swr: move dst_size == 0 handling outside DSP function.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

doc/APIchanges: lengthening a hash to make it non ambigous

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

APIchanges: fillin some missing data

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '39975acc699c83af0a87a7318c0f41e189142938'

* commit '39975acc699c83af0a87a7318c0f41e189142938':
  rtpenc_jpeg: check for color_range too

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'f134b5ec53b4cb51cb69bf0c64de87687ea72b12'

* commit 'f134b5ec53b4cb51cb69bf0c64de87687ea72b12':
  apichanges: fill in changes for lavu 51.19 and 51.20

Conflicts:
doc/APIchanges

Merged-by: Michael Niedermayer <michaelni@gmx.at>

rtpenc_jpeg: check for color_range too

apichanges: fill in changes for lavu 51.19 and 51.20

avcodec/mpegvideo: dont overwrite emu_edge buffer

Fixes vsynth3 fate failures on mips

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/mpegvideo_enc: reduce space between blocks in emu_edge in encode_mb_internal

Revert "x86/videodsp: add emulated_edge_mc_mmxext"

The commit causes minor out of array reads and was mainly intended for
future optimizations which turned out not to be meassurably faster.
Itself it was just 1 cpu cycle faster

Approved-by: jamrial
This reverts commit 057d2704e78b52fec357b4fc646c9de37a085413.

avcodec/mpeg12dec: Limit progressive_seq reinitilaization to where the resolution changes

Fixes Ticket3743

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/vp8: Fix multiple ff_thread_finish_setup() calls

Fixes Ticket3725

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5dd8c08fd5e4c04d7a08d8934f0098a8a4a35c28'

* commit '5dd8c08fd5e4c04d7a08d8934f0098a8a4a35c28':
  mpeg: Change ff_convert_matrix() to take an MpegEncContext parameter

Conflicts:
libavcodec/mpegvideo.h

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'e63b818dbe6060b606ae10eb184859f8b97d2353'

* commit 'e63b818dbe6060b606ae10eb184859f8b97d2353':
  dv: Properly split decoder and encoder initialization

Conflicts:
libavcodec/dv.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

ffmpeg: fix deadlock regression in threading error handing

Commit fc9c857c introduced deadlock regression when processing too many inputs:
  ffmpeg $(seq -f " -f lavfi -i aevalsrc=0:d=%.0f" 70) -vf concat=n=70:v=0:a=1 -f null -
Happens for different number of inputs, depending on available memory size,
overcommit settings, ulimits, etc. Easily noticeable for 32-bit builds,
that exhaust address space allocating 8-10 MB stack for each thread.
Earlier ffmpeg versions exited with unhelpful "Conversion failed!" message.

This patch fixes both problems: it frees the queue to prevent deadlock
and adds a meaningful error message if pthread_create() fails.

Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/dvbsubdec: fix ignoring blank frame by updating got_output

fix ticket 3737

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

mpeg: Change ff_convert_matrix() to take an MpegEncContext parameter

This will come in handy during dsputil splitting.

dv: Properly split decoder and encoder initialization

configure: add missing audiodsp dependency for ra144

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: fix g729 decoder dependencies

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: remove duplicate ape_decoder_select line

Signed-off-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/snow: factor ff_snow_get_buffer() out

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/mxfenc: rename very recently added option to set/force channelcount in MXF D-10

Approved-by: Tomas Härdin <tomas.hardin@codemill.se>
Approved-by: tim nicholson <nichot20@yahoo.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'd2869aea0494d3a20d53d5034cd41dbb488eb133'

* commit 'd2869aea0494d3a20d53d5034cd41dbb488eb133':
  dsputil: Move MMX/SSE2-optimized IDCT bits to the x86 subdirectory

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '24f45c16224d4c5d482e928676714766ffdda4fc'

* commit '24f45c16224d4c5d482e928676714766ffdda4fc':
  fate: Add dependencies for dct/fft/mdct/rdft tests

Conflicts:
libavcodec/fft-test.c
tests/fate/fft.mak

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5ec6d152e26c570c0a16ec72c1f354db95708179'

