xref: /aosp_15_r20/external/libvpx/CHANGELOG (revision fb1b10ab9aebc7c7068eedab379b749d7e3900be)
12024-10-22 v1.15.0 "Wigeon Duck"
2  This release includes new codec control for key frame filtering, more Neon
3  optimizations, improvements to RTC encoding and bug fixes.
4
5  - Upgrading:
6    This release is ABI compatible with the previous release.
7
8    Temporal filtering improvement that can be turned on with the new codec
9    control VP9E_SET_KEY_FRAME_FILTERING, which gives 1+% BD-rate saving with
10    minimal encoder time increase.
11
12    libwebm is upgraded to libwebm-1.0.0.31-10-g3b63004
13
14  - Enhancement:
15    Neon optimization speed up
16      1-3% speed up across speed 5 to 10 for RTC
17      3% speed up for speed 0 and 1 for VoD in standard bitdepth
18      3% and 7% speed up for speed 0 and 1 respectively for VoD in high bitdepth
19    Scene detection is allowed for all RTC speeds (>=5)
20    Support profile guided optimizations
21
22    Delta quantization parameters for UV channels for vp8 is supported in RTC
23    rate control library
24
25    Rate control parameters are reset and maximum QP is enforced on scene
26    changes in SVC when there is no inter-layer prediction
27
28  - Bug fixes:
29    Fix to Uninitialized scalar variable in `vp9_rd_pick_inter_mode_sb()`
30    Fix to Integer-overflow in `resize_multistep`
31    Fix to Heap-buffer-overflow in `vpx_sad64x64_avx2`
32    Fix to Crash in `vpx_sad8x8_sse2`
33    Fix to Assertion in `write_modes`
34    Support profile guided optimizations
35    Fix to Integer-overflow in `encode_frame_to_data_rate`
36    Fix to Integer-overflow in `vp9_svc_check_reset_layer_rc_flag`
37    Fix to core dump error from /usr/bin/tools/tiny_ssim --help
38    Fix to use-of-uninitialized-value in `vp9_setup_tpl_stats`
39    Fix to Undefined-shift in `vp9_cyclic_refresh_setup`
40    Fix to redundant `&& __GNUC__` preproc check
41    Fix to valgrind warning in EncodeAPI.OssFuzz69906
42    Fix to Index-out-of-bounds in `vp8_rd_pick_inter_mode`
43    Fix to Integer-overflow in `vp8_pick_frame_size`
44    Fix to Use-of-uninitialized-value in `vpx_codec_peek_stream_info`
45    Fix to log clutters with the message "Warning: Desired height too large"
46    Fix to Integer-overflow in `vp9_svc_adjust_avg_frame_qindex`
47
48    Fix to integer overflows caused by huge target bitrate, frame rate, or
49    g_timebase numerator or denominator
50
51    Fix to missing license headers
52    Fix to build failure for Android Armv7
53    Fix to integer overflows in image helpers
54    Fix to Integer-overflow in `vp9_calc_iframe_target_size_one_pass_cbr`
55    Fix to Heap-buffer-overflow in `vp9_pick_inter_mode`
56    Fix to Segv in `vp9_multi_thread_tile_init`
57    Fix to Use-of-uninitialized-value in `vp9_row_mt_sync_mem_dealloc`
58    Fix to Crash in `mbloop_filter_vertical_edge_c`
59    Fix to Check failed in CheckUnwind
60    Fix to Heap-buffer-overflow in `write_modes_b` and `vpx_write`
61    Fix to Possible signed integer overflow found in `vpx_codec_encode`
62    Fix to build conflicts between Abseil and libaom/libvpx in Win ARM64 builds
63    Fix to build failures on aarch64
64    Fix to Data race in libvpx ARM NEON
65    Fix to Heap-buffer-overflow in `scale_plane_1_to_2_phase_0`
66    Fix to integer overflow in `encode_mb_row`
67    Fix to Floating-point-exception in `vp8_pick_frame_size`
68    Fix to Heap-buffer-overflow in `vp9_enc_setup_mi`
69    Fix to build failure with --target=arm64-win64-vs17
70    Fix to heap-buffer-overflow write in `vpx_img_read()`
71    Fix to C vs armv8-linux-gcc encode mismatches for `y4m_360p_10bit_input`
72    Fix to Null-dereference READ in `ml_predict_var_rd_partitioning`
73    Fix to Heap-buffer-overflow in `vpx_scaled_2d_ssse3`
74    Fix to Crash in `convolve_horiz`
75    Fix to Ill in `vpx_scaled_2d_ssse3`
76    Fix to Global-buffer-overflow in `cost_coeffs`
77
782024-05-21 v1.14.1 "Venetian Duck"
79  This release includes enhancements and bug fixes.
80
81  - Upgrading:
82    This release is ABI compatible with the previous release.
83
84  - Enhancement:
85    Improved the detection of compiler support for AArch64 extensions,
86    particularly SVE.
87
88    Added vpx_codec_get_global_headers() support for VP9.
89
90  - Bug fixes:
91    Added buffer bounds checks to vpx_writer and vpx_write_bit_buffer.
92    Fix to GetSegmentationData() crash in aq_mode=0 for RTC rate control.
93    Fix to alloc for row_base_thresh_freq_fac.
94    Free row mt memory before freeing cpi->tile_data.
95    Fix to buffer alloc for vp9_bitstream_worker_data.
96    Fix to VP8 race issue for multi-thread with pnsr_calc.
97    Fix to uv width/height in vp9_scale_and_extend_frame_ssse3.
98    Fix to integer division by zero and overflow in calc_pframe_target_size().
99    Fix to integer overflow in vpx_img_alloc() & vpx_img_wrap()(CVE-2024-5197).
100    Fix to UBSan error in vp9_rc_update_framerate().
101    Fix to UBSan errors in vp8_new_framerate().
102    Fix to integer overflow in vp8 encodeframe.c.
103    Handle EINTR from sem_wait().
104
1052024-01-02 v1.14.0 "Venetian Duck"
106  This release drops support for old C compilers, such as Visual Studio 2012
107  and older, that disallow mixing variable declarations and statements (a C99
108  feature). It adds support for run-time CPU feature detection for Arm
109  platforms, as well as support for darwin23 (macOS 14).
