1 xxxx-yy-zz v1.4.0 "Changes for next release"
2 Encoder controls added for 1 pass SVC.
4 2015-04-03 v1.4.0 "Indian Runner Duck"
5 This release includes significant improvements to the VP9 codec.
8 This release is ABI incompatible with 1.3.0. It drops the compatibility
9 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
13 Faster VP9 encoding and decoding
14 Multithreaded VP9 decoding (tile and frame-based)
15 Multithreaded VP9 encoding - on by default
16 YUV 4:2:2 and 4:4:4 support in VP9
17 10 and 12bit support in VP9
18 64bit ARM support by replacing ARM assembly with intrinsics
21 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
25 Frame Parallel decoding fails for segmented and non-420 files.
27 2013-11-15 v1.3.0 "Forest"
28 This release introduces the VP9 codec in a backward-compatible way.
29 All existing users of VP8 can continue to use the library without
30 modification. However, some VP8 options do not map to VP9 in the same manner.
32 The VP9 encoder in this release is not feature complete. Users interested in
33 the encoder are advised to use the git master branch and discuss issues on
37 This release is ABI and API compatible with Duclair (v1.0.0). Users
38 of older releases should refer to the Upgrading notes in this document
42 Get rid of bashisms in the main build scripts
43 Added usage info on command line options
44 Add lossless compression mode
46 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
47 Add option to disable documentation
48 configure: add --enable-external-build support
49 make: support V=1 as short form of verbose=yes
50 configure: support mingw-w64
51 configure: support hardfloat armv7 CHOSTS
52 configure: add support for android x86
53 Add estimated completion time to vpxenc
54 Don't exit on decode errors in vpxenc
55 vpxenc: support scaling prior to encoding
56 vpxdec: support scaling output
57 vpxenc: improve progress indicators with --skip
58 msvs: Don't link to winmm.lib
59 Add a new script for producing vcxproj files
60 Produce Visual Studio 10 and 11 project files
61 Produce Windows Phone project files
62 msvs-build: use msbuild for vs >= 2005
63 configure: default configure log to config.log
64 Add encoding option --static-thresh
67 Miscellaneous speed optimizations for VP8 and VP9.
70 In general, quality is consistent with the Eider release.
73 This release represents approximately a year of engineering effort,
74 and contains multiple bug fixes. Please refer to git history for details.
78 This release acts as a checkpoint for a large amount of internal refactoring
79 and testing. It also contains a number of small bugfixes, so all users are
80 encouraged to upgrade.
83 This release is ABI and API compatible with Duclair (v1.0.0). Users
84 of older releases should refer to the Upgrading notes in this
85 document for that release.
88 VP8 optimizations for MIPS dspr2
89 vpxenc: add -quiet option
92 Encoder and decoder speed is consistent with the Eider release.
95 In general, quality is consistent with the Eider release.
97 Minor tweaks to ARNR filtering
98 Minor improvements to real time encoding with multiple temporal layers
101 Fixes multithreaded encoder race condition in loopfilter
102 Fixes multi-resolution threaded encoding
103 Fix potential encoder dead-lock after picture resize
106 2012-05-09 v1.1.0 "Eider"
107 This introduces a number of enhancements, mostly focused on real-time
108 encoding. In addition, it fixes a decoder bug (first introduced in
109 Duclair) so all users of that release are encouraged to upgrade.
112 This release is ABI and API compatible with Duclair (v1.0.0). Users
113 of older releases should refer to the Upgrading notes in this
114 document for that release.
116 This release introduces a new temporal denoiser, controlled by the
117 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
118 currently take a strength parameter, so the control is effectively
119 a boolean - zero (off) or non-zero (on). For compatibility with
120 existing applications, the values accepted are the same as those
121 for the spatial denoiser (0-6). The temporal denoiser is enabled
122 by default, and the older spatial denoiser may be restored by
123 configuring with --disable-temporal-denoising. The temporal denoiser
124 is more computationally intensive than the spatial one.
126 This release removes support for a legacy, decode only API that was
127 supported, but deprecated, at the initial release of libvpx
128 (v0.9.0). This is not expected to have any impact. If you are
129 impacted, you can apply a reversion to commit 2bf8fb58 locally.
