12023-01-31 v1.13.0 "Ugly Duckling" 2 This release includes more Neon and AVX2 optimizations, adds a new codec 3 control to set per frame QP, upgrades GoogleTest to v1.12.1, and includes 4 numerous bug fixes. 5 6 - Upgrading: 7 This release is ABI incompatible with the previous release. 8 9 New codec control VP9E_SET_QUANTIZER_ONE_PASS to set per frame QP. 10 11 GoogleTest is upgraded to v1.12.1. 12 13 .clang-format is upgraded to clang-format-11. 14 15 VPX_EXT_RATECTRL_ABI_VERSION was bumped due to incompatible changes to the 16 feature of using external rate control models for vp9. 17 18 - Enhancement: 19 Numerous improvements on Neon optimizations. 20 Numerous improvements on AVX2 optimizations. 21 Additional ARM targets added for Visual Studio. 22 23 - Bug fixes: 24 Fix to calculating internal stats when frame dropped. 25 Fix to segfault for external resize test in vp9. 26 Fix to build system with replacing egrep with grep -E. 27 Fix to a few bugs with external RTC rate control library. 28 Fix to make SVC work with VBR. 29 Fix to key frame setting in VP9 external RC. 30 Fix to -Wimplicit-int (Clang 16). 31 Fix to VP8 external RC for buffer levels. 32 Fix to VP8 external RC for dynamic update of layers. 33 Fix to VP9 auto level. 34 Fix to off-by-one error of max w/h in validate_config. 35 Fix to make SVC work for Profile 1. 36 372022-06-17 v1.12.0 "Torrent Duck" 38 This release adds optimizations for Loongarch, adds support for vp8 in the 39 real-time rate control library, upgrades GoogleTest to v1.11.0, updates 40 libwebm to libwebm-1.0.0.28-20-g206d268, and includes numerous bug fixes. 41 42 - Upgrading: 43 This release is ABI compatible with the previous release. 44 45 vp8 support in the real-time rate control library. 46 New codec control VP8E_SET_RTC_EXTERNAL_RATECTRL is added. 47 48 Configure support for darwin21 is added. 49 50 GoogleTest is upgraded to v1.11.0. 51 52 libwebm is updated to libwebm-1.0.0.28-20-g206d268. 53 54 Allow SimpleEncode environment to take target level as input to match 55 the level conformance in vp9. 56 57 - Enhancement: 58 Numerous improvements on checking memory allocations. 59 Optimizations for Loongarch. 60 Code clean-up. 61 62 - Bug fixes: 63 Fix to a crash related to {vp8/vp9}_set_roi_map. 64 Fix to compiling failure with -Wformat-nonliteral. 65 Fix to integer overflow with vp9 with high resolution content. 66 Fix to AddNoiseTest failure with ARMv7. 67 Fix to libvpx Null-dereference READ in vp8. 68 692021-09-27 v1.11.0 "Smew Duck" 70 This maintenance release adds support for VBR mode in VP9 rate control 71 interface, new codec controls to get quantization parameters and loop filter 72 levels, and includes several improvements to NEON and numerous bug fixes. 73 74 - Upgrading: 75 This release is ABI incompatible with the previous release. 76 New codec control is added to get quantization parameters and loop filter 77 levels. 78 79 VBR mode is supported in VP9 rate control library. 80 81 - Enhancement: 82 Numerous improvements for Neon optimizations. 83 Code clean-up and refactoring. 84 Calculation of rd multiplier is changed with BDRATE gains. 85 86 - Bug fixes: 87 Fix to overflow on duration. 88 Fix to several instances of -Wunused-but-set-variable. 89 Fix to avoid chroma resampling for 420mpeg2 input. 90 Fix to overflow in calc_iframe_target_size. 91 Fix to disallow skipping transform and quantization. 92 Fix some -Wsign-compare warnings in simple_encode. 93 Fix input file path in simple_encode_test. 94 Fix valid range for under/over_shoot pct. 95 962021-03-09 v1.10.0 "Ruddy Duck" 97 This maintenance release adds support for darwin20 and new codec controls, as 98 well as numerous bug fixes. 99 100 - Upgrading: 101 This release is ABI incompatible with the previous release. 102 New codec control is added to disable loopfilter for VP9. 103 104 New encoder control is added to disable feature to increase Q on overshoot 105 detection for CBR. 106 107 Configure support for darwin20 is added. 108 109 New codec control is added for VP9 rate control. The control ID of this 110 interface is VP9E_SET_EXTERNAL_RATE_CONTROL. To make VP9 use a customized 111 external rate control model, users will have to implement each callback 112 function in vpx_rc_funcs_t and register them using libvpx API 113 vpx_codec_control_() with the control ID. 114 115 - Enhancement: 116 Use -std=gnu++11 instead of -std=c++11 for c++ files. 117 118 - Bug fixes: 119 Override assembler with --as option of configure for MSVS. 120 Fix several compilation issues with gcc 4.8.5. 