110
111  - Upgrading:
112    This release is ABI incompatible with the previous release.
113
114    Various new features for rate control library for real-time: SVC parallel
115    encoding, loopfilter level, support for frame dropping, and screen content.
116
117    New callback function send_tpl_gop_stats for vp9 external rate control
118    library, which can be used to transmit TPL stats for a group of pictures. A
119    public header vpx_tpl.h is added for the definition of TPL stats used in
120    this callback.
121
122    libwebm is upgraded to libwebm-1.0.0.29-9-g1930e3c.
123
124  - Enhancement:
125    Improvements on Neon optimizations: VoD: 12-35% speed up for bitdepth 8,
126    68%-151% speed up for high bitdepth.
127
128    Improvements on AVX2 and SSE optimizations.
129    Improvements on LSX optimizations for LoongArch.
130    42-49% speedup on speed 0 VoD encoding.
131    Android API level predicates.
132
133  - Bug fixes:
134    Fix to missing prototypes from the rtcd header.
135    Fix to segfault when total size is enlarged but width is smaller.
136    Fix to the build for arm64ec using MSVC.
137    Fix to copy BLOCK_8X8's mi to PICK_MODE_CONTEXT::mic.
138    Fix to -Wshadow warnings.
139    Fix to heap overflow in vpx_get4x4sse_cs_neon.
140    Fix to buffer overrun in highbd Neon subpel variance filters.
141    Added bitexact encode test script.
142    Fix to -Wl,-z,defs with Clang's sanitizers.
143    Fix to decoder stability after error & continued decoding.
144    Fix to mismatch of VP9 encode with NEON intrinsics with C only version.
145    Fix to Arm64 MSVC compile vpx_highbd_fdct4x4_neon.
146    Fix to fragments count before use.
147    Fix to a case where target bandwidth is 0 for SVC.
148    Fix mask in vp9_quantize_avx2,highbd_get_max_lane_eob.
149    Fix to int overflow in vp9_calc_pframe_target_size_one_pass_cbr.
150    Fix to integer overflow in vp8,ratectrl.c.
151    Fix to integer overflow in vp9 svc.
152    Fix to avg_frame_bandwidth overflow.
153    Fix to per frame qp for temporal layers.
154    Fix to unsigned integer overflow in sse computation.
155    Fix to uninitialized mesh feature for BEST mode.
156    Fix to overflow in highbd temporal_filter.
157    Fix to unaligned loads w/w==4 in vpx_convolve_copy_neon.
158    Skip arm64_neon.h workaround w/VS >= 2019.
159    Fix to c vs avx mismatch of diamond_search_sad().
160    Fix to c vs intrinsic mismatch of vpx_hadamard_32x32() function.
161    Fix to a bug in vpx_hadamard_32x32_neon().
162    Fix to Clang -Wunreachable-code-aggressive warnings.
163    Fix to a bug in vpx_highbd_hadamard_32x32_neon().
164    Fix to -Wunreachable-code in mfqe_partition.
165    Force mode search on 64x64 if no mode is selected.
166    Fix to ubsan failure caused by left shift of negative.
167    Fix to integer overflow in calc_pframe_target_size.
168    Fix to float-cast-overflow in vp8_change_config().
169    Fix to a null ptr before use.
170    Conditionally skip using inter frames in speed features.
171    Remove invalid reference frames.
172    Disable intra mode search speed features conditionally.
173    Set nonrd keyframe under dynamic change of deadline for rtc.
174    Fix to scaled reference offsets.
175    Set skip_recode=0 in nonrd_pick_sb_modes.
176    Fix to an edge case when downsizing to one.
177    Fix to a bug in frame scaling.
178    Fix to pred buffer stride.
179    Fix to a bug in simple motion search.
180    Update frame size in actual encoding.
181
1822023-09-29 v1.13.1 "Ugly Duckling"
183  This release contains two security related fixes. One each for VP8 and VP9.
184
185  - Upgrading:
186    This release is ABI compatible with the previous release.
187
188  - Bug fixes:
189    https://crbug.com/1486441 (CVE-2023-5217)
190    Fix to a crash related to VP9 encoding (#1642, CVE-2023-6349)
191
1922023-01-31 v1.13.0 "Ugly Duckling"
193  This release includes more Neon and AVX2 optimizations, adds a new codec
194  control to set per frame QP, upgrades GoogleTest to v1.12.1, and includes
195  numerous bug fixes.
196
197  - Upgrading:
198    This release is ABI incompatible with the previous release.
199
200    New codec control VP9E_SET_QUANTIZER_ONE_PASS to set per frame QP.
201
202    GoogleTest is upgraded to v1.12.1.
203
204    .clang-format is upgraded to clang-format-11.
205
206    VPX_EXT_RATECTRL_ABI_VERSION was bumped due to incompatible changes to the
207    feature of using external rate control models for vp9.
208
209  - Enhancement:
210    Numerous improvements on Neon optimizations.
211    Numerous improvements on AVX2 optimizations.
212    Additional ARM targets added for Visual Studio.
213
214  - Bug fixes:
215    Fix to calculating internal stats when frame dropped.
216    Fix to segfault for external resize test in vp9.
217    Fix to build system with replacing egrep with grep -E.
218    Fix to a few bugs with external RTC rate control library.
219    Fix to make SVC work with VBR.
220    Fix to key frame setting in VP9 external RC.
221    Fix to -Wimplicit-int (Clang 16).
222    Fix to VP8 external RC for buffer levels.
223    Fix to VP8 external RC for dynamic update of layers.
224    Fix to VP9 auto level.
225    Fix to off-by-one error of max w/h in validate_config.
226    Fix to make SVC work for Profile 1.
227
2282022-06-17 v1.12.0 "Torrent Duck"
229  This release adds optimizations for Loongarch, adds support for vp8 in the
230  real-time rate control library, upgrades GoogleTest to v1.11.0, updates
231  libwebm to libwebm-1.0.0.28-20-g206d268, and includes numerous bug fixes.
