| blob | b0d306442b33610fb6220db4a086f41018d740c8 |
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.
4 Decoder control to toggle on/off loopfilter.
6 2015-04-03 v1.4.0 "Indian Runner Duck"
7 This release includes significant improvements to the VP9 codec.
9 - Upgrading:
10 This release is ABI incompatible with 1.3.0. It drops the compatibility
11 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
12 controls for VP9.
14 - Enhancements:
15 Faster VP9 encoding and decoding
16 Multithreaded VP9 decoding (tile and frame-based)
17 Multithreaded VP9 encoding - on by default
18 YUV 4:2:2 and 4:4:4 support in VP9
19 10 and 12bit support in VP9
20 64bit ARM support by replacing ARM assembly with intrinsics
22 - Bug Fixes:
23 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
24 files.
26 - Known Issues:
27 Frame Parallel decoding fails for segmented and non-420 files.
29 2013-11-15 v1.3.0 "Forest"
30 This release introduces the VP9 codec in a backward-compatible way.
31 All existing users of VP8 can continue to use the library without
32 modification. However, some VP8 options do not map to VP9 in the same manner.
34 The VP9 encoder in this release is not feature complete. Users interested in
35 the encoder are advised to use the git master branch and discuss issues on
36 libvpx mailing lists.
38 - Upgrading:
39 This release is ABI and API compatible with Duclair (v1.0.0). Users
40 of older releases should refer to the Upgrading notes in this document
41 for that release.
43 - Enhancements:
44 Get rid of bashisms in the main build scripts
45 Added usage info on command line options
46 Add lossless compression mode
47 Dll build of libvpx
48 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
49 Add option to disable documentation
50 configure: add --enable-external-build support
51 make: support V=1 as short form of verbose=yes
52 configure: support mingw-w64
53 configure: support hardfloat armv7 CHOSTS
54 configure: add support for android x86
55 Add estimated completion time to vpxenc
56 Don't exit on decode errors in vpxenc
57 vpxenc: support scaling prior to encoding
58 vpxdec: support scaling output
59 vpxenc: improve progress indicators with --skip
60 msvs: Don't link to winmm.lib
61 Add a new script for producing vcxproj files
62 Produce Visual Studio 10 and 11 project files
63 Produce Windows Phone project files
64 msvs-build: use msbuild for vs >= 2005
65 configure: default configure log to config.log
66 Add encoding option --static-thresh
68 - Speed:
69 Miscellaneous speed optimizations for VP8 and VP9.
71 - Quality:
72 In general, quality is consistent with the Eider release.
74 - Bug Fixes:
75 This release represents approximately a year of engineering effort,
76 and contains multiple bug fixes. Please refer to git history for details.
79 2012-12-21 v1.2.0
80 This release acts as a checkpoint for a large amount of internal refactoring
81 and testing. It also contains a number of small bugfixes, so all users are
82 encouraged to upgrade.
84 - Upgrading:
85 This release is ABI and API compatible with Duclair (v1.0.0). Users
86 of older releases should refer to the Upgrading notes in this
87 document for that release.
89 - Enhancements:
90 VP8 optimizations for MIPS dspr2
91 vpxenc: add -quiet option
93 - Speed:
94 Encoder and decoder speed is consistent with the Eider release.
96 - Quality:
97 In general, quality is consistent with the Eider release.
99 Minor tweaks to ARNR filtering
100 Minor improvements to real time encoding with multiple temporal layers
102 - Bug Fixes:
103 Fixes multithreaded encoder race condition in loopfilter
104 Fixes multi-resolution threaded encoding
105 Fix potential encoder dead-lock after picture resize
108 2012-05-09 v1.1.0 "Eider"
109 This introduces a number of enhancements, mostly focused on real-time
110 encoding. In addition, it fixes a decoder bug (first introduced in
111 Duclair) so all users of that release are encouraged to upgrade.
113 - Upgrading:
114 This release is ABI and API compatible with Duclair (v1.0.0). Users
115 of older releases should refer to the Upgrading notes in this
116 document for that release.
118 This release introduces a new temporal denoiser, controlled by the
119 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
120 currently take a strength parameter, so the control is effectively
121 a boolean - zero (off) or non-zero (on). For compatibility with
122 existing applications, the values accepted are the same as those
123 for the spatial denoiser (0-6). The temporal denoiser is enabled
124 by default, and the older spatial denoiser may be restored by
125 configuring with --disable-temporal-denoising. The temporal denoiser
126 is more computationally intensive than the spatial one.
