block modes, references
494 vpxdec now silent by default
495 vpxdec --progress shows frame-by-frame timing information
496 vpxenc supports the distinction between --fps and --timebase
497 NASM is now a supported assembler
498 configure: enable PIC for shared libs by default
499 configure: add --enable-small
500 configure: support for ppc32-linux-gcc
501 configure: support for sparc-solaris-gcc
504 Improve handling of invalid frames
505 Fix valgrind errors in the NEON loop filters.
506 Fix loopfilter delta zero transitions
507 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
508 Build fixes for darwin-icc
511 20-40% (average 28%) improvement in libvpx decoder speed,
513 Rewrite vp8_short_walsh4x4_sse2()
514 Optimizations on the loopfilters.
515 Miscellaneous improvements for Atom
516 Add 4-tap version of 2nd-pass ARMv6 MC filter.
517 Improved multithread utilization
518 Better instruction choices on x86
519 reorder data to use wider instructions
520 Update NEON wide idcts
521 Make block access to frame buffer sequential
522 Improved subset block search
523 Bilinear subpixel optimizations for ssse3.
524 Decrease memory footprint
526 Encoder speed improvements (percentage gain not measured):
527 Skip unnecessary search of identical frames
528 Add SSE2 subtract functions
529 Improve bounds checking in vp8_diamond_search_sadx4()
530 Added vp8_fast_quantize_b_sse2
533 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
534 encoding mode, and up to 60% improvement on very noisy, still
535 or slow moving source video
537 Motion compensated temporal filter for Alt-Ref Noise Reduction
538 Improved use of trellis quantization on 2nd order Y blocks
539 Tune effect of motion on KF/GF boost in two pass
540 Allow coefficient optimization for good quality speed 0.
541 Improved control of active min quantizer for two pass.
542 Enable ARFs for non-lagged compress
546 Disable frame dropping by default
547 Improved multithreaded performance
548 Improved Force Key Frame Behaviour
549 Increased rate control buffer level precision
550 Fix bug in 1st pass motion compensation
551 ivfenc: correct fixed kf interval, --disable-kf
553 Changed above and left context data layout
554 Rework idct calling structure.
555 Removed unnecessary MB_MODE_INFO copies
556 x86: SSSE3 sixtap prediction
557 Reworked IDCT to include reconstruction (add) step
558 Swap alt/gold/new/last frame buffer ptrs instead of copying.
559 Improve SSE2 loopfilter functions
560 Change bitreader to use a larger window.
561 Avoid loopfilter reinitialization when possible
563 Normalize quantizer's zero bin and rounding factors
564 Add trellis quantization.
565 Make the quantizer exact.
566 Updates to ARNR filtering algorithm
567 Fix breakout thresh computation for golden & AltRef frames
568 Redo the forward 4x4 dct
569 Improve the accuracy of forward walsh-hadamard transform
570 Further adjustment of RD behaviour with Q and Zbin.
572 Allow linking of libs built with MinGW to MSVC
573 Fix target auto-detection on mingw32
574 Allow --cpu= to work for x86.
575 configure: pass original arguments through to make dist
576 Fix builds without runtime CPU detection
577 msvs: fix install of codec sources
578 msvs: Change devenv.com command line for better msys support
579 msvs: Add vs9 targets.
580 Add x86_64-linux-icc target
582 Potential crashes on older MinGW builds
583 Fix two-pass framrate for Y4M input.
584 Fixed simple loop filter, other crashes on ARM v6
585 arm: fix missing dependency with --enable-shared
586 configure: support directories containing .o
587 Replace pinsrw (SSE) with MMX instructions
588 apple: include proper mach primatives
589 Fixed rate control bug with long key frame interval.
590 Fix DSO link errors on x86-64 when not using a version script
591 Fixed buffer selection for UV in AltRef filtering
596 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
597 * Speed optimizations
600 * Prevent out-of-bounds accesses on invalid data
601 - Build system updates:
602 * Detect toolchain to be used automatically for native builds
603 * Support building shared libraries
604 * Better autotools emulation (--prefix, --libdir, DESTDIR)
606 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
610 - Initial open source release. Welcome to WebM and VP8!