130 Please update to the latest libvpx API if you are affected.
133 Adds a motion compensated temporal denoiser to the encoder, which
134 gives higher quality than the older spatial denoiser. (See above
135 for notes on upgrading).
137 In addition, support for new compilers and platforms were added,
139 improved support for XCode
140 Android x86 NDK build
144 Changing resolution with vpx_codec_enc_config_set() is now
145 supported. Previously, reinitializing the codec was required to
146 change the input resolution.
148 The vpxenc application has initial support for producing multiple
149 encodes from the same input in one call. Resizing is not yet
150 supported, but varying other codec parameters is. Use -- to
151 delineate output streams. Options persist from one stream to the
154 Also, the vpxenc application will now use a keyframe interval of
155 5 seconds by default. Use the --kf-max-dist option to override.
158 Decoder performance improved 2.5% versus Duclair. Encoder speed is
159 consistent with Duclair for most material. Two pass encoding of
160 slideshow-like material will see significant improvements.
162 Large realtime encoding speed gains at a small quality expense are
163 possible by configuring the on-the-fly bitpacking experiment with
164 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
165 faster (ARM) depending on the number of threads and bitrate
166 settings. This technique sees constant gain over the 5-16 speed
167 range. For VC style input the loss seen is up to 0.2dB. See commit
168 52cf4dca for further details.
171 On the whole, quality is consistent with the Duclair release. Some
174 Reduced blockiness in easy sections by applying a penalty to
177 Improved quality of static sections (like slideshows) with
180 Improved keyframe sizing with multiple temporal layers
183 Corrected alt-ref contribution to frame rate for visible updates
184 to the alt-ref buffer. This affected applications making manual
185 usage of the frame reference flags, or temporal layers.
187 Additional constraints were added to disable multi-frame quality
188 enhancement (MFQE) in sections of the frame where there is motion.
191 Fixed corruption issues when vpx_codec_enc_config_set() was called
192 with spatial resampling enabled.
194 Fixed a decoder error introduced in Duclair where the segmentation
195 map was not being reinitialized on keyframes (#378)
198 2012-01-27 v1.0.0 "Duclair"
199 Our fourth named release, focused on performance and features related to
200 real-time encoding. It also fixes a decoder crash bug introduced in
201 v0.9.7, so all users of that release are encouraged to upgrade.
204 This release is ABI incompatible with prior releases of libvpx, so the
205 "major" version number has been bumped to 1. You must recompile your
206 applications against the latest version of the libvpx headers. The
207 API remains compatible, and this should not require code changes in most
211 This release introduces several substantial new features to the encoder,
212 of particular interest to real time streaming applications.
214 Temporal scalability allows the encoder to produce a stream that can
215 be decimated to different frame rates, with independent rate targetting
218 Multiframe quality enhancement postprocessing can make visual quality
219 more consistent in the presence of frames that are substantially
220 different quality than the surrounding frames, as in the temporal
221 scalability case and in some forced keyframe scenarios.
223 Multiple-resolution encoding support allows the encoding of the
224 same content at different resolutions faster than encoding them
228 Optimization targets for this release included the decoder and the real-
229 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
230 this release. Encoder improvements followed a curve where speeds 1-3
231 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
232 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
236 Encoder quality in the single stream case is consistent with the Cayuga
240 This release fixes an OOB read decoder crash bug present in v0.9.7
241 related to the clamping of motion vectors in SPLITMV blocks. This
242 behavior could be triggered by corrupt input or by starting
243 decoding from a P-frame.
246 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
247 This is an incremental bugfix release against Cayuga. All users of that
248 release are strongly encouraged to upgrade.
250 - Fix potential OOB reads (cdae03a)
252 An unbounded out of bounds read was discovered when the
253 decoder was requested to perform error concealment (new in
254 Cayuga) given a frame with corrupt partition sizes.
256 A bounded out of bounds read was discovered affecting all
257 versions of libvpx. Given an multipartition input frame that
258 is truncated between the mode/mv partition and the first
259 residiual paritition (in the block of partition offsets), up
260 to 3 extra bytes could have been read from the source buffer.