128 This release removes support for a legacy, decode only API that was
129 supported, but deprecated, at the initial release of libvpx
130 (v0.9.0). This is not expected to have any impact. If you are
131 impacted, you can apply a reversion to commit 2bf8fb58 locally.
132 Please update to the latest libvpx API if you are affected.
134 - Enhancements:
135 Adds a motion compensated temporal denoiser to the encoder, which
136 gives higher quality than the older spatial denoiser. (See above
137 for notes on upgrading).
139 In addition, support for new compilers and platforms were added,
140 including:
141 improved support for XCode
142 Android x86 NDK build
143 OS/2 support
144 SunCC support
146 Changing resolution with vpx_codec_enc_config_set() is now
147 supported. Previously, reinitializing the codec was required to
148 change the input resolution.
150 The vpxenc application has initial support for producing multiple
151 encodes from the same input in one call. Resizing is not yet
152 supported, but varying other codec parameters is. Use -- to
153 delineate output streams. Options persist from one stream to the
154 next.
156 Also, the vpxenc application will now use a keyframe interval of
157 5 seconds by default. Use the --kf-max-dist option to override.
159 - Speed:
160 Decoder performance improved 2.5% versus Duclair. Encoder speed is
161 consistent with Duclair for most material. Two pass encoding of
162 slideshow-like material will see significant improvements.
164 Large realtime encoding speed gains at a small quality expense are
165 possible by configuring the on-the-fly bitpacking experiment with
166 --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
167 faster (ARM) depending on the number of threads and bitrate
168 settings. This technique sees constant gain over the 5-16 speed
169 range. For VC style input the loss seen is up to 0.2dB. See commit
170 52cf4dca for further details.
172 - Quality:
173 On the whole, quality is consistent with the Duclair release. Some
174 tweaks:
176 Reduced blockiness in easy sections by applying a penalty to
177 intra modes.
179 Improved quality of static sections (like slideshows) with
180 two pass encoding.
182 Improved keyframe sizing with multiple temporal layers
184 - Bug Fixes:
185 Corrected alt-ref contribution to frame rate for visible updates
186 to the alt-ref buffer. This affected applications making manual
187 usage of the frame reference flags, or temporal layers.
189 Additional constraints were added to disable multi-frame quality
190 enhancement (MFQE) in sections of the frame where there is motion.
191 (#392)
193 Fixed corruption issues when vpx_codec_enc_config_set() was called
194 with spatial resampling enabled.
196 Fixed a decoder error introduced in Duclair where the segmentation
197 map was not being reinitialized on keyframes (#378)
200 2012-01-27 v1.0.0 "Duclair"
201 Our fourth named release, focused on performance and features related to
202 real-time encoding. It also fixes a decoder crash bug introduced in
203 v0.9.7, so all users of that release are encouraged to upgrade.
205 - Upgrading:
206 This release is ABI incompatible with prior releases of libvpx, so the
207 "major" version number has been bumped to 1. You must recompile your
208 applications against the latest version of the libvpx headers. The
209 API remains compatible, and this should not require code changes in most
210 applications.
212 - Enhancements:
213 This release introduces several substantial new features to the encoder,
214 of particular interest to real time streaming applications.
216 Temporal scalability allows the encoder to produce a stream that can
217 be decimated to different frame rates, with independent rate targetting
218 for each substream.
220 Multiframe quality enhancement postprocessing can make visual quality
221 more consistent in the presence of frames that are substantially
222 different quality than the surrounding frames, as in the temporal
223 scalability case and in some forced keyframe scenarios.
225 Multiple-resolution encoding support allows the encoding of the
226 same content at different resolutions faster than encoding them
227 separately.
229 - Speed:
230 Optimization targets for this release included the decoder and the real-
231 time modes of the encoder. Decoder speed on x86 has improved 10.5% with
232 this release. Encoder improvements followed a curve where speeds 1-3
233 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
234 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
235 Cayuga release.
237 - Quality:
238 Encoder quality in the single stream case is consistent with the Cayuga
239 release.
241 - Bug Fixes:
242 This release fixes an OOB read decoder crash bug present in v0.9.7
243 related to the clamping of motion vectors in SPLITMV blocks. This
244 behavior could be triggered by corrupt input or by starting
245 decoding from a P-frame.
248 2011-08-15 v0.9.7-p1 "Cayuga" patch 1
249 This is an incremental bugfix release against Cayuga. All users of that
250 release are strongly encouraged to upgrade.
