vvcdec: add vvc decoder

vvc decoder plug-in to avcodec.
split frames into slices/tiles and send them to vvc_thread for further decoding
reorder and wait for the frame decoding to be done and output the frame

Features:
    + Support I, P, B frames
    + Support 8/10/12 bits, chroma 400, 420, 422, and 444 and range extension
    + Support VVC new tools like MIP, CCLM, AFFINE, GPM, DMVR, PROF, BDOF, LMCS, ALF
    + 295 conformace clips passed
    - Not support RPR, IBC, PALETTE, and other minor features yet

Performance:
    C code FPS on an i7-12700K (x86):
        BQTerrace_1920x1080_60_10_420_22_RA.vvc      93.0
        Chimera_8bit_1080P_1000_frames.vvc          184.3
        NovosobornayaSquare_1920x1080.bin           191.3
        RitualDance_1920x1080_60_10_420_32_LD.266   150.7
        RitualDance_1920x1080_60_10_420_37_RA.266   170.0
        Tango2_3840x2160_60_10_420_27_LD.266         33.7

    C code FPS on a M1 Mac Pro (ARM):
        BQTerrace_1920x1080_60_10_420_22_RA.vvc     58.7
        Chimera_8bit_1080P_1000_frames.vvc          153.3
        NovosobornayaSquare_1920x1080.bin           150.3
        RitualDance_1920x1080_60_10_420_32_LD.266   105.0
        RitualDance_1920x1080_60_10_420_37_RA.266   133.0
        Tango2_3840x2160_60_10_420_27_LD.266        21.7

    Asm optimizations still working in progress. please check
    https://github.com/ffvvc/FFmpeg/wiki#performance-data for the latest

Contributors (based on code merge order):
    Nuo Mi <nuomi2021@gmail.com>
    Xu Mu <toxumu@outlook.com>
    Frank Plowman <post@frankplowman.com>
    Shaun Loo <shaunloo10@gmail.com>
    Wu Jianhua <toqsxw@outlook.com>

Thank you for reporting issues and providing performance reports:
    Łukasz Czech <lukaszcz18@wp.pl>
    Xu Fulong <839789740@qq.com>

Thank you for providing review comments:
    Ronald S. Bultje <rsbultje@gmail.com>
    James Almer <jamrial@gmail.com>
    Andreas Rheinhardt <andreas.rheinhardt@outlook.com>

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add CTU thread logical

This is the main entry point for the CTU (Coding Tree Unit) decoder.
The code will divide the CTU decoder into several stages.
It will check the stage dependencies and run the stage decoder.

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add CTU parser

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add dsp init and inv transform

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add LMCS, Deblocking, SAO, and ALF filters

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add intra prediction

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add inv transform 1d

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add inter prediction

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add motion vector decoder

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add reference management

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add cabac decoder

add Context-based Adaptive Binary Arithmetic Coding (CABAC) decoder

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add parameter parser for sps, pps, ph, sh

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

vvcdec: add vvc_data

Co-authored-by: Xu Mu <toxumu@outlook.com>
Co-authored-by: Frank Plowman <post@frankplowman.com>
Co-authored-by: Shaun Loo <shaunloo10@gmail.com>
Co-authored-by: Wu Jianhua <toqsxw@outlook.com>

doc/filters/drawtext: apply second level of escaping on colon

Fix escaping on example.

doc/filters/setpts: add random jitter generation example

lavu/eval: add randomi function to compute random value in interval

lavfi/setpts: use macro to access variables array

lavfi: add quirc filter

lavfi: add qrencode source and filter

lavfi: introduce textutils

Generalize drawtext utilities to make them usable in other filters.
This will be needed to introduce the QR code source and filter without
duplicating functionality.

lavfi/overlay: factorize definition of planar and packed blending functions

avcodec/d3d12va_mpeg2: change the type for the ID3D12Resource_Map input data argument

Fixes -Wincompatible-pointer-types warnings.

Reviewed-by: Wu Jianhua <toqsxw@outlook.com>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/d3d12va_mpeg2: remove unused variables

Reviewed-by: Wu Jianhua <toqsxw@outlook.com>
Signed-off-by: James Almer <jamrial@gmail.com>

configure: correct lensfun header function check

The function lf_db_new was deprecated in lensfun 09dcd3e7ad in 2017
in favour of lf_db_create. The AVfilter was adjusted in 8b78eb312d
but the configure check wasn't changed.

lensfun removed lf_db_new declaration in lf 35c0017593 so configure
fails to find lensfun.

configure: update copyright year

avfilter/vf_format: allow empty pix_fmts list

Which will impose no restriction. This makes sense when using e.g.
`color_ranges=pc` to limit the color range, without also limiting the
pixel format.

avfilter/vf_format: add color_ranges/spaces

Needed for fftools/ffmpeg_filter to be able to force specific output
formats via the filter chain.

avfilter/vf_format: re-use AVFilterFormats for pix_fmt parsing

Rewrite the format parsing code to make it more easily generalizable. In
particular, `invert_formats` does not depend on the type of format list
passed to it, which allows me to re-use this helper in an upcoming
commit.