* commit '5ec6d152e26c570c0a16ec72c1f354db95708179':
  indeo4: B-frames decoding

Conflicts:
libavcodec/ivi_common.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

dsputil: Move MMX/SSE2-optimized IDCT bits to the x86 subdirectory

Merge commit '422e14f721c22cf9c19a8e7aae051ba9d559f6b6'

* commit '422e14f721c22cf9c19a8e7aae051ba9d559f6b6':
  indeo2: rename stride to pitch for consistency with other Indeo decoders

Merged-by: Michael Niedermayer <michaelni@gmx.at>

fate: Add dependencies for dct/fft/mdct/rdft tests

Merge commit 'a5864e9a253143437471dba61a533e7fa9acec1c'

* commit 'a5864e9a253143437471dba61a533e7fa9acec1c':
  output example: free the muxing format context properly

See: 9b211c43dc5f2e618f204c4a7fd184eb2ea51f02
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '182d3f8221c23716ea4eafa460bdb94968f71f26'

* commit '182d3f8221c23716ea4eafa460bdb94968f71f26':
  output example: rewrite encoding logic

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'a7fcd4122b19b0f934020f4e261d0c44c4c32e11'

* commit 'a7fcd4122b19b0f934020f4e261d0c44c4c32e11':
  output example: store the scaling context in the stream context

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '6a927d7aaf5625e83a674072913b9e292a303fd1'

* commit '6a927d7aaf5625e83a674072913b9e292a303fd1':
  output example: use a macro instead of a static variable

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '63fd0d866c8300a8f251a15b1535e9ce40a407fb'

* commit '63fd0d866c8300a8f251a15b1535e9ce40a407fb':
  output example: allocate the audio frame only once

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'edd5f957646dcbf1bb55718bc7bf1e5481c25bcb'

* commit 'edd5f957646dcbf1bb55718bc7bf1e5481c25bcb':
  output example: use OutputStream for audio streams as well

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5e7b125b6ae36893dfd9cb5661c99b67363cbb38'

* commit '5e7b125b6ae36893dfd9cb5661c99b67363cbb38':
  output example: use the new AVFrame API to allocate audio frames

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'ac85f631c9a9cc59aaca1c8dd6894fb1f701c594'

* commit 'ac85f631c9a9cc59aaca1c8dd6894fb1f701c594':
  output example: set output channel layout

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '294daf71a7a1303b5ddd3cbefebed3b732d610f3'

* commit '294daf71a7a1303b5ddd3cbefebed3b732d610f3':
  output example: use a struct to bundle the video stream variables together

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '38d044429dabcd28928693240e955c6d4430cca9'

* commit '38d044429dabcd28928693240e955c6d4430cca9':
  output example: use the new AVFrame API for allocating the video frame

Conflicts:
doc/examples/muxing.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '3f4edf012593c73941caa0ef9b292da00225c3df'

* commit '3f4edf012593c73941caa0ef9b292da00225c3df':
  dump_stream: print the timebase as is, do not reduce it

See: 75511c293add07db1cca58dcd8b08c33fc2f1075
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '1b04eb20f7e3f0a71f73ba91efcc3d60a435e443'

* commit '1b04eb20f7e3f0a71f73ba91efcc3d60a435e443':
  lavc: do not allocate edges in the default get_buffer2()

Conflicts:
libavcodec/utils.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/snow: ensure the buffers have allocated edges

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'a0e1c3517a656dd32293f054a339e0ac73328138'

* commit 'a0e1c3517a656dd32293f054a339e0ac73328138':
  hevc: remove unused array min_cb_addr_zs

Conflicts:
libavcodec/hevc_ps.c

See: f7f1f4c7ce9ce689823e13a53b694eb14cbbf6e7
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '9b60d9197970658e91daf4b586397f450de9af69'

* commit '9b60d9197970658e91daf4b586397f450de9af69':
  hevc: Allow out of bound values for num_reorder_pics

Conflicts:
libavcodec/hevc_ps.c

See: bc21260e643c2249e38f7485bceada1d4bbb2b9a
See: ab296c7a9fe590860dc95ba97e9cbb9dde798f20
Merged-by: Michael Niedermayer <michaelni@gmx.at>

indeo4: B-frames decoding

Signed-off-by: Kostya Shishkov <kostya.shishkov@gmail.com>

indeo2: rename stride to pitch for consistency with other Indeo decoders

x86/videodsp: add emulated_edge_mc_mmxext

This also changes hfix8_mmx and above to use mmx regs instead of
gprs, and makes emulated_edge_mc_sse and emulated_edge_mc_sse2 use
mmxext hfix and hvar functions instead of mmx where possible.

