Merge remote-tracking branch 'qatar/master'

* qatar/master:
  avconv: make the ASCII flow charts narrower to fit onto TTY

Conflicts:
doc/ffmpeg.texi

See: 8fe107609437909ddd2cf8f5690fe0158a0616f2
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '68e95ab81be1aa3f47ab148dceb8711ef5f4212d'

* commit '68e95ab81be1aa3f47ab148dceb8711ef5f4212d':
  dnxhdenc: return meaningful return codes

Conflicts:
libavcodec/dnxhdenc.c

See: b9bedb0b287d3aaa30540d18fdb87b2cc98ca339
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '9e4e35b4d7c43a908944183a58aa389a23116fd6'

* commit '9e4e35b4d7c43a908944183a58aa389a23116fd6':
  avconv_opt: fix avio_open2() return code check

See: 774239be717150909219ad2c0696bfb6a50cf2cb
Merged-by: Michael Niedermayer <michaelni@gmx.at>

lavd: Add QTKit input device.

Signed-off-by: Thilo Borgmann <thilo.borgmann@mail.de>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avconv: make the ASCII flow charts narrower to fit onto TTY

These charts in man page are currently destroyed.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

avformat/wtvenc: advise user when H264 startcode is not present

Fixes ticket #3473.

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/mpegtsenc: move startcode validity check to ff_check_h264_startcode

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/avio: also set generic URL context options

This should have no effect currently as there are no such options yet.

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

avutil/opt: dont crash on av_opt_set_dict() with NULL

This allows calling the function without the need to check if the
dictionary contains any entries

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

dnxhdenc: return meaningful return codes

avconv_opt: fix avio_open2() return code check

avio_open2() only return < 0 when error.

avformat/wtvenc: pad judiciously when writing mpeg2 extradata

Padding rule described here http://msdn.microsoft.com/en-us/library/windows/desktop/dd390707(v=vs.85).aspx

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/wtvdec: demux mpeg2 extradata

Signed-off-by: Peter Ross <pross@xvid.org>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/replaygain: remove unused variable

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  golomb: Fix the implementation of get_se_golomb_long

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

Merge commit '53c20f17c78d1d8a0fc2505868f201e69ff59cc5'

* commit '53c20f17c78d1d8a0fc2505868f201e69ff59cc5':
  vp8: K&R formatting cosmetics

Conflicts:
libavcodec/vp8.c
libavcodec/vp8.h
libavcodec/vp8data.h
libavcodec/vp8dsp.c

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

golomb: Fix the implementation of get_se_golomb_long

This was only used in hevc muxing code so far.

This makes the return values match what get_se_golomb returns for
the same bitstream reader instances.

The logic for producing a signed golomb code out of an unsigned one
was based on the corresponding code in get_se_golomb, which operated
directly on the bitstream reader buffer - not on the equivalent
return value from get_ue_golomb.

CC: libav-stable@libav.org
Signed-off-by: Martin Storsjö <martin@martin.st>

avcodec/vp9dsp: fix typo in mc doxy.

x86/vp9mc: add vp9 namespace.

Merge commit 'lukaszmluki/master^'

* commit 'lukaszmluki/master^':
  lavd/pulse_audio_enc: implement get_device_list callback
  lavd/pulse_audio_dec: implement get_device_list callback
  lavd/pulse_audio_common: add device detecting code
  lavu/mem: add av_dynarray_add_nofree function
  lavd/pulse_audio_enc: implement write_uncoded_frame callback
  tools/uncoded_frame: fix audio codec generation

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

vp8: K&R formatting cosmetics

Signed-off-by: Diego Biurrun <diego@biurrun.de>

avformat/movenc: fix if vs if else

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  movenc: Add dvd subtitle support

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

Merge commit '85698be461c07be10d873dd34348bcfe9ffc56e0'

* commit '85698be461c07be10d873dd34348bcfe9ffc56e0':
  cmdutils: Mark exit_program as av_noreturn

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

lavd/pulse_audio_enc: implement get_device_list callback

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

lavd/pulse_audio_dec: implement get_device_list callback

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

lavd/pulse_audio_common: add device detecting code

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

lavu/mem: add av_dynarray_add_nofree function

av_dynarray_add_nofree function have similar functionality
as existing av_dynarray_add, but it doesn't deallocate memory
on fails.

