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