232
233  - Upgrading:
234    This release is ABI compatible with the previous release.
235
236    vp8 support in the real-time rate control library.
237    New codec control VP8E_SET_RTC_EXTERNAL_RATECTRL is added.
238
239    Configure support for darwin21 is added.
240
241    GoogleTest is upgraded to v1.11.0.
242
243    libwebm is updated to libwebm-1.0.0.28-20-g206d268.
244
245    Allow SimpleEncode environment to take target level as input to match
246    the level conformance in vp9.
247
248  - Enhancement:
249    Numerous improvements on checking memory allocations.
250    Optimizations for Loongarch.
251    Code clean-up.
252
253  - Bug fixes:
254    Fix to a crash related to {vp8/vp9}_set_roi_map.
255    Fix to compiling failure with -Wformat-nonliteral.
256    Fix to integer overflow with vp9 with high resolution content.
257    Fix to AddNoiseTest failure with ARMv7.
258    Fix to libvpx Null-dereference READ in vp8.
259
2602021-09-27 v1.11.0 "Smew Duck"
261  This maintenance release adds support for VBR mode in VP9 rate control
262  interface, new codec controls to get quantization parameters and loop filter
263  levels, and includes several improvements to NEON and numerous bug fixes.
264
265  - Upgrading:
266    This release is ABI incompatible with the previous release.
267    New codec control is added to get quantization parameters and loop filter
268    levels.
269
270    VBR mode is supported in VP9 rate control library.
271
272  - Enhancement:
273    Numerous improvements for Neon optimizations.
274    Code clean-up and refactoring.
275    Calculation of rd multiplier is changed with BDRATE gains.
276
277  - Bug fixes:
278    Fix to overflow on duration.
279    Fix to several instances of -Wunused-but-set-variable.
280    Fix to avoid chroma resampling for 420mpeg2 input.
281    Fix to overflow in calc_iframe_target_size.
282    Fix to disallow skipping transform and quantization.
283    Fix some -Wsign-compare warnings in simple_encode.
284    Fix input file path in simple_encode_test.
285    Fix valid range for under/over_shoot pct.
286
2872021-03-09 v1.10.0 "Ruddy Duck"
288  This maintenance release adds support for darwin20 and new codec controls, as
289  well as numerous bug fixes.
290
291  - Upgrading:
292    This release is ABI incompatible with the previous release.
293    New codec control is added to disable loopfilter for VP9.
294
295    New encoder control is added to disable feature to increase Q on overshoot
296    detection for CBR.
297
298    Configure support for darwin20 is added.
299
300    New codec control is added for VP9 rate control. The control ID of this
301    interface is VP9E_SET_EXTERNAL_RATE_CONTROL. To make VP9 use a customized
302    external rate control model, users will have to implement each callback
303    function in vpx_rc_funcs_t and register them using libvpx API
304    vpx_codec_control_() with the control ID.
305
306  - Enhancement:
307    Use -std=gnu++11 instead of -std=c++11 for c++ files.
308
309  - Bug fixes:
310    Override assembler with --as option of configure for MSVS.
311    Fix several compilation issues with gcc 4.8.5.
312    Fix to resetting rate control for temporal layers.
313    Fix to the rate control stats of SVC example encoder when number of spatial
314    layers is 1.
315    Fix to reusing motion vectors from the base spatial layer in SVC.
316    2 pass related flags removed from SVC example encoder.
317
3182020-07-29 v1.9.0 "Quacking Duck"
319  This release adds support for NV12, a separate library for rate control, as
320  well as incremental improvements.
321
322  - Upgrading:
323    This release is ABI compatible with the previous release.
324    NV12 support is added to this release.
325    A new interface is added for VP9 rate control. The new library libvp9rc.a
326    must be linked by applications.
327    Googletest is updated to v1.10.0.
328    simple_encode.cc is compiled into a new library libsimple_encode.a with
329    CONFIG_RATE_CTRL.
330
331  - Enhancement:
332    Various changes to improve VP9 SVC, rate control, quality and speed to real
333    time encoding.
334
335  - Bug fixes:
336    Fix key frame update refresh simulcast flexible svc.
337    Fix to disable_16x16part speed feature for real time encoding.
338    Fix some signed integer overflows for VP9 rate control.
339    Fix initialization of delta_q_uv.
340    Fix condition in regulate_q for cyclic refresh.
341    Various fixes to dynamic resizing for VP9 SVC.
342
3432019-12-09 v1.8.2 "Pekin Duck"
344  This release collects incremental improvements to many aspects of the library.
345
346  - Upgrading:
347    This release is ABI compatible with the previous release.
348    ARCH_* defines have been removed in favor of VPX_ARCH_*.
349
3502019-07-15 v1.8.1 "Orpington Duck"
351  This release collects incremental improvements to many aspects of the library.
352
353  - Upgrading:
354    This release is ABI incompatible with the previous release.
355    VP8E_SET_CPUUSED now accepts values up to 9 for vp9.
356    VPX_CTRL_VP9E_SET_MAX_INTER_BITRATE_PCT had a spelling fix (was VP8E).
357    The --sdk-path option has been removed. If you were using it to build for
358      Android please read build/make/Android.mk for alternatives.
359    All PPC optimizations have been disabled:
360      https://bugs.chromium.org/p/webm/issues/detail?id=1522.
361
362  - Enhancements:
363    Various changes to improve encoder rate control, quality and speed
364      for practically every use case.
365
366  - Bug fixes:
367    vp9-rtc: Fix color artifacts for speed >= 8.
368
3692019-01-31 v1.8.0 "Northern Shoveler Duck"
370  This release focused on encoding performance for realtime and VOD use cases.
371
372  - Upgrading:
373    This release is ABI incompatible with the previous release. This adds and
374    improves several vp9 controls. Most are related to SVC:
375      VP9E_SET_SVC_FRAME_DROP_LAYER:
376        - Frame dropping in SVC.