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

lavd/pulse_audio_enc: implement write_uncoded_frame callback

Provided implementation doesn't support planar formats yet.

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

tools/uncoded_frame: fix audio codec generation

Codec id is guessed from uninitialized sample format.

Signed-off-by: Lukasz Marek <lukasz.m.luki@gmail.com>

avcodec/vp7: check buffer size

Fixes overread
Fixes Ticket 3501
Found-by: Piotr Bandurski <ami_stuff@o2.pl>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec/xbmdec: redesign parser to handle more cases

The new code is more tolerant on the syntax

Fixes decoding of bm1.xbm

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

avformat/img2_alias_pix: rewrite probe function

Fixes probetest failure

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

avformat/img2dec: try to read PROBE_BUF_MIN instead of just enough for .pix probing

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

avformat/img2dec: Use avformat probing interface to identify format if it has not been otherwise identified

This is used only for distinguishing .pix formats for now.
Which is the only case that has image2 demuxers currently

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat: add image2 brender pix demuxer

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat: add image2 alias pix demuxer

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avformat/img2dec: make image2dec capable to be used from seperate demuxers

Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avfilter/lavfutils: call av_find_stream_info() before reading various information about the stream

No testcase with unmodified source is known

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

cmdutils/filter_codec_opts: do not discard all options for CODEC_ID_NONE

This fixes cases where the codec is detect later than the stream creation
No testcase with unmodified source known

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

movenc: Add dvd subtitle support

Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

fate/aliaspix: Use standard test sample

This also uses samples with odd resolution which should provide better coverage

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  fate: add a bmpparser test

Conflicts:
tests/fate/image.mak

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

Merge commit 'fb5cf145b6bcfa4f83af94398e5560c1132cc410'

* commit 'fb5cf145b6bcfa4f83af94398e5560c1132cc410':
  bmp: add a standalone parser

Conflicts:
Changelog
libavcodec/bmp_parser.c
libavcodec/version.h

See: 011ce89dba5c0c764abc5f39acb6368f580742e3 and others
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '991362fab49b60d34d89b7b7d6dd00fbec3ce022'

* commit '991362fab49b60d34d89b7b7d6dd00fbec3ce022':
  fate: add XBM tests

Conflicts:
tests/fate/image.mak

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

avcodec/xbmdec: support X10 format

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

avcodec/xbmdec: merge ptr increase into dereference

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

Merge commit '678082b409aca711f9cf991df6b0200116489322'

* commit '678082b409aca711f9cf991df6b0200116489322':
  X-Bitmap decoder

Conflicts:
Changelog
libavcodec/Makefile
libavcodec/allcodecs.c
libavcodec/version.h
libavcodec/xbmdec.c

See: a0340dfbd78df7639715056937011d2d414aaeed
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '792e4c21f212979f0e29bcdf107cb6b4f51645a4'

* commit '792e4c21f212979f0e29bcdf107cb6b4f51645a4':
  xbm: use av_frame_free on close

Conflicts:
libavcodec/xbmenc.c

No change as the buggy code is not part of FFmpeg

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

Merge commit 'a7a5e3850ecd94e726ad2272295b9e6c91841cf8'

* commit 'a7a5e3850ecd94e726ad2272295b9e6c91841cf8':
  fate: add PAF audio and video tests

Conflicts:
tests/fate/audio.mak
tests/fate/video.mak

See: 101615c7ff37afe5e6a951838cbbbb92b7485473
See: 409cc4588329633326b32a5245b085c543689471
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit '70daeacd6ef8b354dd7d2d77ad393831a5bbf033'