121 Fix to resetting rate control for temporal layers. 122 Fix to the rate control stats of SVC example encoder when number of spatial 123 layers is 1. 124 Fix to reusing motion vectors from the base spatial layer in SVC. 125 2 pass related flags removed from SVC example encoder. 126 1272020-07-29 v1.9.0 "Quacking Duck" 128 This release adds support for NV12, a separate library for rate control, as 129 well as incremental improvements. 130 131 - Upgrading: 132 This release is ABI compatible with the previous release. 133 NV12 support is added to this release. 134 A new interface is added for VP9 rate control. The new library libvp9rc.a 135 must be linked by applications. 136 Googletest is updated to v1.10.0. 137 simple_encode.cc is compiled into a new library libsimple_encode.a with 138 CONFIG_RATE_CTRL. 139 140 - Enhancement: 141 Various changes to improve VP9 SVC, rate control, quality and speed to real 142 time encoding. 143 144 - Bug fixes: 145 Fix key frame update refresh simulcast flexible svc. 146 Fix to disable_16x16part speed feature for real time encoding. 147 Fix some signed integer overflows for VP9 rate control. 148 Fix initialization of delta_q_uv. 149 Fix condition in regulate_q for cyclic refresh. 150 Various fixes to dynamic resizing for VP9 SVC. 151 1522019-12-09 v1.8.2 "Pekin Duck" 153 This release collects incremental improvements to many aspects of the library. 154 155 - Upgrading: 156 This release is ABI compatible with the previous release. 157 ARCH_* defines have been removed in favor of VPX_ARCH_*. 158 1592019-07-15 v1.8.1 "Orpington Duck" 160 This release collects incremental improvements to many aspects of the library. 161 162 - Upgrading: 163 This release is ABI incompatible with the previous release. 164 VP8E_SET_CPUUSED now accepts values up to 9 for vp9. 165 VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT had a spelling fix (was VP8E). 166 The --sdk-path option has been removed. If you were using it to build for 167 Android please read build/make/Android.mk for alternatives. 168 All PPC optimizations have been disabled: 169 https://bugs.chromium.org/p/webm/issues/detail?id=1522. 170 171 - Enhancements: 172 Various changes to improve encoder rate control, quality and speed 173 for practically every use case. 174 175 - Bug fixes: 176 vp9-rtc: Fix color artifacts for speed >= 8. 177 1782019-01-31 v1.8.0 "Northern Shoveler Duck" 179 This release focused on encoding performance for realtime and VOD use cases. 180 181 - Upgrading: 182 This release is ABI incompatible with the previous release. This adds and 183 improves several vp9 controls. Most are related to SVC: 184 VP9E_SET_SVC_FRAME_DROP_LAYER: 185 - Frame dropping in SVC. 186 VP9E_SET_SVC_INTER_LAYER_PRED: 187 - Inter-layer prediction in SVC. 188 VP9E_SET_SVC_GF_TEMPORAL_REF: 189 - Enable long term temporal reference in SVC. 190 VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG: 191 - Extend and improve this control for better flexibility in setting SVC 192 pattern dynamically. 193 VP9E_SET_POSTENCODE_DROP: 194 - Allow for post-encode frame dropping (applies to non-SVC too). 195 VP9E_SET_SVC_SPATIAL_LAYER_SYNC: 196 - Enable spatial layer sync frames. 197 VP9E_SET_SVC_LAYER_ID: 198 - Extend api to specify temporal id for each spatial layers. 199 VP9E_SET_ROI_MAP: 200 - Extend Region of Interest functionality to VP9. 201 202 - Enhancements: 203 2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6, 204 we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1, 205 the gains are approximately 4% for VBR and 5% for CQ. 206 207 For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at 208 screen sharing have improved when the content changes significantly (slide 209 sharing) or scrolls. There is a new speed 9 setting for mobile devices which 210 is about 10-20% faster than speed 8. 211 212 - Bug fixes: 213 VP9 denoiser issue. 214 VP9 partition issue for 1080p. 215 VP9 rate control improvments. 216 Postprocessing Multi Frame Quality Enhancement (MFQE) issue. 217 VP8 multithread decoder issues. 218 A variety of fuzzing issues. 219 2202018-01-04 v1.7.0 "Mandarin Duck" 221 This release focused on high bit depth performance (10/12 bit) and vp9 222 encoding improvements. 223 224 - Upgrading: 225 This release is ABI incompatible due to new vp9 encoder features. 226 227 Frame parallel decoding for vp9 has been removed. 228 229 - Enhancements: 230 vp9 encoding supports additional threads with --row-mt. This can be greater 231 than the number of tiles. 