377      VP9E_SET_SVC_INTER_LAYER_PRED:
378        - Inter-layer prediction in SVC.
379      VP9E_SET_SVC_GF_TEMPORAL_REF:
380        - Enable long term temporal reference in SVC.
381      VP9E_SET_SVC_REF_FRAME_CONFIG/VP9E_GET_SVC_REF_FRAME_CONFIG:
382        - Extend and improve this control for better flexibility in setting SVC
383          pattern dynamically.
384      VP9E_SET_POSTENCODE_DROP:
385        - Allow for post-encode frame dropping (applies to non-SVC too).
386      VP9E_SET_SVC_SPATIAL_LAYER_SYNC:
387        - Enable spatial layer sync frames.
388      VP9E_SET_SVC_LAYER_ID:
389        - Extend api to specify temporal id for each spatial layers.
390      VP9E_SET_ROI_MAP:
391        - Extend Region of Interest functionality to VP9.
392
393  - Enhancements:
394    2 pass vp9 encoding has improved substantially. When using --auto-alt-ref=6,
395    we see approximately 8% for VBR and 10% for CQ. When using --auto-alt-ref=1,
396    the gains are approximately 4% for VBR and 5% for CQ.
397
398    For real-time encoding, speed 7 has improved by ~5-10%. Encodes targeted at
399    screen sharing have improved when the content changes significantly (slide
400    sharing) or scrolls. There is a new speed 9 setting for mobile devices which
401    is about 10-20% faster than speed 8.
402
403  - Bug fixes:
404    VP9 denoiser issue.
405    VP9 partition issue for 1080p.
406    VP9 rate control improvments.
407    Postprocessing Multi Frame Quality Enhancement (MFQE) issue.
408    VP8 multithread decoder issues.
409    A variety of fuzzing issues.
410
4112018-01-04 v1.7.0 "Mandarin Duck"
412  This release focused on high bit depth performance (10/12 bit) and vp9
413  encoding improvements.
414
415  - Upgrading:
416    This release is ABI incompatible due to new vp9 encoder features.
417
418    Frame parallel decoding for vp9 has been removed.
419
420  - Enhancements:
421    vp9 encoding supports additional threads with --row-mt. This can be greater
422    than the number of tiles.
423
424    Two new vp9 encoder options have been added:
425      --corpus-complexity
426      --tune-content=film
427
428    Additional tooling for respecting the vp9 "level" profiles has been added.
429
430  - Bug fixes:
431    A variety of fuzzing issues.
432    vp8 threading fix for ARM.
433    Codec control VP9_SET_SKIP_LOOP_FILTER fixed.
434    Reject invalid multi resolution configurations.
435
4362017-01-09 v1.6.1 "Long Tailed Duck"
437  This release improves upon the VP9 encoder and speeds up the encoding and
438  decoding processes.
439
440  - Upgrading:
441    This release is ABI compatible with 1.6.0.
442
443  - Enhancements:
444    Faster VP9 encoding and decoding.
445    High bit depth builds now provide similar speed for 8 bit encode and decode
446    for x86 targets. Other platforms and higher bit depth improvements are in
447    progress.
448
449  - Bug Fixes:
450    A variety of fuzzing issues.
451
4522016-07-20 v1.6.0 "Khaki Campbell Duck"
453  This release improves upon the VP9 encoder and speeds up the encoding and
454  decoding processes.
455
456  - Upgrading:
457    This release is ABI incompatible with 1.5.0 due to a new 'color_range' enum
458    in vpx_image and some minor changes to the VP8_COMP structure.
459
460    The default key frame interval for VP9 has changed from 128 to 9999.
461
462  - Enhancement:
463    A core focus has been performance for low end Intel processors. SSSE3
464    instructions such as 'pshufb' have been avoided and instructions have been
465    reordered to better accommodate the more constrained pipelines.
466
467    As a result, devices based on Celeron processors have seen substantial
468    decoding improvements. From Indian Runner Duck to Javan Whistling Duck,
469    decoding speed improved between 10 and 30%. Between Javan Whistling Duck
470    and Khaki Campbell Duck, it improved another 10 to 15%.
471
472    While Celeron benefited most, Core-i5 also improved 5% and 10% between the
473    respective releases.
474
475    Realtime performance for WebRTC for both speed and quality has received a
476    lot of attention.
477
478  - Bug Fixes:
479    A number of fuzzing issues, found variously by Mozilla, Chromium and others,
480    have been fixed and we strongly recommend updating.
481
4822015-11-09 v1.5.0 "Javan Whistling Duck"
483  This release improves upon the VP9 encoder and speeds up the encoding and
484  decoding processes.
485
486  - Upgrading:
487    This release is ABI incompatible with 1.4.0. It drops deprecated VP8
488    controls and adds a variety of VP9 controls for testing.
489
490    The vpxenc utility now prefers VP9 by default.
491
492  - Enhancements:
493    Faster VP9 encoding and decoding
494    Smaller library size by combining functions used by VP8 and VP9
495
496  - Bug Fixes:
497    A variety of fuzzing issues
498
4992015-04-03 v1.4.0 "Indian Runner Duck"
500  This release includes significant improvements to the VP9 codec.
501
502  - Upgrading:
503    This release is ABI incompatible with 1.3.0. It drops the compatibility
504    layer, requiring VPX_IMG_FMT_* instead of IMG_FMT_*, and adds several codec
505    controls for VP9.
506
507  - Enhancements:
508    Faster VP9 encoding and decoding
509    Multithreaded VP9 decoding (tile and frame-based)
510    Multithreaded VP9 encoding - on by default
511    YUV 4:2:2 and 4:4:4 support in VP9
512    10 and 12bit support in VP9
513    64bit ARM support by replacing ARM assembly with intrinsics
514
515  - Bug Fixes:
516    Fixes a VP9 bitstream issue in Profile 1. This only affected non-YUV 4:2:0
517    files.
