vc2enc_dwt: use 32 bit coefficients by default

The problem is that with particularly complex images and especially at
high bit depths and 5-level transforms the coefficients would overflow,
causing huge artifacts to appear. This was discovered thanks to the fate
tests, which will have to be redone as this fixes a multitude of
problems and increases PSNR.

There is a slight performance drop associated with this change, making
the encoder slower by 1.15 times, however this is necessary in order to
avoid undefined behavior and overflows.

It would be worth to template the transforms to keep the performance for
8 bit images as 32 bit coefficients are unnecessary for that case, but
the primary use of the encoder is to encode video at 10 bits.

Reviewed-by: Christophe Gisquet <christophe.gisquet@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avcodec/dxva2_h264: fix slice offset in long slice struct after ca2f19b9

The GetBitContext now includes the NAL header, which offsets the slice by
one byte, which needs to be accounted for here.

vaapi_h264: Fix bit offset of slice data.

Commit ca2f19b9cc37be509d85f05c8f902860475905f8 modified the meaning of
H264SliceContext.gb: it is now initialised at the start of the NAL unit
header, rather than at the start of the slice header.  The VAAPI slice
decoder uses the offset after parsing to determine the offset of the
slice data in the bitstream, so with the changed meaning we no longer
need to add the extra byte to account for the NAL unit header because
it is now included directly.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

fate: Remove duplicate wmv8_x8intra.wmv test

Also temporary enable the test so we get updated fate failure statistics

Note, this does not work on all platforms, it fails on MIPS
and ml archives indicate it failed on x86 openbsd with some compilers as well

Reviewed-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

x86: lossless audio: SSE4 madd 32bits

The unique user so far is wmalossless 24bits. The few samples tested show an
order of 8, so more unrolling or an avx2 version do not make sense.

Timings: 68 -> 49 cycles

Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Merge commit 'ca8c7591735c0f80cc29e31e2e92cb10228e14c7'

* commit 'ca8c7591735c0f80cc29e31e2e92cb10228e14c7':
  intrax8: Remove mpegvideo dependency

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '6ebd06a9b2508747a135ee4c880d8f612e08932b'

* commit '6ebd06a9b2508747a135ee4c880d8f612e08932b':
  intrax8: Drop lots of pointless parentheses

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '9b57995cdd489a4cff51dcc1a1f08ac77ec5a58c'

* commit '9b57995cdd489a4cff51dcc1a1f08ac77ec5a58c':
  intrax8: Drop MB emulation code

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '9fa888c02801fff2e8817c24068f5296bbe60000'

* commit '9fa888c02801fff2e8817c24068f5296bbe60000':
  intrax8: Keep a reference to the decoder blocks

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'c2084ffcbfc11d1b6ed3a4a0df9cafd56fbb896f'

* commit 'c2084ffcbfc11d1b6ed3a4a0df9cafd56fbb896f':
  intrax8: Use the generic horizband function

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'b1268e0f032a3af3912fe3fb8d3855e12d7ea83b'

* commit 'b1268e0f032a3af3912fe3fb8d3855e12d7ea83b':
  intrax8: Pass macroblock coordinates to ff_intrax8_decode_picture

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'd0540fd02171a6233d2016b199d013299debf7e3'

* commit 'd0540fd02171a6233d2016b199d013299debf7e3':
  intrax8: Pass macroblock size to ff_intrax8_common_init

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '9f4d99138df434a73b097c997fb4cafc65f4ff54'

* commit '9f4d99138df434a73b097c997fb4cafc65f4ff54':
  fate: Add test for WMV2 with jframes

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

h265_parse: skip zero sized NAL units

Avoids extra error checks later on and/or invalid reads.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

h264: do not return an error when NAL parsing of extradata failed

Fixes AnnexB samples with broken extradata, but proper in-band parameter sets

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'ca2f19b9cc37be509d85f05c8f902860475905f8'

* commit 'ca2f19b9cc37be509d85f05c8f902860475905f8':
  h264: switch to h2645_parse for NAL parsing

Merged-by: Hendrik Leppkes <h.leppkes@gmail.com>

h2645_parse: allow partial escaping

This ports the fix from 033a533 to the new parser module in prepartion
of using it for the h264 decoder.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

h2645_parse: initialize the GetBitContext to the proper size

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

hevc: fix size condition in ptl parsing

When only one sublayer is present, no information is coded. Only when at least two
are present, all 8 sublayers are written.

Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

avformat/tee: Use ref instead copy in write_packet

Replace av_copy_packet and deprecated av_dup_packet by
creating reference using av_packet_ref.

Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Jan Sebechlebsky <sebechlebskyjan@gmail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

avformat/tee: Fix TeeSlave.bsfs pointer array size

TeeSlave.bsfs is array of pointers to AVBitStreamFilterContext,
so element size should be really size of a pointer, not size
of TeeSlave structure.

Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Jan Sebechlebsky <sebechlebskyjan@gmail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>

ffplay: force setting alsa buffer size

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

riff: add comment for g721,g723 codec tags

Signed-off-by: Piotr Bandurski <ami_stuff@o2.pl>

avformat/riff: add M102 FourCC

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

avcodec/m101: simplify 8bit code

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

avcodec/m101: remove unneeded zeroing of priv_data_size

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

avcodec/m101: Simplify if() condition

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

fate: fix dcadec test dependencies

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

configure: Only utvideo <= 15.3.0 is supported.

Fixes ticket #5511.

avcodec/avcodec: Move AV_CODEC_ID_M101 to the 2nd group of video codecs

Found-by: durandal_1707
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avformat/riff: add M101

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

avcodec: add M101 decoder

Fixes Ticket 2611

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

avcodec/nvenc: don't set profile in lossless mode

avcodec/dcadsp: use LOCAL_ALIGNED_32 instead of LOCAL_ALIGNED(32, ...)

avcodec: Add "sar" alias to "aspect" option of video encoders

It is impossible to pass "aspect" parameter to encoder from ffmpeg CLI
because option from lavc/options_table.h is eclipsed by option with same
name in ffmpeg_opt.c, which has different meaning (DAR, not SAR).

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

avfilter/graphparser: add '\r' as whitespace

for compatibility with platforms that treat it
as newline

Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>

avutil/parsing: add '\r' as whitespace

for compatibility with platforms that treat it
as newline

Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>

vc2enc: prevent random data

The slice prefix is 0 in the reference encoder and the decoder ignores it.
Writing 0 there seems like the best temporary solution.

The padding could have contained uninitialized data, but reference VC2
encoders put 0xFF there, hence the memset value.

Overall this allows producing bistreams with no random data for use by fate.

fate: Add test for vorbis encoder

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

avcodec/alac: fix 20-bit support

Signed-off-by: Paul B Mahol <onemda@gmail.com>

lavd/avfoundation: use AVCodecParameters

Fixes "Could not find codec parameters for stream" error (#5494)

Signed-off-by: Rick Kern <kernrj@gmail.com>

avfilter/vf_waveform: fix order of graticule scale items

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avfilter/avf_showcqt: add .

fate: Add test for broken SPS (Ticket 2580)

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

vc2enc: do not print the lavc version if the bitexact flag is enabled

Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>

avformat/dtsdec: reduce the scope of crctab

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

avformat/dtsdec: detect core-less streams

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

avformat/mpegtsenc: fix usage of AVStream.codec

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

avcodec/hevc_parser: fix packet_split function name

Fixes compilation of hevc_parser without hevc_decoder

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

avformat/h264dec: Avoid get_ue_golomb_31( ) due to ff_ prefixed table from golomb.c

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

avformat/utils: Do not detect video codecs when audio is expected

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

avformat/h264dec: Check pps_id/sps_id fields from parameter sets

Fixes a misdetection in wav.detected.as.h264.error.wav

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

lavc/videotoolboxenc: Set colorimetry values

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Support pixel aspect ratio

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Support for forced I-frames

Setting AVFrame.pic_type to AV_PICTURE_TYPE_I will force an I-frame.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: add concatentation properties

Add frames_before and frames_after as hints that there will be frames before
or after the frames produced in this session. This may help with
concatenation issues like bit rate spikes.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Add realtime encoding property

Hint to the encoder that encoding should be done in real-time, even at the
expense of quality.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Handle out-of-memory and fix memory leak

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Fix DTS

Some devices output an invalid DTS when B-frames aren't used. Using PTS for
the DTS in this case.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Fix AVCodecContext.has_b_frames usage.

Now set by the encoder, not used as an input parameter.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Add entropy setting

Add an entropy setting to choose between CAVLC and CABAC.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Require hardware encoding

Software encoding can be allowed by setting allow_sw to 1.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Handle hwaccel format as input

Handle AV_PIX_FMT_VIDEOTOOLBOX.
This results in better energy usage and faster encoding, especially on iOS.
When the buffer comes from the media server, no memcpy's are needed.

Signed-off-by: Rick Kern <kernrj@gmail.com>

lavc/videotoolboxenc: Use shared pixel buffer pool

This reduces the chance of a memcpy in the media server.

Signed-off-by: Rick Kern <kernrj@gmail.com>

avcodec/adpcm: fix decoding of stereo non 4-bit ADPCM IMA WAV

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avcodec/adpcm: replace char/short/long with int8_t/int16_t/int

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avcodec/dca: convert to AVCRC

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

lavf/mpegtsenc: Define service_type "HEVC digital television service".