232 233 Two new vp9 encoder options have been added: 234 --corpus-complexity 235 --tune-content=film 236 237 Additional tooling for respecting the vp9 "level" profiles has been added. 238 239 - Bug fixes: 240 A variety of fuzzing issues. 241 vp8 threading fix for ARM. 242 Codec control VP9_SET_SKIP_LOOP_FILTER fixed. 243 Reject invalid multi resolution configurations. 244 2452017-01-09 v1.6.1 "Long Tailed Duck" 246 This release improves upon the VP9 encoder and speeds up the encoding and 247 decoding processes. 248 249 - Upgrading: 250 This release is ABI compatible with 1.6.0. 251 252 - Enhancements: 253 Faster VP9 encoding and decoding. 254 High bit depth builds now provide similar speed for 8 bit encode and decode 255 for x86 targets. Other platforms and higher bit depth improvements are in 256 progress. 257 258 - Bug Fixes: 259 A variety of fuzzing issues. 260 2612016-07-20 v1.6.0 "Khaki Campbell Duck" 262 This release improves upon the VP9 encoder and speeds up the encoding and 263 decoding processes. 264 265 - Upgrading: 266 This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum 267 in vpx_image and some minor changes to the VP8_COMP structure. 268 269 The default key frame interval for VP9 has changed from 128 to 9999. 270 271 - Enhancement: 272 A core focus has been performance for low end Intel processors. SSSE3 273 instructions such as 'pshufb' have been avoided and instructions have been 274 reordered to better accommodate the more constrained pipelines. 275 276 As a result, devices based on Celeron processors have seen substantial 277 decoding improvements. From Indian Runner Duck to Javan Whistling Duck, 278 decoding speed improved between 10 and 30%. Between Javan Whistling Duck 279 and Khaki Campbell Duck, it improved another 10 to 15%. 280 281 While Celeron benefited most, Core-i5 also improved 5% and 10% between the 282 respective releases. 283 284 Realtime performance for WebRTC for both speed and quality has received a 285 lot of attention. 286 287 - Bug Fixes: 288 A number of fuzzing issues, found variously by Mozilla, Chromium and others, 289 have been fixed and we strongly recommend updating. 290 2912015-11-09 v1.5.0 "Javan Whistling Duck" 292 This release improves upon the VP9 encoder and speeds up the encoding and 293 decoding processes. 294 295 - Upgrading: 296 This release is ABI incompatible with 1.4.0. It drops deprecated VP8 297 controls and adds a variety of VP9 controls for testing. 298 299 The vpxenc utility now prefers VP9 by default. 300 301 - Enhancements: 302 Faster VP9 encoding and decoding 303 Smaller library size by combining functions used by VP8 and VP9 304 305 - Bug Fixes: 306 A variety of fuzzing issues 307 3082015-04-03 v1.4.0 "Indian Runner Duck" 309 This release includes significant improvements to the VP9 codec. 310 311 - Upgrading: 312 This release is ABI incompatible with 1.3.0. It drops the compatibility 313 layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec 314 controls for VP9. 315 316 - Enhancements: 317 Faster VP9 encoding and decoding 318 Multithreaded VP9 decoding (tile and frame-based) 319 Multithreaded VP9 encoding - on by default 320 YUV 4:2:2 and 4:4:4 support in VP9 321 10 and 12bit support in VP9 322 64bit ARM support by replacing ARM assembly with intrinsics 323 324 - Bug Fixes: 325 Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0 326 files. 327 328 - Known Issues: 329 Frame Parallel decoding fails for segmented and non-420 files. 330 3312013-11-15 v1.3.0 "Forest" 332 This release introduces the VP9 codec in a backward-compatible way. 333 All existing users of VP8 can continue to use the library without 334 modification. However, some VP8 options do not map to VP9 in the same manner. 335 336 The VP9 encoder in this release is not feature complete. Users interested in 337 the encoder are advised to use the git master branch and discuss issues on 338 libvpx mailing lists. 339 340 - Upgrading: 341 This release is ABI and API compatible with Duclair (v1.0.0). Users 342 of older releases should refer to the Upgrading notes in this document 343 for that release. 