518
519  - Known Issues:
520    Frame Parallel decoding fails for segmented and non-420 files.
521
5222013-11-15 v1.3.0 "Forest"
523  This release introduces the VP9 codec in a backward-compatible way.
524  All existing users of VP8 can continue to use the library without
525  modification. However, some VP8 options do not map to VP9 in the same manner.
526
527  The VP9 encoder in this release is not feature complete. Users interested in
528  the encoder are advised to use the git master branch and discuss issues on
529  libvpx mailing lists.
530
531  - Upgrading:
532    This release is ABI and API compatible with Duclair (v1.0.0). Users
533    of older releases should refer to the Upgrading notes in this document
534    for that release.
535
536  - Enhancements:
537      Get rid of bashisms in the main build scripts
538      Added usage info on command line options
539      Add lossless compression mode
540      Dll build of libvpx
541      Add additional Mac OS X targets: 10.7, 10.8 and 10.9 (darwin11-13)
542      Add option to disable documentation
543      configure: add --enable-external-build support
544      make: support V=1 as short form of verbose=yes
545      configure: support mingw-w64
546      configure: support hardfloat armv7 CHOSTS
547      configure: add support for android x86
548      Add estimated completion time to vpxenc
549      Don't exit on decode errors in vpxenc
550      vpxenc: support scaling prior to encoding
551      vpxdec: support scaling output
552      vpxenc: improve progress indicators with --skip
553      msvs: Don't link to winmm.lib
554      Add a new script for producing vcxproj files
555      Produce Visual Studio 10 and 11 project files
556      Produce Windows Phone project files
557      msvs-build: use msbuild for vs >= 2005
558      configure: default configure log to config.log
559      Add encoding option --static-thresh
560
561  - Speed:
562      Miscellaneous speed optimizations for VP8 and VP9.
563
564  - Quality:
565      In general, quality is consistent with the Eider release.
566
567  - Bug Fixes:
568      This release represents approximately a year of engineering effort,
569      and contains multiple bug fixes. Please refer to git history for details.
570
571
5722012-12-21 v1.2.0
573  This release acts as a checkpoint for a large amount of internal refactoring
574  and testing. It also contains a number of small bugfixes, so all users are
575  encouraged to upgrade.
576
577  - Upgrading:
578    This release is ABI and API compatible with Duclair (v1.0.0). Users
579    of older releases should refer to the Upgrading notes in this
580    document for that release.
581
582  - Enhancements:
583      VP8 optimizations for MIPS dspr2
584      vpxenc: add -quiet option
585
586  - Speed:
587      Encoder and decoder speed is consistent with the Eider release.
588
589  - Quality:
590      In general, quality is consistent with the Eider release.
591
592      Minor tweaks to ARNR filtering
593      Minor improvements to real time encoding with multiple temporal layers
594
595  - Bug Fixes:
596      Fixes multithreaded encoder race condition in loopfilter
597      Fixes multi-resolution threaded encoding
598      Fix potential encoder dead-lock after picture resize
599
600
6012012-05-09 v1.1.0 "Eider"
602  This introduces a number of enhancements, mostly focused on real-time
603  encoding. In addition, it fixes a decoder bug (first introduced in
604  Duclair) so all users of that release are encouraged to upgrade.
605
606  - Upgrading:
607    This release is ABI and API compatible with Duclair (v1.0.0). Users
608    of older releases should refer to the Upgrading notes in this
609    document for that release.
610
611    This release introduces a new temporal denoiser, controlled by the
612    VP8E_SET_NOISE_SENSITIVITY control. The temporal denoiser does not
613    currently take a strength parameter, so the control is effectively
614    a boolean - zero (off) or non-zero (on). For compatibility with
615    existing applications, the values accepted are the same as those
616    for the spatial denoiser (0-6). The temporal denoiser is enabled
617    by default, and the older spatial denoiser may be restored by
618    configuring with --disable-temporal-denoising. The temporal denoiser
619    is more computationally intensive than the spatial one.
620
621    This release removes support for a legacy, decode only API that was
622    supported, but deprecated, at the initial release of libvpx
623    (v0.9.0). This is not expected to have any impact. If you are
624    impacted, you can apply a reversion to commit 2bf8fb58 locally.
625    Please update to the latest libvpx API if you are affected.
626
627  - Enhancements:
628      Adds a motion compensated temporal denoiser to the encoder, which
629      gives higher quality than the older spatial denoiser. (See above
630      for notes on upgrading).
631
632      In addition, support for new compilers and platforms were added,
633      including:
634        improved support for XCode
635        Android x86 NDK build
636        OS/2 support
637        SunCC support
638
639      Changing resolution with vpx_codec_enc_config_set() is now
640      supported. Previously, reinitializing the codec was required to
641      change the input resolution.
642
643      The vpxenc application has initial support for producing multiple
644      encodes from the same input in one call. Resizing is not yet
645      supported, but varying other codec parameters is. Use -- to
646      delineate output streams. Options persist from one stream to the
647      next.
648
649      Also, the vpxenc application will now use a keyframe interval of
650      5 seconds by default. Use the --kf-max-dist option to override.
651
652  - Speed:
653      Decoder performance improved 2.5% versus Duclair. Encoder speed is
654      consistent with Duclair for most material. Two pass encoding of
655      slideshow-like material will see significant improvements.
656
657      Large realtime encoding speed gains at a small quality expense are
658      possible by configuring the on-the-fly bitpacking experiment with
659      --enable-onthefly-bitpacking. Realtime encoder can be up to 13%
660      faster (ARM) depending on the number of threads and bitrate
661      settings. This technique sees constant gain over the 5-16 speed
662      range. For VC style input the loss seen is up to 0.2dB. See commit
663      52cf4dca for further details.
664
665  - Quality:
666      On the whole, quality is consistent with the Duclair release. Some
667      tweaks:
668
669        Reduced blockiness in easy sections by applying a penalty to
670        intra modes.
671
672        Improved quality of static sections (like slideshows) with
673        two pass encoding.