This is mostly in preparation for an ssse3 version.

Signed-off-by: James Almer <jamrial@gmail.com>
code is about 1 cpu cycle faster approximately

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

output example: free the muxing format context properly

output example: rewrite encoding logic

Properly generate pts for the frames sent to the encoder, avoid
using private and deprecated AVStream.pts.

output example: store the scaling context in the stream context

output example: use a macro instead of a static variable

output example: allocate the audio frame only once

output example: use OutputStream for audio streams as well

output example: use the new AVFrame API to allocate audio frames

output example: set output channel layout

output example: use a struct to bundle the video stream variables together

output example: use the new AVFrame API for allocating the video frame

dump_stream: print the timebase as is, do not reduce it

It makes more sense to print the timebase exactly as it is set. Also,
this avoids a divide by zero when av_dump_format() is called on a format
context before writing the header.

lavc: do not allocate edges in the default get_buffer2()

hevc: remove unused array min_cb_addr_zs

Signed-off-by: Anton Khirnov <anton@khirnov.net>

avcodec/cavs: fix B_SUB_DIRECT col-MB is Intra MB case.

in this case current MB size is forced to 16x16 (AVS standard section 9.9.1)

Signed-off-by: Yao Wang <jiayaowang@gmail.com>
Fixes Ticket 1901

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

hevc: Allow out of bound values for num_reorder_pics

This fixes decoding for a sample that cannot be shared

Signed-off-by: Anton Khirnov <anton@khirnov.net>

Merge commit '593d2326ef985cdffe413df629419938f7b07c4c'

* commit '593d2326ef985cdffe413df629419938f7b07c4c':
  dv: Replace a magic number by sizeof()

Conflicts:
libavcodec/dv.c

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '5ab03e41e553452118113d0c224fa32b325e45e5'

* commit '5ab03e41e553452118113d0c224fa32b325e45e5':
  x86: h264dsp: Fix link failure with optimizations disabled

Merged-by: Michael Niedermayer <michaelni@gmx.at>

avformat/mxfenc: set/force channelcount in MXF D-10

There are interoperability issues with D-10 related to the channelcount property in the generic sound essence descriptor.

On one side, SMPTE 386M requires channel count to be 4 or 8, other values being prohibited.
The most widespread value is 8, which seems straightforward as it is the actual size of the allocated structure/disk space.
At the end, it appears that some vendors or workflows do require this descriptor to be 8, and otherwise just "fail".

On the other side, at least AVID and ffmpeg do write/set the channel count to the exact number of channels really "used",
usually 2 or 4, or any other value. And on the decoding side, ffmpeg (for example) make use of the channel count for probing
and only expose this limited number of audio streams
(which make sense but has strong impact on ffmpeg command line usage, output, and downstream workflow).

At the end, I find it pretty usefull to simply give ffmpeg the ability to force/set the channel count to any value the user wants.
(there are turnaround using complex filters, pans, amerge etc., but it is quite boring and requires the command line to be adapted to the input file properties)

Reviewed-by: Matthieu Bouron <matthieu.bouron@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

dv: Replace a magic number by sizeof()

x86: h264dsp: Fix link failure with optimizations disabled

With optimzations disabled compilers have trouble doing dead code
elimination on 'if (foo && 0)' expressions, while 'if (0 && foo)'
still works, so use the latter to avoid problems.