252 - Fix potential OOB reads (cdae03a)
254 An unbounded out of bounds read was discovered when the
255 decoder was requested to perform error concealment (new in
256 Cayuga) given a frame with corrupt partition sizes.
258 A bounded out of bounds read was discovered affecting all
259 versions of libvpx. Given an multipartition input frame that
260 is truncated between the mode/mv partition and the first
261 residiual paritition (in the block of partition offsets), up
262 to 3 extra bytes could have been read from the source buffer.
263 The code will not take any action regardless of the contents
264 of these undefined bytes, as the truncated buffer is detected
265 immediately following the read based on the calculated
266 starting position of the coefficient partition.
268 - Fix potential error concealment crash when the very first frame
269 is missing or corrupt (a609be5)
271 - Fix significant artifacts in error concealment (a4c2211, 99d870a)
273 - Revert 1-pass CBR rate control changes (e961317)
274 Further testing showed this change produced undesirable visual
275 artifacts, rolling back for now.
278 2011-08-02 v0.9.7 "Cayuga"
279 Our third named release, focused on a faster, higher quality, encoder.
281 - Upgrading:
282 This release is backwards compatible with Aylesbury (v0.9.5) and
283 Bali (v0.9.6). Users of older releases should refer to the Upgrading
284 notes in this document for that release.
286 - Enhancements:
287 Stereo 3D format support for vpxenc
288 Runtime detection of available processor cores.
289 Allow specifying --end-usage by enum name
290 vpxdec: test for frame corruption
291 vpxenc: add quantizer histogram display
292 vpxenc: add rate histogram display
293 Set VPX_FRAME_IS_DROPPABLE
294 update configure for ios sdk 4.3
295 Avoid text relocations in ARM vp8 decoder
296 Generate a vpx.pc file for pkg-config.
297 New ways of passing encoded data between encoder and decoder.
299 - Speed:
300 This release includes across-the-board speed improvements to the
301 encoder. On x86, these measure at approximately 11.5% in Best mode,
302 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
303 On ARM Cortex A9 with Neon extensions, real-time encoding of video
304 telephony content is 35% faster than Bali on single core and 48%
305 faster on multi-core. On the NVidia Tegra2 platform, real time
306 encoding is 40% faster than Bali.
308 Decoder speed was not a priority for this release, but improved
309 approximately 8.4% on x86.
311 Reduce motion vector search on alt-ref frame.
312 Encoder loopfilter running in its own thread
313 Reworked loopfilter to precalculate more parameters
314 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
315 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
316 Removed redundant checks
317 Reduced structure sizes
318 utilize preload in ARMv6 MC/LPF/Copy routines
319 ARM optimized quantization, dfct, variance, subtract
320 Increase chrow row alignment to 16 bytes.
321 disable trellis optimization for first pass
322 Write SSSE3 sub-pixel filter function
323 Improve SSE2 half-pixel filter funtions
324 Add vp8_sub_pixel_variance16x8_ssse3 function
325 Reduce unnecessary distortion computation
326 Use diamond search to replace full search
327 Preload reference area in sub-pixel motion search (real-time mode)
329 - Quality:
330 This release focused primarily on one-pass use cases, including
331 video conferencing. Low latency data rate control was significantly
332 improved, improving streamability over bandwidth constrained links.
333 Added support for error concealment, allowing frames to maintain
334 visual quality in the presence of substantial packet loss.
336 Add rc_max_intra_bitrate_pct control
337 Limit size of initial keyframe in one-pass.
338 Improve framerate adaptation
339 Improved 1-pass CBR rate control
340 Improved KF insertion after fades to still.
341 Improved key frame detection.
342 Improved activity masking (lower PSNR impact for same SSIM boost)
343 Improved interaction between GF and ARFs
344 Adding error-concealment to the decoder.
345 Adding support for independent partitions
346 Adjusted rate-distortion constants
349 - Bug Fixes:
350 Removed firstpass motion map
351 Fix parallel make install
352 Fix multithreaded encoding for 1 MB wide frame
353 Fixed iwalsh_neon build problems with RVDS4.1
354 Fix semaphore emulation, spin-wait intrinsics on Windows
355 Fix build with xcode4 and simplify GLOBAL.
356 Mark ARM asm objects as allowing a non-executable stack.
357 Fix vpxenc encoding incorrect webm file header on big endian
360 2011-03-07 v0.9.6 "Bali"
361 Our second named release, focused on a faster, higher quality, encoder.
363 - Upgrading:
364 This release is backwards compatible with Aylesbury (v0.9.5). Users
365 of older releases should refer to the Upgrading notes in this
366 document for that release.