674
675        Improved keyframe sizing with multiple temporal layers
676
677  - Bug Fixes:
678      Corrected alt-ref contribution to frame rate for visible updates
679      to the alt-ref buffer. This affected applications making manual
680      usage of the frame reference flags, or temporal layers.
681
682      Additional constraints were added to disable multi-frame quality
683      enhancement (MFQE) in sections of the frame where there is motion.
684      (#392)
685
686      Fixed corruption issues when vpx_codec_enc_config_set() was called
687      with spatial resampling enabled.
688
689      Fixed a decoder error introduced in Duclair where the segmentation
690      map was not being reinitialized on keyframes (#378)
691
692
6932012-01-27 v1.0.0 "Duclair"
694  Our fourth named release, focused on performance and features related to
695  real-time encoding. It also fixes a decoder crash bug introduced in
696  v0.9.7, so all users of that release are encouraged to upgrade.
697
698  - Upgrading:
699      This release is ABI incompatible with prior releases of libvpx, so the
700      "major" version number has been bumped to 1. You must recompile your
701      applications against the latest version of the libvpx headers. The
702      API remains compatible, and this should not require code changes in most
703      applications.
704
705  - Enhancements:
706      This release introduces several substantial new features to the encoder,
707      of particular interest to real time streaming applications.
708
709      Temporal scalability allows the encoder to produce a stream that can
710      be decimated to different frame rates, with independent rate targeting
711      for each substream.
712
713      Multiframe quality enhancement postprocessing can make visual quality
714      more consistent in the presence of frames that are substantially
715      different quality than the surrounding frames, as in the temporal
716      scalability case and in some forced keyframe scenarios.
717
718      Multiple-resolution encoding support allows the encoding of the
719      same content at different resolutions faster than encoding them
720      separately.
721
722  - Speed:
723      Optimization targets for this release included the decoder and the real-
724      time modes of the encoder. Decoder speed on x86 has improved 10.5% with
725      this release. Encoder improvements followed a curve where speeds 1-3
726      improved 4.0%-1.5%, speeds 4-8 improved <1%, and speeds 9-16 improved
727      1.5% to 10.5%, respectively. "Best" mode speed is consistent with the
728      Cayuga release.
729
730  - Quality:
731      Encoder quality in the single stream case is consistent with the Cayuga
732      release.
733
734  - Bug Fixes:
735      This release fixes an OOB read decoder crash bug present in v0.9.7
736      related to the clamping of motion vectors in SPLITMV blocks. This
737      behavior could be triggered by corrupt input or by starting
738      decoding from a P-frame.
739
740
7412011-08-15 v0.9.7-p1 "Cayuga" patch 1
742  This is an incremental bugfix release against Cayuga. All users of that
743  release are strongly encouraged to upgrade.
744
745    - Fix potential OOB reads (cdae03a)
746
747          An unbounded out of bounds read was discovered when the
748          decoder was requested to perform error concealment (new in
749          Cayuga) given a frame with corrupt partition sizes.
750
751          A bounded out of bounds read was discovered affecting all
752          versions of libvpx. Given an multipartition input frame that
753          is truncated between the mode/mv partition and the first
754          residiual paritition (in the block of partition offsets), up
755          to 3 extra bytes could have been read from the source buffer.
756          The code will not take any action regardless of the contents
757          of these undefined bytes, as the truncated buffer is detected
758          immediately following the read based on the calculated
759          starting position of the coefficient partition.
760
761    - Fix potential error concealment crash when the very first frame
762      is missing or corrupt (a609be5)
763
764    - Fix significant artifacts in error concealment (a4c2211, 99d870a)
765
766    - Revert 1-pass CBR rate control changes (e961317)
767      Further testing showed this change produced undesirable visual
768      artifacts, rolling back for now.
769
770
7712011-08-02 v0.9.7 "Cayuga"
772  Our third named release, focused on a faster, higher quality, encoder.
773
774  - Upgrading:
775    This release is backwards compatible with Aylesbury (v0.9.5) and
776    Bali (v0.9.6). Users of older releases should refer to the Upgrading
777    notes in this document for that release.
778
779  - Enhancements:
780          Stereo 3D format support for vpxenc
781          Runtime detection of available processor cores.
782          Allow specifying --end-usage by enum name
783          vpxdec: test for frame corruption
784          vpxenc: add quantizer histogram display
785          vpxenc: add rate histogram display
786          Set VPX_FRAME_IS_DROPPABLE
787          update configure for ios sdk 4.3
788          Avoid text relocations in ARM vp8 decoder
789          Generate a vpx.pc file for pkg-config.
790          New ways of passing encoded data between encoder and decoder.
791
792  - Speed:
793      This release includes across-the-board speed improvements to the
794      encoder. On x86, these measure at approximately 11.5% in Best mode,
795      21.5% in Good mode (speed 0), and 22.5% in Realtime mode (speed 6).
796      On ARM Cortex A9 with Neon extensions, real-time encoding of video
797      telephony content is 35% faster than Bali on single core and 48%
798      faster on multi-core. On the NVidia Tegra2 platform, real time
799      encoding is 40% faster than Bali.
800
801      Decoder speed was not a priority for this release, but improved
802      approximately 8.4% on x86.
803
804          Reduce motion vector search on alt-ref frame.
805          Encoder loopfilter running in its own thread
806          Reworked loopfilter to precalculate more parameters
807          SSE2/SSSE3 optimizations for build_predictors_mbuv{,_s}().
808          Make hor UV predict ~2x faster (73 vs 132 cycles) using SSSE3.
809          Removed redundant checks
810          Reduced structure sizes
811          utilize preload in ARMv6 MC/LPF/Copy routines
812          ARM optimized quantization, dfct, variance, subtract
813          Increase chrow row alignment to 16 bytes.
814          disable trellis optimization for first pass
815          Write SSSE3 sub-pixel filter function
816          Improve SSE2 half-pixel filter funtions
817          Add vp8_sub_pixel_variance16x8_ssse3 function
818          Reduce unnecessary distortion computation
819          Use diamond search to replace full search
820          Preload reference area in sub-pixel motion search (real-time mode)
821
822  - Quality:
823      This release focused primarily on one-pass use cases, including
824      video conferencing. Low latency data rate control was significantly
825      improved, improving streamability over bandwidth constrained links.