344 345 - Enhancements: 346 Get rid of bashisms in the main build scripts 347 Added usage info on command line options 348 Add lossless compression mode 349 Dll build of libvpx 350 Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13) 351 Add option to disable documentation 352 configure: add --enable-external-build support 353 make: support V=1 as short form of verbose=yes 354 configure: support mingw-w64 355 configure: support hardfloat armv7 CHOSTS 356 configure: add support for android x86 357 Add estimated completion time to vpxenc 358 Don't exit on decode errors in vpxenc 359 vpxenc: support scaling prior to encoding 360 vpxdec: support scaling output 361 vpxenc: improve progress indicators with --skip 362 msvs: Don't link to winmm.lib 363 Add a new script for producing vcxproj files 364 Produce Visual Studio 10 and 11 project files 365 Produce Windows Phone project files 366 msvs-build: use msbuild for vs >= 2005 367 configure: default configure log to config.log 368 Add encoding option --static-thresh 369 370 - Speed: 371 Miscellaneous speed optimizations for VP8 and VP9. 372 373 - Quality: 374 In general, quality is consistent with the Eider release. 375 376 - Bug Fixes: 377 This release represents approximately a year of engineering effort, 378 and contains multiple bug fixes. Please refer to git history for details. 379 380 3812012-12-21 v1.2.0 382 This release acts as a checkpoint for a large amount of internal refactoring 383 and testing. It also contains a number of small bugfixes, so all users are 384 encouraged to upgrade. 385 386 - Upgrading: 387 This release is ABI and API compatible with Duclair (v1.0.0). Users 388 of older releases should refer to the Upgrading notes in this 389 document for that release. 390 391 - Enhancements: 392 VP8 optimizations for MIPS dspr2 393 vpxenc: add -quiet option 394 395 - Speed: 396 Encoder and decoder speed is consistent with the Eider release. 397 398 - Quality: 399 In general, quality is consistent with the Eider release. 400 401 Minor tweaks to ARNR filtering 402 Minor improvements to real time encoding with multiple temporal layers 403 404 - Bug Fixes: 405 Fixes multithreaded encoder race condition in loopfilter 406 Fixes multi-resolution threaded encoding 407 Fix potential encoder dead-lock after picture resize 408 409 4102012-05-09 v1.1.0 "Eider" 411 This introduces a number of enhancements, mostly focused on real-time 412 encoding. In addition, it fixes a decoder bug (first introduced in 413 Duclair) so all users of that release are encouraged to upgrade. 414 415 - Upgrading: 416 This release is ABI and API compatible with Duclair (v1.0.0). Users 417 of older releases should refer to the Upgrading notes in this 418 document for that release. 419 420 This release introduces a new temporal denoiser, controlled by the 421 VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not 422 currently take a strength parameter, so the control is effectively 423 a boolean - zero (off) or non-zero (on). For compatibility with 424 existing applications, the values accepted are the same as those 425 for the spatial denoiser (0-6). The temporal denoiser is enabled 426 by default, and the older spatial denoiser may be restored by 427 configuring with --disable-temporal-denoising. The temporal denoiser 428 is more computationally intensive than the spatial one. 429 430 This release removes support for a legacy, decode only API that was 431 supported, but deprecated, at the initial release of libvpx 432 (v0.9.0). This is not expected to have any impact. If you are 433 impacted, you can apply a reversion to commit 2bf8fb58 locally. 434 Please update to the latest libvpx API if you are affected. 435 436 - Enhancements: 437 Adds a motion compensated temporal denoiser to the encoder, which 438 gives higher quality than the older spatial denoiser. (See above 439 for notes on upgrading). 440 441 In addition, support for new compilers and platforms were added, 442 including: 443 improved support for XCode 444 Android x86 NDK build 445 OS/2 support 446 SunCC support 447 448 Changing resolution with vpx_codec_enc_config_set() is now 449 supported. Previously, reinitializing the codec was required to 450 change the input resolution. 451 452 The vpxenc application has initial support for producing multiple 453 encodes from the same input in one call. Resizing is not yet 454 supported, but varying other codec parameters is. Use -- to 455 delineate output streams. Options persist from one stream to the 456 next. 457 458 Also, the vpxenc application will now use a keyframe interval of 459 5 seconds by default. Use the --kf-max-dist option to override. 460 461 - Speed: 462 Decoder performance improved 2.5% versus Duclair. Encoder speed is 463 consistent with Duclair for most material. Two pass encoding of 464 slideshow-like material will see significant improvements. 