261 The code will not take any action regardless of the contents
262 of these undefined bytes, as the truncated buffer is detected
263 immediately following the read based on the calculated
264 starting position of the coefficient partition.
266 - Fix potential error concealment crash when the very first frame
267 is missing or corrupt (a609be5)
269 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
271 - Revert 1-pass CBR rate control changes (e961317)
272 Further testing showed this change produced undesirable visual
273 artifacts, rolling back for now.
276 2011-08-02 v0.9.7 "Cayuga"
277 Our third named release, focused on a faster, higher quality, encoder.
280 This release is backwards compatible with Aylesbury (v0.9.5) and
281 Bali (v0.9.6). Users of older releases should refer to the Upgrading
282 notes in this document for that release.
285 Stereo 3D format support for vpxenc
286 Runtime detection of available processor cores.
287 Allow specifying --end-usage by enum name
288 vpxdec: test for frame corruption
289 vpxenc: add quantizer histogram display
290 vpxenc: add rate histogram display
291 Set VPX_FRAME_IS_DROPPABLE
292 update configure for ios sdk 4.3
293 Avoid text relocations in ARM vp8 decoder
294 Generate a vpx.pc file for pkg-config.
295 New ways of passing encoded data between encoder and decoder.
298 This release includes across-the-board speed improvements to the
299 encoder. On x86, these measure at approximately 11.5% in Best mode,
300 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
301 On ARM Cortex A9 with Neon extensions, real-time encoding of video
302 telephony content is 35% faster than Bali on single core and 48%
303 faster on multi-core. On the NVidia Tegra2 platform, real time
304 encoding is 40% faster than Bali.
306 Decoder speed was not a priority for this release, but improved
307 approximately 8.4% on x86.
309 Reduce motion vector search on alt-ref frame.
310 Encoder loopfilter running in its own thread
311 Reworked loopfilter to precalculate more parameters
312 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
313 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
314 Removed redundant checks
315 Reduced structure sizes
316 utilize preload in ARMv6 MC/LPF/Copy routines
317 ARM optimized quantization, dfct, variance, subtract
318 Increase chrow row alignment to 16 bytes.
319 disable trellis optimization for first pass
320 Write SSSE3 sub-pixel filter function
321 Improve SSE2 half-pixel filter funtions
322 Add vp8_sub_pixel_variance16x8_ssse3 function
323 Reduce unnecessary distortion computation
324 Use diamond search to replace full search
325 Preload reference area in sub-pixel motion search (real-time mode)
328 This release focused primarily on one-pass use cases, including
329 video conferencing. Low latency data rate control was significantly
330 improved, improving streamability over bandwidth constrained links.
331 Added support for error concealment, allowing frames to maintain
332 visual quality in the presence of substantial packet loss.
334 Add rc_max_intra_bitrate_pct control
335 Limit size of initial keyframe in one-pass.
336 Improve framerate adaptation
337 Improved 1-pass CBR rate control
338 Improved KF insertion after fades to still.
339 Improved key frame detection.
340 Improved activity masking (lower PSNR impact for same SSIM boost)
341 Improved interaction between GF and ARFs
342 Adding error-concealment to the decoder.
343 Adding support for independent partitions
344 Adjusted rate-distortion constants
348 Removed firstpass motion map
349 Fix parallel make install
350 Fix multithreaded encoding for 1 MB wide frame
351 Fixed iwalsh_neon build problems with RVDS4.1
352 Fix semaphore emulation, spin-wait intrinsics on Windows
353 Fix build with xcode4 and simplify GLOBAL.
354 Mark ARM asm objects as allowing a non-executable stack.
355 Fix vpxenc encoding incorrect webm file header on big endian
358 2011-03-07 v0.9.6 "Bali"
359 Our second named release, focused on a faster, higher quality, encoder.
362 This release is backwards compatible with Aylesbury (v0.9.5). Users
363 of older releases should refer to the Upgrading notes in this
364 document for that release.