826      Added support for error concealment, allowing frames to maintain
827      visual quality in the presence of substantial packet loss.
828
829          Add rc_max_intra_bitrate_pct control
830          Limit size of initial keyframe in one-pass.
831          Improve framerate adaptation
832          Improved 1-pass CBR rate control
833          Improved KF insertion after fades to still.
834          Improved key frame detection.
835          Improved activity masking (lower PSNR impact for same SSIM boost)
836          Improved interaction between GF and ARFs
837          Adding error-concealment to the decoder.
838          Adding support for independent partitions
839          Adjusted rate-distortion constants
840
841
842  - Bug Fixes:
843          Removed firstpass motion map
844          Fix parallel make install
845          Fix multithreaded encoding for 1 MB wide frame
846          Fixed iwalsh_neon build problems with RVDS4.1
847          Fix semaphore emulation, spin-wait intrinsics on Windows
848          Fix build with xcode4 and simplify GLOBAL.
849          Mark ARM asm objects as allowing a non-executable stack.
850          Fix vpxenc encoding incorrect webm file header on big endian
851
852
8532011-03-07 v0.9.6 "Bali"
854  Our second named release, focused on a faster, higher quality, encoder.
855
856  - Upgrading:
857    This release is backwards compatible with Aylesbury (v0.9.5). Users
858    of older releases should refer to the Upgrading notes in this
859    document for that release.
860
861  - Enhancements:
862      vpxenc --psnr shows a summary when encode completes
863      --tune=ssim option to enable activity masking
864      improved postproc visualizations for development
865      updated support for Apple iOS to SDK 4.2
866      query decoder to determine which reference frames were updated
867      implemented error tracking in the decoder
868      fix pipe support on windows
869
870  - Speed:
871      Primary focus was on good quality mode, speed 0. Average improvement
872      on x86 about 40%, up to 100% on user-generated content at that speed.
873      Best quality mode speed improved 35%, and realtime speed 10-20%. This
874      release also saw significant improvement in realtime encoding speed
875      on ARM platforms.
876
877        Improved encoder threading
878        Dont pick encoder filter level when loopfilter is disabled.
879        Avoid double copying of key frames into alt and golden buffer
880        FDCT optimizations.
881        x86 sse2 temporal filter
882        SSSE3 version of fast quantizer
883        vp8_rd_pick_best_mbsegmentation code restructure
884        Adjusted breakout RD for SPLITMV
885        Changed segmentation check order
886        Improved rd_pick_intra4x4block
887        Adds armv6 optimized variance calculation
888        ARMv6 optimized sad16x16
889        ARMv6 optimized half pixel variance calculations
890        Full search SAD function optimization in SSE4.1
891        Improve MV prediction accuracy to achieve performance gain
892        Improve MV prediction in vp8_pick_inter_mode() for speed>3
893
894  - Quality:
895      Best quality mode improved PSNR 6.3%, and SSIM 6.1%. This release
896      also includes support for "activity masking," which greatly improves
897      SSIM at the expense of PSNR. For now, this feature is available with
898      the --tune=ssim option. Further experimentation in this area
899      is ongoing. This release also introduces a new rate control mode
900      called "CQ," which changes the allocation of bits within a clip to
901      the sections where they will have the most visual impact.
902
903        Tuning for the more exact quantizer.
904        Relax rate control for last few frames
905        CQ Mode
906        Limit key frame quantizer for forced key frames.
907        KF/GF Pulsing
908        Add simple version of activity masking.
909        make rdmult adaptive for intra in quantizer RDO
910        cap the best quantizer for 2nd order DC
911        change the threshold of DC check for encode breakout
912
913  - Bug Fixes:
914      Fix crash on Sparc Solaris.
915      Fix counter of fixed keyframe distance
916      ARNR filter pointer update bug fix
917      Fixed use of motion percentage in KF/GF group calc
918      Changed condition for using RD in Intra Mode
919      Fix encoder real-time only configuration.
920      Fix ARM encoder crash with multiple token partitions
921      Fixed bug first cluster timecode of webm file is wrong.
922      Fixed various encoder bugs with odd-sized images
923      vp8e_get_preview fixed when spatial resampling enabled
924      quantizer: fix assertion in fast quantizer path
925      Allocate source buffers to be multiples of 16
926      Fix for manual Golden frame frequency
927      Fix drastic undershoot in long form content
928
929
9302010-10-28 v0.9.5 "Aylesbury"
931  Our first named release, focused on a faster decoder, and a better encoder.
932
933  - Upgrading:
934    This release incorporates backwards-incompatible changes to the
935    ivfenc and ivfdec tools. These tools are now called vpxenc and vpxdec.
936
937    vpxdec
938      * the -q (quiet) option has been removed, and replaced with
939        -v (verbose). the output is quiet by default. Use -v to see
940        the version number of the binary.
941
942      * The default behavior is now to write output to a single file
943        instead of individual frames. The -y option has been removed.
944        Y4M output is the default.
945
946      * For raw I420/YV12 output instead of Y4M, the --i420 or --yv12
947        options must be specified.
948
949          $ ivfdec -o OUTPUT INPUT
950          $ vpxdec --i420 -o OUTPUT INPUT
951
952      * If an output file is not specified, the default is to write
953        Y4M to stdout. This makes piping more natural.
954
955          $ ivfdec -y -o - INPUT | ...
956          $ vpxdec INPUT | ...
957
958      * The output file has additional flexibility for formatting the
959        filename. It supports escape characters for constructing a
960        filename from the width, height, and sequence number. This
961        replaces the -p option. To get the equivalent:
962
963          $ ivfdec -p frame INPUT
964          $ vpxdec --i420 -o frame-%wx%h-%4.i420 INPUT
965
966    vpxenc
967      * The output file must be specified with -o, rather than as the
968        last argument.