465 466 Large realtime encoding speed gains at a small quality expense are 467 possible by configuring the on-the-fly bitpacking experiment with 468 --enable-onthefly-bitpacking. Realtime encoder can be up to 13% 469 faster (ARM) depending on the number of threads and bitrate 470 settings. This technique sees constant gain over the 5-16 speed 471 range. For VC style input the loss seen is up to 0.2dB. See commit 472 52cf4dca for further details. 473 474 - Quality: 475 On the whole, quality is consistent with the Duclair release. Some 476 tweaks: 477 478 Reduced blockiness in easy sections by applying a penalty to 479 intra modes. 480 481 Improved quality of static sections (like slideshows) with 482 two pass encoding. 483 484 Improved keyframe sizing with multiple temporal layers 485 486 - Bug Fixes: 487 Corrected alt-ref contribution to frame rate for visible updates 488 to the alt-ref buffer. This affected applications making manual 489 usage of the frame reference flags, or temporal layers. 490 491 Additional constraints were added to disable multi-frame quality 492 enhancement (MFQE) in sections of the frame where there is motion. 493 (#392) 494 495 Fixed corruption issues when vpx_codec_enc_config_set() was called 496 with spatial resampling enabled. 497 498 Fixed a decoder error introduced in Duclair where the segmentation 499 map was not being reinitialized on keyframes (#378) 500 501 5022012-01-27 v1.0.0 "Duclair" 503 Our fourth named release, focused on performance and features related to 504 real-time encoding. It also fixes a decoder crash bug introduced in 505 v0.9.7, so all users of that release are encouraged to upgrade. 506 507 - Upgrading: 508 This release is ABI incompatible with prior releases of libvpx, so the 509 "major" version number has been bumped to 1. You must recompile your 510 applications against the latest version of the libvpx headers. The 511 API remains compatible, and this should not require code changes in most 512 applications. 513 514 - Enhancements: 515 This release introduces several substantial new features to the encoder, 516 of particular interest to real time streaming applications. 517 518 Temporal scalability allows the encoder to produce a stream that can 519 be decimated to different frame rates, with independent rate targeting 520 for each substream. 521 522 Multiframe quality enhancement postprocessing can make visual quality 523 more consistent in the presence of frames that are substantially 524 different quality than the surrounding frames, as in the temporal 525 scalability case and in some forced keyframe scenarios. 526 527 Multiple-resolution encoding support allows the encoding of the 528 same content at different resolutions faster than encoding them 529 separately. 530 531 - Speed: 532 Optimization targets for this release included the decoder and the real- 533 time modes of the encoder. Decoder speed on x86 has improved 10.5% with 534 this release. Encoder improvements followed a curve where speeds 1-3 535 improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved 536 1.5% to 10.5%, respectively. "Best" mode speed is consistent with the 537 Cayuga release. 538 539 - Quality: 540 Encoder quality in the single stream case is consistent with the Cayuga 541 release. 542 543 - Bug Fixes: 544 This release fixes an OOB read decoder crash bug present in v0.9.7 545 related to the clamping of motion vectors in SPLITMV blocks. This 546 behavior could be triggered by corrupt input or by starting 547 decoding from a P-frame. 548 549 5502011-08-15 v0.9.7-p1 "Cayuga" patch 1 551 This is an incremental bugfix release against Cayuga. All users of that 552 release are strongly encouraged to upgrade. 553 554 - Fix potential OOB reads (cdae03a) 555 556 An unbounded out of bounds read was discovered when the 557 decoder was requested to perform error concealment (new in 558 Cayuga) given a frame with corrupt partition sizes. 559 560 A bounded out of bounds read was discovered affecting all 561 versions of libvpx. Given an multipartition input frame that 562 is truncated between the mode/mv partition and the first 563 residiual paritition (in the block of partition offsets), up 564 to 3 extra bytes could have been read from the source buffer. 565 The code will not take any action regardless of the contents 566 of these undefined bytes, as the truncated buffer is detected 567 immediately following the read based on the calculated 568 starting position of the coefficient partition. 