Slightly shortens the code, at the sole cost of doing several malloc
(ff_add_format) instead of a single malloc.

avfilter/buffersink: add color space/range accessors

Needed for fftools.

avfilter/vf_libplacebo: switch to colorspace negotiation API

avfilter/vf_libplacebo: don't force dovi frames to bt.2020-ncl

This is at odds with the YUV matrix negotiation API, in which such
dynamic changes in YUV encoding are no longer easily possible. There is
also no really strong motivating reason to do this, since the choice of
YUV matrix is essentially arbitrary and not actually related to the
Dolby Vision decoding process.

avfilter/vf_zscale: switch to colorspace negotiation API

Following the same design as vf_scale.

avfilter/vf_zscale: remove unused variables

Only assigned, never read.

vf_scale: use colorspace negotiation API

This filter will always accept any input format, even if the user sets
a specific in_range/in_color_matrix. This is to preserve status quo with
current behavior, where passing a specific in_color_matrix merely
overrides the incoming frames' attributes. (Use `vf_format` to force
a specific input range)

Because changing colorspace and color_range now requires reconfiguring
the link, we can lift sws_setColorspaceDetails out of scale_frame and
into config_props. (This will also get re-called if the input frame
properties change)

fftools/ffmpeg_filter: configure buffersrc with csp/range

Propagates input metadata to the input filter graph.

avfilter/buffersrc: add color_space/range parameters

To allow adding proper negotiation, in particular, to fftools.

These values will simply be negotiated downstream for YUV formats, and
ignored otherwise.

fftools/ffmpeg_filter: don't clear buffersrc params

This memset is bogus, it accomplishes nothing in the best case and
regresses future additions to AVBufferSrcParameters in the worst case.

avfilter: add negotiation API for color space/range

Motivated by YUVJ removal. This change will allow full negotiation
between color ranges and matrices as needed. By default, all ranges and
matrices are marked as supported.

Because grayscale formats are currently handled very inconsistently (and
in particular, assumed as forced full-range by swscale), we exclude them
from negotiation altogether for the time being, to get this API merged.

After filter negotiation is available, we can relax the
grayscale-is-forced-jpeg restriction again, when it will be more
feasible to do so without breaking a million test cases.

Note that this commit updates one FATE test as a consequence of the
sanity fallback for non-YUV formats. In particular, the test case now
writes rgb24(pc, gbr/unspecified/unspecified) to the matroska file,
instead of rgb24(unspecified/unspecified/unspecified) as before.

avfilter/formats: set audio fmt lists for vaf filters

Currently, the logic inside the FF_FILTER_FORMATS_QUERY_FUNC branch
prevents this code from running in the event that we have a filter with
a single video input and a single audio output, as the resulting audio
output link will not have its channel counts / samplerates correctly
initialized to their default values, possibly triggering a segfault
downstream.

An example of such a filter is vaf_spectrumsynth. Although this
particular filter already sets up the channel counts and samplerates as
part of the query function and therefore avoids triggering this bug, the
bug still exists in principle. (And importantly, sets a wrong precedent)

avfilter: always call ff_default_query_formats

Even if a query func is set. This is safe to do, because
ff_default_query_formats is documented not to touch any filter lists
that were already set by the query func.

The reason to do this is because it allows us to extend
AVFilterFormatsConfig without having to touch every filter in existence.
An alternative implementation of this commit would be to explicitly add
a `ff_default_query_formats` call at the end of every query_formats
function, but that would end up functionally equivalent to this change
while touching a whole lot more code paths for no reason.

As a bonus, eliminates some code/logic duplication from this function.

avfilter/formats: document ff_default_query_formats

In particular, make it clear that this function will not touch format
lists which were already set by the caller before calling into this
function.

Add jpeg2000 subdescriptor (V2).

Signed-off-by: Cedric Le Barz <clebarz@ektacom.com>

avcodec/mpegvideo_enc: Use ptrdiff_t for stride

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/mpegvideo_enc: Dont copy beyond the image

Fixes: out of array access
Fixes: tickets/10754/poc17ffmpeg

Discovered by Zeng Yunxiang.