969
970          $ ivfenc <options> INPUT OUTPUT
971          $ vpxenc <options> -o OUTPUT INPUT
972
973      * The output defaults to webm. To get IVF output, use the --ivf
974        option.
975
976          $ ivfenc <options> INPUT OUTPUT.ivf
977          $ vpxenc <options> -o OUTPUT.ivf --ivf INPUT
978
979
980  - Enhancements:
981      ivfenc and ivfdec have been renamed to vpxenc, vpxdec.
982      vpxdec supports .webm input
983      vpxdec writes .y4m by default
984      vpxenc writes .webm output by default
985      vpxenc --psnr now shows the average/overall PSNR at the end
986      ARM platforms now support runtime cpu detection
987      vpxdec visualizations added for motion vectors, block modes, references
988      vpxdec now silent by default
989      vpxdec --progress shows frame-by-frame timing information
990      vpxenc supports the distinction between --fps and --timebase
991      NASM is now a supported assembler
992      configure: enable PIC for shared libs by default
993      configure: add --enable-small
994      configure: support for ppc32-linux-gcc
995      configure: support for sparc-solaris-gcc
996
997  - Bugs:
998      Improve handling of invalid frames
999      Fix valgrind errors in the NEON loop filters.
1000      Fix loopfilter delta zero transitions
1001      Fix valgrind errors in vp8_sixtap_predict8x4_armv6().
1002      Build fixes for darwin-icc
1003
1004  - Speed:
1005      20-40% (average 28%) improvement in libvpx decoder speed,
1006      including:
1007        Rewrite vp8_short_walsh4x4_sse2()
1008        Optimizations on the loopfilters.
1009        Miscellaneous improvements for Atom
1010        Add 4-tap version of 2nd-pass ARMv6 MC filter.
1011        Improved multithread utilization
1012        Better instruction choices on x86
1013        reorder data to use wider instructions
1014        Update NEON wide idcts
1015        Make block access to frame buffer sequential
1016        Improved subset block search
1017        Bilinear subpixel optimizations for ssse3.
1018        Decrease memory footprint
1019
1020      Encoder speed improvements (percentage gain not measured):
1021        Skip unnecessary search of identical frames
1022        Add SSE2 subtract functions
1023        Improve bounds checking in vp8_diamond_search_sadx4()
1024        Added vp8_fast_quantize_b_sse2
1025
1026  - Quality:
1027      Over 7% overall PSNR improvement (6.3% SSIM) in "best" quality
1028      encoding mode, and up to 60% improvement on very noisy, still
1029      or slow moving source video
1030
1031        Motion compensated temporal filter for Alt-Ref Noise Reduction
1032        Improved use of trellis quantization on 2nd order Y blocks
1033        Tune effect of motion on KF/GF boost in two pass
1034        Allow coefficient optimization for good quality speed 0.
1035        Improved control of active min quantizer for two pass.
1036        Enable ARFs for non-lagged compress
1037
10382010-09-02 v0.9.2
1039  - Enhancements:
1040      Disable frame dropping by default
1041      Improved multithreaded performance
1042      Improved Force Key Frame Behaviour
1043      Increased rate control buffer level precision
1044      Fix bug in 1st pass motion compensation
1045      ivfenc: correct fixed kf interval, --disable-kf
1046  - Speed:
1047      Changed above and left context data layout
1048      Rework idct calling structure.
1049      Removed unnecessary MB_MODE_INFO copies
1050      x86: SSSE3 sixtap prediction
1051      Reworked IDCT to include reconstruction (add) step
1052      Swap alt/gold/new/last frame buffer ptrs instead of copying.
1053      Improve SSE2 loopfilter functions
1054      Change bitreader to use a larger window.
1055      Avoid loopfilter reinitialization when possible
1056  - Quality:
1057      Normalize quantizer's zero bin and rounding factors
1058      Add trellis quantization.
1059      Make the quantizer exact.
1060      Updates to ARNR filtering algorithm
1061      Fix breakout thresh computation for golden & AltRef frames
1062      Redo the forward 4x4 dct
1063      Improve the accuracy of forward walsh-hadamard transform
1064      Further adjustment of RD behaviour with Q and Zbin.
1065  - Build System:
1066      Allow linking of libs built with MinGW to MSVC
1067      Fix target auto-detection on mingw32
1068      Allow --cpu= to work for x86.
1069      configure: pass original arguments through to make dist
1070      Fix builds without runtime CPU detection
1071      msvs: fix install of codec sources
1072      msvs: Change devenv.com command line for better msys support
1073      msvs: Add vs9 targets.
1074      Add x86_64-linux-icc target
1075  - Bugs:
1076      Potential crashes on older MinGW builds
1077      Fix two-pass framrate for Y4M input.
1078      Fixed simple loop filter, other crashes on ARM v6
1079      arm: fix missing dependency with --enable-shared
1080      configure: support directories containing .o
1081      Replace pinsrw (SSE) with MMX instructions
1082      apple: include proper mach primatives
1083      Fixed rate control bug with long key frame interval.
1084      Fix DSO link errors on x86-64 when not using a version script
1085      Fixed buffer selection for UV in AltRef filtering
1086
1087
10882010-06-17 v0.9.1
1089  - Enhancements:
1090      * ivfenc/ivfdec now support YUV4MPEG2 input and pipe I/O
1091      * Speed optimizations
1092  - Bugfixes:
1093      * Rate control
1094      * Prevent out-of-bounds accesses on invalid data
1095  - Build system updates:
1096      * Detect toolchain to be used automatically for native builds
1097      * Support building shared libraries
1098      * Better autotools emulation (--prefix, --libdir, DESTDIR)
1099  - Updated LICENSE
1100      * http://webmproject.blogspot.com/2010/06/changes-to-webm-open-source-license.html
1101
1102
11032010-05-18 v0.9.0
1104  - Initial open source release. Welcome to WebM and VP8!
1105
1106