569 570 - Fix potential error concealment crash when the very first frame 571 is missing or corrupt (a609be5) 572 573 - Fix significant artifacts in error concealment (a4c2211, 99d870a) 574 575 - Revert 1-pass CBR rate control changes (e961317) 576 Further testing showed this change produced undesirable visual 577 artifacts, rolling back for now. 578 579 5802011-08-02 v0.9.7 "Cayuga" 581 Our third named release, focused on a faster, higher quality, encoder. 582 583 - Upgrading: 584 This release is backwards compatible with Aylesbury (v0.9.5) and 585 Bali (v0.9.6). Users of older releases should refer to the Upgrading 586 notes in this document for that release. 587 588 - Enhancements: 589 Stereo 3D format support for vpxenc 590 Runtime detection of available processor cores. 591 Allow specifying --end-usage by enum name 592 vpxdec: test for frame corruption 593 vpxenc: add quantizer histogram display 594 vpxenc: add rate histogram display 595 Set VPX_FRAME_IS_DROPPABLE 596 update configure for ios sdk 4.3 597 Avoid text relocations in ARM vp8 decoder 598 Generate a vpx.pc file for pkg-config. 599 New ways of passing encoded data between encoder and decoder. 600 601 - Speed: 602 This release includes across-the-board speed improvements to the 603 encoder. On x86, these measure at approximately 11.5% in Best mode, 604 21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6). 605 On ARM Cortex A9 with Neon extensions, real-time encoding of video 606 telephony content is 35% faster than Bali on single core and 48% 607 faster on multi-core. On the NVidia Tegra2 platform, real time 608 encoding is 40% faster than Bali. 609 610 Decoder speed was not a priority for this release, but improved 611 approximately 8.4% on x86. 612 613 Reduce motion vector search on alt-ref frame. 614 Encoder loopfilter running in its own thread 615 Reworked loopfilter to precalculate more parameters 616 SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}(). 617 Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3. 618 Removed redundant checks 619 Reduced structure sizes 620 utilize preload in ARMv6 MC/LPF/Copy routines 621 ARM optimized quantization, dfct, variance, subtract 622 Increase chrow row alignment to 16 bytes. 623 disable trellis optimization for first pass 624 Write SSSE3 sub-pixel filter function 625 Improve SSE2 half-pixel filter funtions 626 Add vp8_sub_pixel_variance16x8_ssse3 function 627 Reduce unnecessary distortion computation 628 Use diamond search to replace full search 629 Preload reference area in sub-pixel motion search (real-time mode) 630 631 - Quality: 632 This release focused primarily on one-pass use cases, including 633 video conferencing. Low latency data rate control was significantly 634 improved, improving streamability over bandwidth constrained links. 635 Added support for error concealment, allowing frames to maintain 636 visual quality in the presence of substantial packet loss. 637 638 Add rc_max_intra_bitrate_pct control 639 Limit size of initial keyframe in one-pass. 640 Improve framerate adaptation 641 Improved 1-pass CBR rate control 642 Improved KF insertion after fades to still. 643 Improved key frame detection. 644 Improved activity masking (lower PSNR impact for same SSIM boost) 645 Improved interaction between GF and ARFs 646 Adding error-concealment to the decoder. 647 Adding support for independent partitions 648 Adjusted rate-distortion constants 649 650 651 - Bug Fixes: 652 Removed firstpass motion map 653 Fix parallel make install 654 Fix multithreaded encoding for 1 MB wide frame 655 Fixed iwalsh_neon build problems with RVDS4.1 656 Fix semaphore emulation, spin-wait intrinsics on Windows 657 Fix build with xcode4 and simplify GLOBAL. 658 Mark ARM asm objects as allowing a non-executable stack. 659 Fix vpxenc encoding incorrect webm file header on big endian 660 661 6622011-03-07 v0.9.6 "Bali" 663 Our second named release, focused on a faster, higher quality, encoder. 664 665 - Upgrading: 666 This release is backwards compatible with Aylesbury (v0.9.5). Users 667 of older releases should refer to the Upgrading notes in this 668 document for that release. 669 670 - Enhancements: 671 vpxenc --psnr shows a summary when encode completes 672 --tune=ssim option to enable activity masking 673 improved postproc visualizations for development 674 updated support for Apple iOS to SDK 4.2 675 query decoder to determine which reference frames were updated 676 implemented error tracking in the decoder 677 fix pipe support on windows 678 679 - Speed: 680 Primary focus was on good quality mode, speed 0. Average improvement 681 on x86 about 40%, up to 100% on user-generated content at that speed. 