Uses 'buf' directly instead of a separate iterator variable 'in'.

Uses the native sample format for the codec instead of left-shifting all
samples by 8.

Make the code easier to read.

This fixes bind(8080): Address family not supported by protocol.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

On OS X, av_malloc(0) returns pointers that cause crashes when
freed.
Signed-off-by: Martin Storsjö <martin@martin.st>

Copy the Inter 8x8 scaling list as second 8x8 matrix into DXVA2's
quantization matrix data structure instead of a potentially unset
Intra chroma scaling matrix.

Fix dxva2 decoding for some H264 samples.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>

Signed-off-by: Mans Rullgard <mans@mansr.com>

There is no valid reason the user should ever send such packets in the
first place, but the documentation for CODEC_CAP_DELAY states that the
codec is guaranteed not to get a NULL packet unless that capability is
set. That isn't true without preventing this case.

The workaround needs to be enabled per PCI ID which cannot be detected inside
libavcodec. So add a flag to manually enable the alternate behavior.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>

Signed-off-by: Janne Grunau <janne-libav@jannau.net>

Also use the post-header data size to control termination of the main
decoding loop.

Also use the post-header data size to control termination of the main
decoding loop.

Also properly clip the right channel step_index.

Speeds up the ADPCM IMA WAV decoder by 15-20% overall.

Add Multimedia Wiki link.
Mark dead links with [dead]. Some can still be accessed through archive.org.
Update URLs for pages which have moved.
Replace duplicated links in adpcmenc.c with a note to see the ADPCM decoder
reference documents.

This is easier to understand. It also avoids returning existing samples mixed
with new samples when the packet is too small.

show_bits() is undefined when the number of bits is less than or equal to
zero.

Writing zeros to the high entries in the array need only be
done once as the cutoff position is constant throughout the
loop.
Signed-off-by: Mans Rullgard <mans@mansr.com>

The previous default value 0x0 was not good, since it is not even
valid.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

In apply_unsharp(), when y is >= height, prevent out-of-buffer reading
from src, read from the last buffer line in src2 instead.

The check was implemented in the original unsharp libmpcodecs code and
lost in the port.

This also fixes output discrepancy between the two filters.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

Signed-off-by: Mans Rullgard <mans@mansr.com>

Signed-off-by: Mans Rullgard <mans@mansr.com>

Signed-off-by: Mans Rullgard <mans@mansr.com>

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

Also drop obfuscating CHROMA_WIDTH/HEIGHT macros.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

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

More consistent with the original libmpcodecs code, and the name
"unsharpen" was confusing.
Signed-off-by: Anton Khirnov <anton@khirnov.net>