* commit '70daeacd6ef8b354dd7d2d77ad393831a5bbf033':
  PAF demuxer and decoder

Conflicts:
Changelog
doc/general.texi
libavcodec/avcodec.h
libavcodec/codec_desc.c
libavcodec/paf.c
libavcodec/version.h
libavformat/Makefile
libavformat/allformats.c
libavformat/paf.c
libavformat/version.h

See: 7de4a16508cb9bd2d40855f7114dea14329ef62a, and others
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'd37c96213a2a9e1fd8669122d5405f4ce6a99ed8'

* commit 'd37c96213a2a9e1fd8669122d5405f4ce6a99ed8':
  lavc: restore copy_block{4,16} functions

Conflicts:
libavcodec/copy_block.h

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

cmdutils: Mark exit_program as av_noreturn

dshow: show device name when outputting buffer overflow log message

Signed-off-by: rogerdpack <rogerpack2005@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

fate: add a bmpparser test

bmp: add a standalone parser

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

avcodec/libx264: move where x264opts is applied down so it isnt overridden by avctx & defaults

fixes x264opts opengop=1

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

fate: add XBM tests

X-Bitmap decoder

xbm: use av_frame_free on close

fate: add PAF audio and video tests

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

PAF demuxer and decoder

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>

lavc: restore copy_block{4,16} functions

They were removed in 9e31729d692f1e721b7ed1a3a0f51b68c064d68f.

x86: convert DNxHDenc inline asm to yasm

Signed-off-by: Timothy Gu <timothygu99@gmail.com>
Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

dnxhdenc: make get_pixel_8x4_sym accept ptrdiff_t as stride

Signed-off-by: Timothy Gu <timothygu99@gmail.com>
Reviewed-by: "Ronald S. Bultje" <rsbultje@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

avcodec: Make ff_print_debug_info2() independant of Picture struct

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

avcodec/brenderpix: propagate error codes

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

avcodec/brenderpix: remove unused variable

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  fate: add BRender PIX tests

Conflicts:
tests/fate/image.mak

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

Merge commit 'ae17878fb2ab100264226c84c58f5b95a703312f'

* commit 'ae17878fb2ab100264226c84c58f5b95a703312f':
  BRender PIX image decoder

Conflicts:
doc/general.texi
libavcodec/Makefile
libavcodec/version.h

See: 492a5f835c59ab7c2e347d3579d8bf6148ae0a0, and others
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'c3a0b3eb64be441ca897629e8ecd80d5b51fded7'

* commit 'c3a0b3eb64be441ca897629e8ecd80d5b51fded7':
  arm: build: Maintain decoder objects separate from infrastructure objects

Conflicts:
libavcodec/arm/Makefile

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

fate: add BRender PIX tests

BRender PIX image decoder

Further enhancements by Vittorio Giovara and Paul B Mahol.

Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>

arm: build: Maintain decoder objects separate from infrastructure objects

avcodec/vp7: Fix null pointer dereference in vp7_decode_frame_header()

This simply copies the "interframe without a prior keyframe" check
thats done later into vp7_decode_frame_header()

Found-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

configure: fix VP7 standalone build

Found-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Fix spelling errors in texi files: more informations --> more information allows to --> allows one to

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

Fix texinfo error due to wrong @subsubsection

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

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  fate: add Alias PIX tests

Conflicts:
tests/fate/image.mak

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

Merge commit 'e50f5d3cf9ef9a16982a5cb4d8b1916cd963aa5b'

* commit 'e50f5d3cf9ef9a16982a5cb4d8b1916cd963aa5b':
  Alias PIX image encoder and decoder

Conflicts:
Changelog
libavcodec/allcodecs.c
libavcodec/version.h
libavformat/img2enc.c

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

libavformat/aviobuf: keep track of the original buffer-size and restore it after probe/ensure-seekback

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

avformat/aviobuf: ffio_ensure_seekback: only copy the initialized part of the buffer

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

fate: add Alias PIX tests

Alias PIX image encoder and decoder

Merge remote-tracking branch 'qatar/master'

* qatar/master:
  truehd: add hand-scheduled ARM asm version of ff_mlp_pack_output.