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_minterpolate: Check pts before division

Fixes: FPE
Fixes: tickets/10758/poc20ffmpeg

Discovered by Zeng Yunxiang

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/avf_showwaves: Check history_nb_samples

Fixes: out of array access
Fixes: tickets/10756/poc18ffmpeg

Discovered by Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

libavfilter/vf_dnn_detect: Add two outputs ssd support

For this kind of model, we can directly use its output as final result
just like ssd model. The difference is that it splits output into two
tensors. [x_min, y_min, x_max, y_max, confidence] and [lable_id].

Model example refer to: https://github.com/openvinotoolkit/open_model_zoo/tree/master/models/intel/person-detection-0106

Signed-off-by: Wenbin Chen <wenbin.chen@intel.com>
Reviewed-by: Guo Yejun <yejun.guo@intel.com>

libavfilter/dnn_backend_openvino: Add dynamic output support

Add dynamic outputs support. Some models don't have fixed output size.
Its size changes according to result. Now openvino can run these kinds of
models.

Signed-off-by: Wenbin Chen <wenbin.chen@intel.com>
Reviewed-by: Guo Yejun <yejun.guo@intel.com>

avformat/flacdec: Avoid double AVERRORS

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_vidstabdetect: Avoid double AVERRORS

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/vaapi_encode: Avoid double AVERRORS

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_swaprect: round coordinates down

Fixes: out of array access:
Fixes: tickets/10745/poc12ffmpeg

Found-by: Li Zeyuan and Zeng Yunxiang.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_swaprect: Use height for vertical variables

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_swaprect: assert that rectangles are within memory

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/af_alimiter: Check nextpos before use

Fixes: out of array read
Fixes: tickets/10744/poc11ffmpeg

Found-by: Li Zeyuan and Zeng Yunxiang.
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/f_reverse: Apply PTS compensation only when pts is available

Fixes: out of array access
Fixes: tickets/10753/poc16ffmpeg

Regression since: 45dc668aea0edac34969b5a1ff76cf9ad3a09be1
Found-by: Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/af_stereowiden: Check length

Fixes: out of array access
Fixes: tickets/10746/poc13ffmpeg

Found-by: Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/mov: Fix MSAN issue with stsd_id

Fixes: use of uninitialized value
Fixes: bbb-320x240-2video-2audio.mp4

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/jpegxl_parser: Check get_vlc2()

Fixes: shift exponent -1 is negative
Fixes: 63889/clusterfuzz-testcase-minimized-ffmpeg_DEMUXER_fuzzer-6009343056936960

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_weave: Fix odd height handling

Fixes: out of array access
Fixes: tickets/10743/poc10ffmpeg

Found-by: Zeng Yunxiang and Li Zeyuan
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/edge_template: Fix small inputs with gaussian_blur()

Fixes: out of array access
Fixes: Ticket10699
Fixes: poc5ffmpeg

Found-by: Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/af_stereotools: round-up max size of buffer

Fixes: out of array access
Fixes: tickets/10747/poc14ffmpeg

Found-by: Zeng Yunxiang and Song Jiaxuan
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/vf_gradfun: Do not overread last line

The code works in steps of 2 lines and lacks support for odd height
Implementing odd height support is better but for now this fixes the
out of array access

Fixes: out of array access
Fixes: tickets/10702/poc6ffmpe

Found-by: Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/avf_showspectrum: fix off by 1 error

Fixes: out of array access
Fixes: tickets/10749/poc15ffmpeg

Regression since: 81df787b53eb5c6433731f6eaaf7f2a94d8a8c80

Found-by: Zeng Yunxiang
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avfilter/af_stereowiden: Round length to nearest

Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/leaddec: Check remaining bits in decode_block()

Fixes: Timeout
Fixes: 64163/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_LEAD_fuzzer-6418925835124736

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/jpegxl_parser: Add padding to cs_buffer

Fixes: out of array access
Fixes: 64081/clusterfuzz-testcase-minimized-ffmpeg_DEMUXER_fuzzer-6151006496620544

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tools/target_dec_fuzzer: Adjust threshold for QPEG

Fixes: 63712/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_QPEG_fuzzer-456353017272729
Fixes: Timeout

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/mov: do not set sign bit for chunk_offsets

Fixes: signed integer overflow: 2314885530818453536 - -7412889664301817824 cannot be represented in type 'long'
Fixes: 64296/clusterfuzz-testcase-minimized-ffmpeg_dem_MOV_fuzzer-6304027146846208

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/jpeglsdec: Check Jpeg-LS LSE