682 Best quality mode speed improved 35%, and realtime speed 10-20%. This 683 release also saw significant improvement in realtime encoding speed 684 on ARM platforms. 685 686 Improved encoder threading 687 Dont pick encoder filter level when loopfilter is disabled. 688 Avoid double copying of key frames into alt and golden buffer 689 FDCT optimizations. 690 x86 sse2 temporal filter 691 SSSE3 version of fast quantizer 692 vp8_rd_pick_best_mbsegmentation code restructure 693 Adjusted breakout RD for SPLITMV 694 Changed segmentation check order 695 Improved rd_pick_intra4x4block 696 Adds armv6 optimized variance calculation 697 ARMv6 optimized sad16x16 698 ARMv6 optimized half pixel variance calculations 699 Full search SAD function optimization in SSE4.1 700 Improve MV prediction accuracy to achieve performance gain 701 Improve MV prediction in vp8_pick_inter_mode() for speed>3 702 703 - Quality: 704 Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release 705 also includes support for "activity masking," which greatly improves 706 SSIM at the expense of PSNR. For now, this feature is available with 707 the --tune=ssim option. Further experimentation in this area 708 is ongoing. This release also introduces a new rate control mode 709 called "CQ," which changes the allocation of bits within a clip to 710 the sections where they will have the most visual impact. 711 712 Tuning for the more exact quantizer. 713 Relax rate control for last few frames 714 CQ Mode 715 Limit key frame quantizer for forced key frames. 716 KF/GF Pulsing 717 Add simple version of activity masking. 718 make rdmult adaptive for intra in quantizer RDO 719 cap the best quantizer for 2nd order DC 720 change the threshold of DC check for encode breakout 721 722 - Bug Fixes: 723 Fix crash on Sparc Solaris. 724 Fix counter of fixed keyframe distance 725 ARNR filter pointer update bug fix 726 Fixed use of motion percentage in KF/GF group calc 727 Changed condition for using RD in Intra Mode 728 Fix encoder real-time only configuration. 729 Fix ARM encoder crash with multiple token partitions 730 Fixed bug first cluster timecode of webm file is wrong. 731 Fixed various encoder bugs with odd-sized images 732 vp8e_get_preview fixed when spatial resampling enabled 733 quantizer: fix assertion in fast quantizer path 734 Allocate source buffers to be multiples of 16 735 Fix for manual Golden frame frequency 736 Fix drastic undershoot in long form content 737 738 7392010-10-28 v0.9.5 "Aylesbury" 740 Our first named release, focused on a faster decoder, and a better encoder. 741 742 - Upgrading: 743 This release incorporates backwards-incompatible changes to the 744 ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec. 745 746 vpxdec 747 * the -q (quiet) option has been removed, and replaced with 748 -v (verbose). the output is quiet by default. Use -v to see 749 the version number of the binary. 750 751 * The default behavior is now to write output to a single file 752 instead of individual frames. The -y option has been removed. 753 Y4M output is the default. 754 755 * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12 756 options must be specified. 757 758 $ ivfdec -o OUTPUT INPUT 759 $ vpxdec --i420 -o OUTPUT INPUT 760 761 * If an output file is not specified, the default is to write 762 Y4M to stdout. This makes piping more natural. 763 764 $ ivfdec -y -o - INPUT | ... 765 $ vpxdec INPUT | ... 766 767 * The output file has additional flexibility for formatting the 768 filename. It supports escape characters for constructing a 769 filename from the width, height, and sequence number. This 770 replaces the -p option. To get the equivalent: 771 772 $ ivfdec -p frame INPUT 773 $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT 774 775 vpxenc 776 * The output file must be specified with -o, rather than as the 777 last argument. 778 779 $ ivfenc <options> INPUT OUTPUT 780 $ vpxenc <options> -o OUTPUT INPUT 781 782 * The output defaults to webm. To get IVF output, use the --ivf 783 option. 784 785 $ ivfenc <options> INPUT OUTPUT.ivf 786 $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT 787 788 789 - Enhancements: 790 ivfenc and ivfdec have been renamed to vpxenc, vpxdec. 