367 vpxenc --psnr shows a summary when encode completes
368 --tune=ssim option to enable activity masking
369 improved postproc visualizations for development
370 updated support for Apple iOS to SDK 4.2
371 query decoder to determine which reference frames were updated
372 implemented error tracking in the decoder
373 fix pipe support on windows
376 Primary focus was on good quality mode, speed 0. Average improvement
377 on x86 about 40%, up to 100% on user-generated content at that speed.
378 Best quality mode speed improved 35%, and realtime speed 10-20%. This
379 release also saw significant improvement in realtime encoding speed
382 Improved encoder threading
383 Dont pick encoder filter level when loopfilter is disabled.
384 Avoid double copying of key frames into alt and golden buffer
386 x86 sse2 temporal filter
387 SSSE3 version of fast quantizer
388 vp8_rd_pick_best_mbsegmentation code restructure
389 Adjusted breakout RD for SPLITMV
390 Changed segmentation check order
391 Improved rd_pick_intra4x4block
392 Adds armv6 optimized variance calculation
393 ARMv6 optimized sad16x16
394 ARMv6 optimized half pixel variance calculations
395 Full search SAD function optimization in SSE4.1
396 Improve MV prediction accuracy to achieve performance gain
397 Improve MV prediction in vp8_pick_inter_mode() for speed>3
400 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
401 also includes support for "activity masking," which greatly improves
402 SSIM at the expense of PSNR. For now, this feature is available with
403 the --tune=ssim option. Further experimentation in this area
404 is ongoing. This release also introduces a new rate control mode
405 called "CQ," which changes the allocation of bits within a clip to
406 the sections where they will have the most visual impact.
408 Tuning for the more exact quantizer.
409 Relax rate control for last few frames
411 Limit key frame quantizer for forced key frames.
413 Add simple version of activity masking.
414 make rdmult adaptive for intra in quantizer RDO
415 cap the best quantizer for 2nd order DC
416 change the threshold of DC check for encode breakout
419 Fix crash on Sparc Solaris.
420 Fix counter of fixed keyframe distance
421 ARNR filter pointer update bug fix
422 Fixed use of motion percentage in KF/GF group calc
423 Changed condition for using RD in Intra Mode
424 Fix encoder real-time only configuration.
425 Fix ARM encoder crash with multiple token partitions
426 Fixed bug first cluster timecode of webm file is wrong.
427 Fixed various encoder bugs with odd-sized images
428 vp8e_get_preview fixed when spatial resampling enabled
429 quantizer: fix assertion in fast quantizer path
430 Allocate source buffers to be multiples of 16
431 Fix for manual Golden frame frequency
432 Fix drastic undershoot in long form content
435 2010-10-28 v0.9.5 "Aylesbury"
436 Our first named release, focused on a faster decoder, and a better encoder.
439 This release incorporates backwards-incompatible changes to the
440 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
443 * the -q (quiet) option has been removed, and replaced with
444 -v (verbose). the output is quiet by default. Use -v to see
445 the version number of the binary.
447 * The default behavior is now to write output to a single file
448 instead of individual frames. The -y option has been removed.
449 Y4M output is the default.
451 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
452 options must be specified.
454 $ ivfdec -o OUTPUT INPUT
455 $ vpxdec --i420 -o OUTPUT INPUT
457 * If an output file is not specified, the default is to write
458 Y4M to stdout. This makes piping more natural.
460 $ ivfdec -y -o - INPUT | ...
463 * The output file has additional flexibility for formatting the
464 filename. It supports escape characters for constructing a
465 filename from the width, height, and sequence number. This
466 replaces the -p option. To get the equivalent:
468 $ ivfdec -p frame INPUT
469 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
472 * The output file must be specified with -o, rather than as the
475 $ ivfenc <options> INPUT OUTPUT
476 $ vpxenc <options> -o OUTPUT INPUT
478 * The output defaults to webm. To get IVF output, use the --ivf
481 $ ivfenc <options> INPUT OUTPUT.ivf
482 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
486 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
487 vpxdec supports .webm input
488 vpxdec writes .y4m by default
489 vpxenc writes .webm output by default
490 vpxenc --psnr now shows the average/overall PSNR at the end
491 ARM platforms now support runtime cpu detection
492 vpxdec visualizations added for motion vectors, block modes, references
493 vpxdec now silent by default
494 vpxdec --progress shows frame-by-frame timing information
495 vpxenc supports the distinction between --fps and --timebase
496 NASM is now a supported assembler
497 configure: enable PIC for shared libs by default
498 configure: add --enable-small
499 configure: support for ppc32-linux-gcc
500 configure: support for sparc-solaris-gcc
503 Improve handling of invalid frames
504 Fix valgrind errors in the NEON loop filters.
505 Fix loopfilter delta zero transitions
506 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
507 Build fixes for darwin-icc
510 20-40% (average 28%) improvement in libvpx decoder speed,
512 Rewrite vp8_short_walsh4x4_sse2()
513 Optimizations on the loopfilters.
514 Miscellaneous improvements for Atom
515 Add 4-tap version of 2nd-pass ARMv6 MC filter.
516 Improved multithread utilization
517 Better instruction choices on x86
518 reorder data to use wider instructions
519 Update NEON wide idcts
520 Make block access to frame buffer sequential
521 Improved subset block search
522 Bilinear subpixel optimizations for ssse3.
523 Decrease memory footprint
525 Encoder speed improvements (percentage gain not measured):
526 Skip unnecessary search of identical frames
527 Add SSE2 subtract functions
528 Improve bounds checking in vp8_diamond_search_sadx4()
529 Added vp8_fast_quantize_b_sse2
532 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
533 encoding mode, and up to 60% improvement on very noisy, still
534 or slow moving source video
536 Motion compensated temporal filter for Alt-Ref Noise Reduction
537 Improved use of trellis quantization on 2nd order Y blocks
538 Tune effect of motion on KF/GF boost in two pass
539 Allow coefficient optimization for good quality speed 0.
540 Improved control of active min quantizer for two pass.
541 Enable ARFs for non-lagged compress
545 Disable frame dropping by default
546 Improved multithreaded performance
547 Improved Force Key Frame Behaviour
548 Increased rate control buffer level precision
549 Fix bug in 1st pass motion compensation
550 ivfenc: correct fixed kf interval, --disable-kf
552 Changed above and left context data layout
553 Rework idct calling structure.
554 Removed unnecessary MB_MODE_INFO copies
555 x86: SSSE3 sixtap prediction
556 Reworked IDCT to include reconstruction (add) step
557 Swap alt/gold/new/last frame buffer ptrs instead of copying.
558 Improve SSE2 loopfilter functions
559 Change bitreader to use a larger window.
560 Avoid loopfilter reinitialization when possible
562 Normalize quantizer's zero bin and rounding factors
563 Add trellis quantization.
564 Make the quantizer exact.
565 Updates to ARNR filtering algorithm
566 Fix breakout thresh computation for golden & AltRef frames
567 Redo the forward 4x4 dct
568 Improve the accuracy of forward walsh-hadamard transform
569 Further adjustment of RD behaviour with Q and Zbin.
571 Allow linking of libs built with MinGW to MSVC
572 Fix target auto-detection on mingw32
573 Allow --cpu= to work for x86.
574 configure: pass original arguments through to make dist
575 Fix builds without runtime CPU detection
576 msvs: fix install of codec sources
577 msvs: Change devenv.com command line for better msys support
578 msvs: Add vs9 targets.
579 Add x86_64-linux-icc target
581 Potential crashes on older MinGW builds
582 Fix two-pass framrate for Y4M input.
583 Fixed simple loop filter, other crashes on ARM v6
584 arm: fix missing dependency with --enable-shared
585 configure: support directories containing .o
586 Replace pinsrw (SSE) with MMX instructions
587 apple: include proper mach primatives
588 Fixed rate control bug with long key frame interval.
589 Fix DSO link errors on x86-64 when not using a version script
590 Fixed buffer selection for UV in AltRef filtering
595 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
596 * Speed optimizations
599 * Prevent out-of-bounds accesses on invalid data
600 - Build system updates:
601 * Detect toolchain to be used automatically for native builds
602 * Support building shared libraries
603 * Better autotools emulation (--prefix, --libdir, DESTDIR)
605 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
609 - Initial open source release. Welcome to WebM and VP8!