368 - Enhancements:
369 vpxenc --psnr shows a summary when encode completes
370 --tune=ssim option to enable activity masking
371 improved postproc visualizations for development
372 updated support for Apple iOS to SDK 4.2
373 query decoder to determine which reference frames were updated
374 implemented error tracking in the decoder
375 fix pipe support on windows
377 - Speed:
378 Primary focus was on good quality mode, speed 0. Average improvement
379 on x86 about 40%, up to 100% on user-generated content at that speed.
380 Best quality mode speed improved 35%, and realtime speed 10-20%. This
381 release also saw significant improvement in realtime encoding speed
382 on ARM platforms.
384 Improved encoder threading
385 Dont pick encoder filter level when loopfilter is disabled.
386 Avoid double copying of key frames into alt and golden buffer
387 FDCT optimizations.
388 x86 sse2 temporal filter
389 SSSE3 version of fast quantizer
390 vp8_rd_pick_best_mbsegmentation code restructure
391 Adjusted breakout RD for SPLITMV
392 Changed segmentation check order
393 Improved rd_pick_intra4x4block
394 Adds armv6 optimized variance calculation
395 ARMv6 optimized sad16x16
396 ARMv6 optimized half pixel variance calculations
397 Full search SAD function optimization in SSE4.1
398 Improve MV prediction accuracy to achieve performance gain
399 Improve MV prediction in vp8_pick_inter_mode() for speed>3
401 - Quality:
402 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
403 also includes support for "activity masking," which greatly improves
404 SSIM at the expense of PSNR. For now, this feature is available with
405 the --tune=ssim option. Further experimentation in this area
406 is ongoing. This release also introduces a new rate control mode
407 called "CQ," which changes the allocation of bits within a clip to
408 the sections where they will have the most visual impact.
410 Tuning for the more exact quantizer.
411 Relax rate control for last few frames
412 CQ Mode
413 Limit key frame quantizer for forced key frames.
414 KF/GF Pulsing
415 Add simple version of activity masking.
416 make rdmult adaptive for intra in quantizer RDO
417 cap the best quantizer for 2nd order DC
418 change the threshold of DC check for encode breakout
420 - Bug Fixes:
421 Fix crash on Sparc Solaris.
422 Fix counter of fixed keyframe distance
423 ARNR filter pointer update bug fix
424 Fixed use of motion percentage in KF/GF group calc
425 Changed condition for using RD in Intra Mode
426 Fix encoder real-time only configuration.
427 Fix ARM encoder crash with multiple token partitions
428 Fixed bug first cluster timecode of webm file is wrong.
429 Fixed various encoder bugs with odd-sized images
430 vp8e_get_preview fixed when spatial resampling enabled
431 quantizer: fix assertion in fast quantizer path
432 Allocate source buffers to be multiples of 16
433 Fix for manual Golden frame frequency
434 Fix drastic undershoot in long form content
437 2010-10-28 v0.9.5 "Aylesbury"
438 Our first named release, focused on a faster decoder, and a better encoder.
440 - Upgrading:
441 This release incorporates backwards-incompatible changes to the
442 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
444 vpxdec
445 * the -q (quiet) option has been removed, and replaced with
446 -v (verbose). the output is quiet by default. Use -v to see
447 the version number of the binary.
449 * The default behavior is now to write output to a single file
450 instead of individual frames. The -y option has been removed.
451 Y4M output is the default.
453 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
454 options must be specified.
456 $ ivfdec -o OUTPUT INPUT
457 $ vpxdec --i420 -o OUTPUT INPUT
459 * If an output file is not specified, the default is to write
460 Y4M to stdout. This makes piping more natural.
462 $ ivfdec -y -o - INPUT | ...
463 $ vpxdec INPUT | ...
465 * The output file has additional flexibility for formatting the
466 filename. It supports escape characters for constructing a
467 filename from the width, height, and sequence number. This
468 replaces the -p option. To get the equivalent:
470 $ ivfdec -p frame INPUT
471 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
473 vpxenc
474 * The output file must be specified with -o, rather than as the
475 last argument.
477 $ ivfenc <options> INPUT OUTPUT
478 $ vpxenc <options> -o OUTPUT INPUT
480 * The output defaults to webm. To get IVF output, use the --ivf
481 option.