791 vpxdec supports .webm input 792 vpxdec writes .y4m by default 793 vpxenc writes .webm output by default 794 vpxenc --psnr now shows the average/overall PSNR at the end 795 ARM platforms now support runtime cpu detection 796 vpxdec visualizations added for motion vectors, block modes, references 797 vpxdec now silent by default 798 vpxdec --progress shows frame-by-frame timing information 799 vpxenc supports the distinction between --fps and --timebase 800 NASM is now a supported assembler 801 configure: enable PIC for shared libs by default 802 configure: add --enable-small 803 configure: support for ppc32-linux-gcc 804 configure: support for sparc-solaris-gcc 805 806 - Bugs: 807 Improve handling of invalid frames 808 Fix valgrind errors in the NEON loop filters. 809 Fix loopfilter delta zero transitions 810 Fix valgrind errors in vp8_sixtap_predict8x4_armv6(). 811 Build fixes for darwin-icc 812 813 - Speed: 814 20-40% (average 28%) improvement in libvpx decoder speed, 815 including: 816 Rewrite vp8_short_walsh4x4_sse2() 817 Optimizations on the loopfilters. 818 Miscellaneous improvements for Atom 819 Add 4-tap version of 2nd-pass ARMv6 MC filter. 820 Improved multithread utilization 821 Better instruction choices on x86 822 reorder data to use wider instructions 823 Update NEON wide idcts 824 Make block access to frame buffer sequential 825 Improved subset block search 826 Bilinear subpixel optimizations for ssse3. 827 Decrease memory footprint 828 829 Encoder speed improvements (percentage gain not measured): 830 Skip unnecessary search of identical frames 831 Add SSE2 subtract functions 832 Improve bounds checking in vp8_diamond_search_sadx4() 833 Added vp8_fast_quantize_b_sse2 834 835 - Quality: 836 Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality 837 encoding mode, and up to 60% improvement on very noisy, still 838 or slow moving source video 839 840 Motion compensated temporal filter for Alt-Ref Noise Reduction 841 Improved use of trellis quantization on 2nd order Y blocks 842 Tune effect of motion on KF/GF boost in two pass 843 Allow coefficient optimization for good quality speed 0. 844 Improved control of active min quantizer for two pass. 845 Enable ARFs for non-lagged compress 846 8472010-09-02 v0.9.2 848 - Enhancements: 849 Disable frame dropping by default 850 Improved multithreaded performance 851 Improved Force Key Frame Behaviour 852 Increased rate control buffer level precision 853 Fix bug in 1st pass motion compensation 854 ivfenc: correct fixed kf interval, --disable-kf 855 - Speed: 856 Changed above and left context data layout 857 Rework idct calling structure. 858 Removed unnecessary MB_MODE_INFO copies 859 x86: SSSE3 sixtap prediction 860 Reworked IDCT to include reconstruction (add) step 861 Swap alt/gold/new/last frame buffer ptrs instead of copying. 862 Improve SSE2 loopfilter functions 863 Change bitreader to use a larger window. 864 Avoid loopfilter reinitialization when possible 865 - Quality: 866 Normalize quantizer's zero bin and rounding factors 867 Add trellis quantization. 868 Make the quantizer exact. 869 Updates to ARNR filtering algorithm 870 Fix breakout thresh computation for golden & AltRef frames 871 Redo the forward 4x4 dct 872 Improve the accuracy of forward walsh-hadamard transform 873 Further adjustment of RD behaviour with Q and Zbin. 874 - Build System: 875 Allow linking of libs built with MinGW to MSVC 876 Fix target auto-detection on mingw32 877 Allow --cpu= to work for x86. 878 configure: pass original arguments through to make dist 879 Fix builds without runtime CPU detection 880 msvs: fix install of codec sources 881 msvs: Change devenv.com command line for better msys support 882 msvs: Add vs9 targets. 883 Add x86_64-linux-icc target 884 - Bugs: 885 Potential crashes on older MinGW builds 886 Fix two-pass framrate for Y4M input. 887 Fixed simple loop filter, other crashes on ARM v6 888 arm: fix missing dependency with --enable-shared 889 configure: support directories containing .o 890 Replace pinsrw (SSE) with MMX instructions 891 apple: include proper mach primatives 892 Fixed rate control bug with long key frame interval. 893 Fix DSO link errors on x86-64 when not using a version script 894 Fixed buffer selection for UV in AltRef filtering 895 896 8972010-06-17 v0.9.1 898 - Enhancements: 899 * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O 900 * Speed optimizations 901 - Bugfixes: 902 * Rate control 903 * Prevent out-of-bounds accesses on invalid data 904 - Build system updates: 905 * Detect toolchain to be used automatically for native builds 906 * Support building shared libraries 907 * Better autotools emulation (--prefix, --libdir, DESTDIR) 908 - Updated LICENSE 909 * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html 910 911 9122010-05-18 v0.9.0 913 - Initial open source release. Welcome to WebM and VP8! 914 915