Fixes ticket #5455.

rtpenc: packetizer for VC-2 HQ RTP payload format (draft v1)

avcodec/h264: Put the removed SPS handling code back

reverts one hunk from 7966ddfc0bb7ee87dc2606b7b146701db6f6c717
The new code from 7966ddfc0bb7ee87dc2606b7b146701db6f6c717 only covers extradata based SPS

Fixes: ffplay -ss 13 58af5798-fa2c-42a2-997d-dc8e49de2d8a.flv
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

Merge commit 'a7829a2a3f8e6ec0b9f2673c11f56916800aeb33'

* commit 'a7829a2a3f8e6ec0b9f2673c11f56916800aeb33':
  h264: reimplement 3aa661ec5 in a more explicit way

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

lavf/mpegtsenc: Fix stream_type for low sample rate MP2/MP3.

Merge commit 'add1467e5e447b79e8743a0b05c54dcf58c61dfe'

* commit 'add1467e5e447b79e8743a0b05c54dcf58c61dfe':
  svq3: drop the build dependency on the h264 decoder

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'a4d126dc59c39bb03e5e444432d1b27af26a45b4'

* commit 'a4d126dc59c39bb03e5e444432d1b27af26a45b4':
  svq3: eliminate remaining H264Context usage.

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '7bbdae81e895a49125dba58bad01b98389966c39'

* commit '7bbdae81e895a49125dba58bad01b98389966c39':
  svq3: move block_offset to SVQ3Context

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '1848a154a49d18c1f31f54ee75c7445dc49a7ecc'

* commit '1848a154a49d18c1f31f54ee75c7445dc49a7ecc':
  svq3: stop using H264Context.gb

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'ea6ab02a174bcc11f3eaa1b840c9a4c895968690'

* commit 'ea6ab02a174bcc11f3eaa1b840c9a4c895968690':
  svq3: move the frame num variables to the SVQ3Context

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '21b746932241246be846a133abb3c5f91b1cab85'

* commit '21b746932241246be846a133abb3c5f91b1cab85':
  svq3: eliminate H264Context.cur_pic usage

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '939b388383db8d0db5b2ff483e3a197c27b79791'

* commit '939b388383db8d0db5b2ff483e3a197c27b79791':
  svq3: eliminate remaining H264SliceContext usage

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '5a5db90edf71ef4c60db8ad7b0ebaa9a810c2d9e'

* commit '5a5db90edf71ef4c60db8ad7b0ebaa9a810c2d9e':
  svq3: move pict_type to the SVQ3Context

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit '12f13ecb2dcddfa3ee930167395370d3c6fff90c'

* commit '12f13ecb2dcddfa3ee930167395370d3c6fff90c':
  svq3: move mb strides/sizes to the SVQ3Context

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Merge commit 'ad9d3384de08f02967d6eb11196ee8c78e8b2dba'

* commit 'ad9d3384de08f02967d6eb11196ee8c78e8b2dba':
  svq3: move the dequant buffer to SVQ3Context

Merged-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

lavf/mpegts: Add E-AC3 registered stream specifier "EAC3".

Related to ticket #5501.

wmalossless: allow calling madd_int16

This is done by actually handling the "prev_values" in the cascaded LMS data
as if it were int16_t, thus requiring switching at various locations the
computations.

avformat/riff: add M702

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

avcodec/error_resilience: Improve missing slice handling for mpeg2

Fixes: m702_2.avi
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dca: fix sync word search error condition

This didn't actually check if sync word was found and always errored out
with "-err_detect explode" option enabled.

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

avcodec/options_table: fix strict compliance constant flags to match the strict field

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

avformat/options_table: Add missing identifier for very strict compliance

Fixes Ticket5443

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

fate: wma: add lossless 24bits tests

Should evaluate coefficients and raw pcm tiles.

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

avcodec/mjpegdec: Do not try to detect last scan but apply idct after all scans for progressive jpeg

Fixes: IMG-20160418-WA0002.jpg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/mjpeg2jpeg_bsf: Check ff_bsf_get_packet success

This fixes ticket #5487 - mjpeg2jpeg bitstream filter causes
segmentation fault with header-less mjpeg.

Signed-off-by: Jan Sebechlebsky <sebechlebskyjan@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>

avcodec/dca: move channel counter utility into dca.h

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

avcodec/dca: move huffman data into separate object file

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

avcodec/noise_bsf: set correct size of .priv_data_size field

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avcodec/wmalosslessdec: Fix memset sizeof

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

avcodec/smc: Mark pixel pointer as const

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

avformat/wsddec: set bit_rate, fixes duration estimation

Signed-off-by: Paul B Mahol <onemda@gmail.com>

avformat/dsfdec: set bit_rate, fixes duration estimation

Signed-off-by: Paul B Mahol <onemda@gmail.com>

lavf/mpegts: Return small probe score for very short transport streams.

Fixes Debian bug 823098.