483 $ ivfenc <options> INPUT OUTPUT.ivf
484 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
487 - Enhancements:
488 ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
489 vpxdec supports .webm input
490 vpxdec writes .y4m by default
491 vpxenc writes .webm output by default
492 vpxenc --psnr now shows the average/overall PSNR at the end
493 ARM platforms now support runtime cpu detection
494 vpxdec visualizations added for motion vectors, block modes, references
495 vpxdec now silent by default
496 vpxdec --progress shows frame-by-frame timing information
497 vpxenc supports the distinction between --fps and --timebase
498 NASM is now a supported assembler
499 configure: enable PIC for shared libs by default
500 configure: add --enable-small
501 configure: support for ppc32-linux-gcc
502 configure: support for sparc-solaris-gcc
504 - Bugs:
505 Improve handling of invalid frames
506 Fix valgrind errors in the NEON loop filters.
507 Fix loopfilter delta zero transitions
508 Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
509 Build fixes for darwin-icc
511 - Speed:
512 20-40% (average 28%) improvement in libvpx decoder speed,
513 including:
514 Rewrite vp8_short_walsh4x4_sse2()
515 Optimizations on the loopfilters.
516 Miscellaneous improvements for Atom
517 Add 4-tap version of 2nd-pass ARMv6 MC filter.
518 Improved multithread utilization
519 Better instruction choices on x86
520 reorder data to use wider instructions
521 Update NEON wide idcts
522 Make block access to frame buffer sequential
523 Improved subset block search
524 Bilinear subpixel optimizations for ssse3.
525 Decrease memory footprint
527 Encoder speed improvements (percentage gain not measured):
528 Skip unnecessary search of identical frames
529 Add SSE2 subtract functions
530 Improve bounds checking in vp8_diamond_search_sadx4()
531 Added vp8_fast_quantize_b_sse2
533 - Quality:
534 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
535 encoding mode, and up to 60% improvement on very noisy, still
536 or slow moving source video
538 Motion compensated temporal filter for Alt-Ref Noise Reduction
539 Improved use of trellis quantization on 2nd order Y blocks
540 Tune effect of motion on KF/GF boost in two pass
541 Allow coefficient optimization for good quality speed 0.
542 Improved control of active min quantizer for two pass.
543 Enable ARFs for non-lagged compress
545 2010-09-02 v0.9.2
546 - Enhancements:
547 Disable frame dropping by default
548 Improved multithreaded performance
549 Improved Force Key Frame Behaviour
550 Increased rate control buffer level precision
551 Fix bug in 1st pass motion compensation
552 ivfenc: correct fixed kf interval, --disable-kf
553 - Speed:
554 Changed above and left context data layout
555 Rework idct calling structure.
556 Removed unnecessary MB_MODE_INFO copies
557 x86: SSSE3 sixtap prediction
558 Reworked IDCT to include reconstruction (add) step
559 Swap alt/gold/new/last frame buffer ptrs instead of copying.
560 Improve SSE2 loopfilter functions
561 Change bitreader to use a larger window.
562 Avoid loopfilter reinitialization when possible
563 - Quality:
564 Normalize quantizer's zero bin and rounding factors
565 Add trellis quantization.
566 Make the quantizer exact.
567 Updates to ARNR filtering algorithm
568 Fix breakout thresh computation for golden & AltRef frames
569 Redo the forward 4x4 dct
570 Improve the accuracy of forward walsh-hadamard transform
571 Further adjustment of RD behaviour with Q and Zbin.
572 - Build System:
573 Allow linking of libs built with MinGW to MSVC
574 Fix target auto-detection on mingw32
575 Allow --cpu= to work for x86.
576 configure: pass original arguments through to make dist
577 Fix builds without runtime CPU detection
578 msvs: fix install of codec sources
579 msvs: Change devenv.com command line for better msys support
580 msvs: Add vs9 targets.
581 Add x86_64-linux-icc target
582 - Bugs:
583 Potential crashes on older MinGW builds
584 Fix two-pass framrate for Y4M input.
585 Fixed simple loop filter, other crashes on ARM v6
586 arm: fix missing dependency with --enable-shared
587 configure: support directories containing .o
588 Replace pinsrw (SSE) with MMX instructions
589 apple: include proper mach primatives
590 Fixed rate control bug with long key frame interval.
591 Fix DSO link errors on x86-64 when not using a version script
592 Fixed buffer selection for UV in AltRef filtering
595 2010-06-17 v0.9.1
596 - Enhancements:
597 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
598 * Speed optimizations
599 - Bugfixes:
600 * Rate control
601 * Prevent out-of-bounds accesses on invalid data
602 - Build system updates:
603 * Detect toolchain to be used automatically for native builds
604 * Support building shared libraries
605 * Better autotools emulation (--prefix, --libdir, DESTDIR)
606 - Updated LICENSE
607 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
610 2010-05-18 v0.9.0
611 - Initial open source release. Welcome to WebM and VP8!