Bug-Id: 707

avfilter/x86/vf_pullup: fix old typo

This makes C and MMX match, no change to fate as the differences where
apparently not sufficient to show up in fate

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/vf_pullup: use ptrdiff_t as stride argument for dsp functions

This should avoid issues on x86_64

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/x86/hevc_idct: fix function name in comment

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/x86/hevc_idct: fix number of sse registers

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/x86/hevc: add avx2 dc idct

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

doc/build_system: document examples and config targets

Signed-off-by: Lukasz Marek <lukasz.m.luki2@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge remote-tracking branch 'cus/stable'

* cus/stable:
  ffplay: decrease audio_diff_threshold
  ffplay: decrease max audio callbacks per second
  ffplay: calculate SDL audio buffer size based on sample rate
  ffplay: pass simple integers to calculate_display_rect and set_default_window_size
  ffplay: eliminate pictq_prev_picture

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '2deb614272e6faad8802c5341971d08c7272f74d'

* commit '2deb614272e6faad8802c5341971d08c7272f74d':
  mjpegdec: Properly set the context colorspace info

See: c11043aca7365add6c4c80b7cff80d22dde0eb94
Merged-by: Michael Niedermayer <michaelni@gmx.at>

ffplay: decrease audio_diff_threshold

Since audio clock calculations are more accurate now, it is safe to decrease
the sync treshold to compensate the larger buffers caused by less frequent
audio callbacks.

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

ffplay: decrease max audio callbacks per second

Too many audio callbacks per second can cause buffer underruns especially under
load. As now we take into accound the elapsed time after an audio callback when
determining current audio clock, it is not that important to use small buffer
sizes and frequent audio callbacks, so lets remove the comment.

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

ffplay: calculate SDL audio buffer size based on sample rate

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

ffplay: pass simple integers to calculate_display_rect and set_default_window_size

No change in functionality.

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

ffplay: eliminate pictq_prev_picture

Instead of directly rolling back the frame queue, keep the last displayed
picture in the queue and use a boolean variable to keep track if it is
displayed or not. This makes the code cleaner because it removes the
complicated logic in pictq_prev_picture.

There should be no change in functionality.

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

avfilter/vf_deshake: fix macro ()

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/vf_colormatrix: fix macro ()

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

mjpegdec: Properly set the context colorspace info

The JPEG spec requires it to be this.

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

mjpegdec: Properly set the context colorspace info

The JPEG spec requires it to be this.

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

avformat/nutdec: improve probe speed by 30%

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/wavenc: simplify malloc failure checking

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/wavenc: more specific error return for "Writing 16 bit peak for 8 bit audio"

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/wavenc: do not hardcode array size in memset and other functions

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/wavenc: use the bitexact flag from avformat instead of the one from avcodec

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

hevc/pps: optimized size of min_tb_addr_zs
reduce computation too
(cherry picked from commit 39c4d45c7788081c45c7fae51b7c5d0bcbaece9d)

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/hevc_ps: remove min_cb_addr_zs

(cherry picked from commit ff7926d5092f9d4158108963e977e8c992322ba4)

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Peak Envelope Chunk encoder: Indent

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

EBU Tech 3285 - Supplement 3 - Peak Envelope Chunk encoder

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/vf_hqx: optimize table init

5389024880 -> 1389386610 dezicycles

This surely can be optimized more, i just didnt want to cause a slowdown
when trying to make the fate test bitexact.
Further optimization left to ubitux

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/vf_hqx: avoid floats

this likely fixes fate failures

2748170880 -> 5389024880 dezicycles

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/x86/h264_qpel_10bit: locally define pb_0

somehow old llvm-gcc manages to ignore the alignment from ff_pb_0 causing a crash on freebsd

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/h264: do not leave ret random on minor failures, causing major failure

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '772d150a6e82542c06b0c251e73dd299d98d1027'

* commit '772d150a6e82542c06b0c251e73dd299d98d1027':
  h264: error out from decode_nal_units() when AV_EF_EXPLODE is set

Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '10306e9c5fcc28bd9310a9b38f21540e9e1433e9'

* commit '10306e9c5fcc28bd9310a9b38f21540e9e1433e9':
  jpeg2000: fix dereferencing invalid pointers during cleanup

Conflicts:
libavcodec/jpeg2000.c

See: 09927f3eaa93e31dd90ab8ee66dc1ad867b3365f
See: 912ce9dd2080c5837285a471d750fa311e09b555
See: 9e477a37703318cb86d8ed1d426929235aa02b67
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'ab72eda15e98197cf148abc08574206cfde0d9b0'

* commit 'ab72eda15e98197cf148abc08574206cfde0d9b0':
  pixfmt: mark the reserved values

Conflicts:
libavutil/version.h

Merged-by: Michael Niedermayer <michaelni@gmx.at>