Fixes: signed integer overflow: 2147478526 + 33924 cannot be represented in type 'int'
Fixes: shift exponent 32 is too large for 32-bit type 'unsigned int'
Fixes: 64243/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_JPEGLS_fuzzer-5195717848989696

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

tools/target_dec_fuzzer: Adjust Threshold for VP6A

Fixes: Timeout
Fixes: 64220/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VP6A_fuzzer-5653856213925888

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/osq: Implement flush()

Fixes: out of array access
Fixes: 62164/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6227491892887552
Fixes: 62164/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6268561729126400
Fixes: 62164/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6414805046788096
Fixes: 62164/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6538151088488448
Fixes: 62164/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-6608131540779008

Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/hevc: Add init for sao_edge_filter

Forgot to init c->sao_edge_filter[idx] when idx=0/1/2/3.
After this patch, the speedup of decoding H265 4K 30FPS
30Mbps on 3A6000 is about 7% (42fps==>45fps).

Change-Id: I521999b397fa72b931a23c165cf45f276440cdfb
Reviewed-by: yinshiyou-hf@loongson.cn
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/bitpacked_dec: optimize bitpacked_decode_yuv422p10

Rework the code a bit to speed up the 10-bit bitpacked decoding
routine.  This is probably about as fast as I can get it without
switching to assembly language.

Demonstratable with:

./ffmpeg -f lavfi -i "smptehdbars=size=3840x2160" -c bitpacked -f image2 -frames:v 1 source.yuv
./ffmpeg -f bitpacked -pix_fmt yuv422p10le -s 3840x2160 -c:v bitpacked -i source.yuv -pix_fmt yuv422p10le out.yuv

On my development system, it went from 80ms for a 2160p frame
down to 20ms (i.e. a 4X speedup).  Good enough for now, I hope...

Comments from Marton:

Originally on my system better performance could be achieved by simply
switching to the cached bitstream reader, but for Devin it was slower than
his direct byte operations.

I changed the order of writing output from u/y/v/y to u/v/y/y, and that made
the code faster than the cached bitstream reader on my system as well.

TIMER measurement of the decode loop on Ryzen 5 3600 with command line:

./ffmpeg -stream_loop 256 -threads 1 -f bitpacked -pix_fmt yuv422p10le -s 3840x2160 -c:v bitpacked -i source.yuv -pix_fmt yuv422p10le -f null none -loglevel error

Before: 823204127 decicycles in YUV,     256 runs,      0 skips
After:  315070524 decicycles in YUV,     256 runs,      0 skips

Signed-off-by: Devin Heitmueller <dheitmueller@ltnglobal.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

avcodec/mjpegdec: avoid indirection when accessing avctx

Signed-off-by: Marton Balint <cus@passwd.hu>

avcodec/mjpegdec: use memset to clear alpha

Signed-off-by: Marton Balint <cus@passwd.hu>

tests/dca: remove usage of deprecated request_channel_layout option

Signed-off-by: James Almer <jamrial@gmail.com>

tests/truehd: remove usage of deprecated request_channel_layout option

Signed-off-by: James Almer <jamrial@gmail.com>

avformat/riffenc: don't force WAVEFORMATEXTENSIBLE for flt/dbl LPCM

2c2a167ca7 forced WAVEFORMATEXTENSIBLE for all LPCM streams with greater
than 16 bits per sample. However, WAVEFORMATEX allows IEEE Float samples
or any depth where raw depth == coded depth, see Remarks section at
https://learn.microsoft.com/en-us/windows/win32/api/mmreg/ns-mmreg-waveformatex
and samples M1F1-float32-AFsp, M1F1-float64-AFsp at
https://www.mmsp.ece.mcgill.ca/Documents/AudioFormats/WAVE/Samples.html

There are hardware devices and likely software players requiring float samples
that fail to qualify files with WAVEFORMATEXTENSIBLE headers.

avfilter/vf_showinfo: add udu_sei_as_ascii option

Some encoders (e.g., libx264) dump encoder configuration as user
data unregistered SEI message. This option try to print it as
ascii character when possible.

Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>

avcodec/jpegxl_parser: check ANS cluster alphabet size vs bundle size

The specification doesn't mention that clusters cannot have alphabet
sizes greater than 1 << bundle->log_alphabet_size, but the reference
implementation rejects these entropy streams as invalid, so we should
too. Refusing to do so can overflow a stack variable that should be
large enough otherwise.

Fixes #10738.

Found-by: Zeng Yunxiang and Li Zeyuan
Signed-off-by: Leo Izen <leo.izen@gmail.com>

x86/takdsp: add missing wrappers to AVX2 functions

Fixes compilation with old yasm.