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

Merge commit 'b9eb03416d93a5c4ece27ffef5e6e11c81bec6fa'

* commit 'b9eb03416d93a5c4ece27ffef5e6e11c81bec6fa':
  truehd: break out part of output_data into platform-specific callback.

Conflicts:
libavcodec/mlpdec.c

See: b01a2562ae3ff50198ba189c878023ee4ac34c72
Merged-by: Michael Niedermayer <michaelni@gmx.at>

Merge commit 'fcf5fc444522d24caa9907225802817ae788f511'

* commit 'fcf5fc444522d24caa9907225802817ae788f511':
  truehd: tune VLC decoding for ARM.

Conflicts:
libavcodec/mlpdec.c

See: e555e1bc39ea71d46a07d0bb847fb426c2dd4d80
Merged-by: Michael Niedermayer <michaelni@gmx.at>

truehd: break out part of output_data into platform-specific callback.

Verified with profiling that this doesn't have a measurable effect upon
overall performance.

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

Merge commit '483321fe789566dcb27b6387c00ea16dd86bc587'

* commit '483321fe789566dcb27b6387c00ea16dd86bc587':
  truehd: add hand-scheduled ARM asm version of ff_mlp_rematrix_channel.

See: 89135716fd4c2963e01e0155547c47bf709f1aa3
Merged-by: Michael Niedermayer <michaelni@gmx.at>

truehd: add hand-scheduled ARM asm version of ff_mlp_rematrix_channel.

Profiling results for overall audio decode and the rematrix_channels function
in particular are as follows:

              Before          After
              Mean   StdDev   Mean   StdDev  Confidence  Change
6:2 total     370.8  17.0     348.8  20.1    99.9%       +6.3%
6:2 function  46.4   8.4      45.8   6.6     18.0%       +1.2%  (insignificant)
8:2 total     343.2  19.0     339.1  15.4    54.7%       +1.2%  (insignificant)
8:2 function  38.9   3.9      40.2   6.9     52.4%       -3.2%  (insignificant)
6:6 total     658.4  15.7     604.6  20.8    100.0%      +8.9%
6:6 function  109.0  8.7      59.5   5.4     100.0%      +83.3%
8:8 total     896.2  24.5     766.4  17.6    100.0%      +16.9%
8:8 function  223.4  12.8     93.8   5.0     100.0%      +138.3%

The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.

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

Merge commit '4e5aa080bb8d83cb6de1ffbdd7b37ec34bc6b30b'

* commit '4e5aa080bb8d83cb6de1ffbdd7b37ec34bc6b30b':
  truehd: break out part of rematrix_channels into platform-specific callback.

See: 3f4e73afe92743e96a633aa5b80416e4714e4783
Merged-by: Michael Niedermayer <michaelni@gmx.at>

truehd: break out part of rematrix_channels into platform-specific callback.

Verified with profiling that this doesn't have a measurable effect upon
overall performance.

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

Merge commit '15a29c39d9ef15b0783c04b3228e1c55f6701ee3'

* commit '15a29c39d9ef15b0783c04b3228e1c55f6701ee3':
  truehd: add hand-scheduled ARM asm version of mlp_filter_channel.

Conflicts:
libavcodec/arm/Makefile
libavcodec/arm/mlpdsp_init_arm.c

See: 87b128d5ef6aeb01cc8bba167c929c74cbf2e7c7
Merged-by: Michael Niedermayer <michaelni@gmx.at>

truehd: add hand-scheduled ARM asm version of mlp_filter_channel.

Profiling results for overall audio decode and the mlp_filter_channel(_arm)
function in particular are as follows:

              Before          After
              Mean   StdDev   Mean   StdDev  Confidence  Change
6:2 total     380.4  22.0     370.8  17.0    87.4%       +2.6%  (insignificant)
6:2 function  60.7   7.2      36.6   8.1     100.0%      +65.8%
8:2 total     357.0  17.5     343.2  19.0    97.8%       +4.0%  (insignificant)
8:2 function  60.3   8.8      37.3   3.8     100.0%      +61.8%
6:6 total     717.2  23.2     658.4  15.7    100.0%      +8.9%
6:6 function  140.4  12.9     81.5   9.2     100.0%      +72.4%
8:8 total     981.9  16.2     896.2  24.5    100.0%      +9.6%
8:8 function  193.4  15.0     103.3  11.5    100.0%      +87.2%

Experiments with adding preload instructions to this function yielded no
useful benefit, so these have not been included.

The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.

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

avfilter/vf_pullup: add comment to explain memset(0)

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

truehd: add hand-scheduled ARM asm version of ff_mlp_pack_output.

Profiling results for overall decode and the output_data function in
particular are as follows:

              Before          After
              Mean   StdDev   Mean   StdDev  Confidence  Change
6:2 total     339.6  15.1     329.3  16.0    95.8%       +3.1%  (insignificant)
6:2 function  24.6   6.0      9.9    3.1     100.0%      +148.5%
8:2 total     324.5  15.5     323.6  14.3    15.2%       +0.3%  (insignificant)
8:2 function  20.4   3.9      9.9    3.4     100.0%      +104.7%
6:6 total     572.8  20.6     539.9  24.2    100.0%      +6.1%
6:6 function  54.5   5.6      16.0   3.8     100.0%      +240.9%
8:8 total     741.5  21.2     702.5  18.5    100.0%      +5.6%
8:8 function  63.9   7.6      18.4   4.8     100.0%      +247.3%

The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.

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

truehd: break out part of output_data into platform-specific callback.

Verified with profiling that this doesn't have a measurable effect upon
overall performance.

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

truehd: tune VLC decoding for ARM.

Profiling on a Raspberry Pi revealed the best performance to correspond
with VLC_BITS = 5. Results for overall audio decode and the get_vlc2 function
in particular are as follows:

              Before          After
              Mean   StdDev   Mean   StdDev  Confidence  Change
6:2 total     348.8  20.1     339.6  15.1    88.8%       +2.7%  (insignificant)
6:2 function  38.1   8.1      26.4   4.1     100.0%      +44.5%
8:2 total     339.1  15.4     324.5  15.5    99.4%       +4.5%
8:2 function  33.8   7.0      27.3   5.6     99.7%       +23.6%
6:6 total     604.6  20.8     572.8  20.6    100.0%      +5.6%
6:6 function  95.8   8.4      68.9   8.2     100.0%      +39.1%
8:8 total     766.4  17.6     741.5  21.2    100.0%      +3.4%
8:8 function  106.0  11.4     86.1   9.9     100.0%      +23.1%

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

truehd: add hand-scheduled ARM asm version of ff_mlp_rematrix_channel.

Profiling results for overall audio decode and the rematrix_channels function
in particular are as follows:

              Before          After
              Mean   StdDev   Mean   StdDev  Confidence  Change
6:2 total     370.8  17.0     348.8  20.1    99.9%       +6.3%
6:2 function  46.4   8.4      45.8   6.6     18.0%       +1.2%  (insignificant)
8:2 total     343.2  19.0     339.1  15.4    54.7%       +1.2%  (insignificant)
8:2 function  38.9   3.9      40.2   6.9     52.4%       -3.2%  (insignificant)
6:6 total     658.4  15.7     604.6  20.8    100.0%      +8.9%
6:6 function  109.0  8.7      59.5   5.4     100.0%      +83.3%
8:8 total     896.2  24.5     766.4  17.6    100.0%      +16.9%
8:8 function  223.4  12.8     93.8   5.0     100.0%      +138.3%

The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.

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

truehd: break out part of rematrix_channels into platform-specific callback.

Verified with profiling that this doesn't have a measurable effect upon
overall performance.

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