Signed-off-by: James Almer <jamrial@gmail.com>

configure: Enable section_data_rel_ro for FreeBSD and NetBSD aarch64 / arm

Fixes the build. It's a requirement when utilizing PIE.

Signed-off-by: Brad Smith <brad@comstyle.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Remove myself from maintainers.

x86/takdsp: add avx2 versions of all functions

On an Intel Core i7 12700k:

decorrelate_ls_c: 814.3
decorrelate_ls_sse2: 165.8
decorrelate_ls_avx2: 101.3
decorrelate_sf_c: 1602.6
decorrelate_sf_sse4: 640.1
decorrelate_sf_avx2: 324.6
decorrelate_sm_c: 1564.8
decorrelate_sm_sse2: 379.3
decorrelate_sm_avx2: 203.3
decorrelate_sr_c: 785.3
decorrelate_sr_sse2: 176.3
decorrelate_sr_avx2: 99.8

Tested-by: Lynne <dev@lynne.ee>
Signed-off-by: James Almer <jamrial@gmail.com>

avcodec/libjxlenc: Set AV_CODEC_CAP_DR1

This encoder uses ff_get_encode_buffer() to allocate the packet buffer.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>

avcodec/libjxlenc: Don't refer to decoder in comments

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>

avcodec/libjxldec: emit proper PTS to decoded AVFrame

If a sequence of JXL images is encapsulated in a container that has PTS
information, we should use the PTS information from the container. At
this time there is no container that does this, but if JPEG XL support
is ever added to NUT, AVTransport, or some other container, this commit
should allow the PTS information those containers provide to work as
expected.

Signed-off-by: Leo Izen <leo.izen@gmail.com>

avcodec/libjxlenc: accept rgbf32 and rgbaf32 frames

These pixel formats have always been supported by libjxl, but at the
time this plugin was written, they were not in FFmpeg yet. Now that
they are in FFmpeg, we should support them.

Signed-off-by: Leo Izen <leo.izen@gmail.com>

avcodec/libjxldec: produce rgbf32 and rgbaf32 frames

These pixel formats have always been supported by libjxl, but at the
time this plugin was written, they were not in FFmpeg yet. Now that
they are in FFmpeg, we should support them.

Signed-off-by: Leo Izen <leo.izen@gmail.com>

avcodec/pngdec: improve handling of bad cICP range tags

FFmpeg doesn't support tv-range RGB throughout most of its pipeline, so
we should keep the warning. However, in case something does support it
we should at least keep it tagged properly. Additionally, the encoder
writes this tag if the space is tagged as such so this makes a round
trip work as it should.

Also, PNG doesn't support nonzero matrices but we only warn and ignore
in that case, so we have no reason to error out for illegal cICP ranges
either (i.e. greater than 1).

Signed-off-by: Leo Izen <leo.izen@gmail.com>
Reported-by: Kacper Michajłow <kasper93@gmail.com>

lavc/takdsp: R-V V decorrelate_sm

C908:
decorrelate_sm_c: 130.0
decorrelate_sm_rvv_i32: 43.2

Signed-off-by: Rémi Denis-Courmont <remi@remlab.net>
(with minor changes)

all: Don't set AVClass.item_name to its default value

Unnecessary since acf63d5350adeae551d412db699f8ca03f7e76b9;
also avoids relocations.

Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>

avcodec/takdsp: fix const correctness

Signed-off-by: James Almer <jamrial@gmail.com>

checkasm/takdsp: add decorrelate_sf test

Signed-off-by: James Almer <jamrial@gmail.com>

checkasm: Fix the takdsp tests

For memcpy and memcmp, we need to multiply by the element size,
otherwise we're copying and comparing only a fraction of the buffer.

For decorrelate_sr, the buffer p1 is the one that is mutated;
copy and check p1 instead of p2.

For decorrelate_sm, both buffers are mutated, so copy and check
both of them.

For decorrelate_sm, the memcpy initialization of p1 and p1_2 was
reversed - p1 is filled with randomize, but then memcpy copies from
p1_2 to p1. As p1_2 is uninitialized at this point, clang concluded
that the copy was bogus and omitted it entirely, triggering failures
in this test on x86 (where there was an existing assembly implementation
to test).

Signed-off-by: Martin Storsjö <martin@martin.st>

fftools/ffmpeg: remove deprecated -[av]bsf

They were marked as deprecated over 10 years ago.

fftools/ffmpeg: mark -vsync for future removal

It has already been deprecated over a year ago.

fftools/ffmpeg_opt: fix -dn flags

It's a data, not video option.

fftools/ffmpeg_opt: print a section for data-stream options

Only show it with show_advanced (triggered by -h long), since data
streams themselves are an advanced topic.