This is similar to int32_to_float_fmul_scalar, but
loads a new scalar multiplier every 8 input samples.
This enables the use of much larger input arrays, which
is important for pipelining on some CPUs (such as
ARMv6).
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 1175.0 4.4 366.2 18.3 +220.8%
Overall 19285.5 292.0 18420.5 489.1 +4.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 9295.0 114.9 4853.2 83.5 +91.5%
Overall 23699.8 397.6 19285.5 292.0 +22.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
Prior to this it was possible that format reduction was ended
before it fully propagated leading to failure later in picking
formats.
No testcase with unmodified source exists, the case was reproduced
with less aggressive list merging though.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
This causes a race condition with VLC. Its plausible that other
applications also would have races with it and its just fixing a memleak when
the user application forgets to free the codec. It causes more
problems than it solves in its current form, thus the revert.
Better solutions are welcome
This reverts commit 0f229f9b91.
This matches the matroska defintion of stereo_mode, with
no metadata written if no info exist in sei
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
libx264 has a few data exports which require X264_API_IMPORTS
to be defined if we link to libx264 dynamically on Windows.
In a similar fashion to how we handle our compat snprintf
implementation, if we define it all the time, the compiler
will first try and link to __imp_x264_symbol_name, and failing
that, as in the case of a static libx264, will attempt to link
to the non-prefixed symbol, which has already been pulled in by
other x264 functions' object files.
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>
* qatar/master:
dsicinav: Clip the source size to the expected maximum
Clipping the compressed size based on the uncompressed size is not correct
thus this commit is not merged, and the merge is for git metadata only
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* commit '36fb0d02a1faa11eaee51de01fb4061ad6092af9':
rtsp: Support multicast source filters (RFC 4570)
rtpproto: Check the source IP if one single source has been specified
rtpproto: Support IGMPv3 source specific multicast inclusion
Conflicts:
libavformat/rtpproto.c
libavformat/rtsp.c
libavformat/rtsp.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Theoretically using start_time should also work if seeking is available and we
could determine that the next packet after a flush packet is the first packet
of a stream, but I could not think of an easy and clean way to do that, that is
why I sticked to the no seeking available condition for now.
Fixes ticket #2647.
Signed-off-by: Marton Balint <cus@passwd.hu>
Previously we estimated the audio packet pts instead of the frame pts,
therefore it only worked within a single packet (containing multiple frames).
The new method works accross seperate audio packets as well and also handles
better the case if a decoder buffers several packets before outputting a
decoded frame.
Signed-off-by: Marton Balint <cus@passwd.hu>
Also use negative stream_index for signaling obsolete audio packets. Using the
size alone is not enough, because size is 0 for null packets as well.
Signed-off-by: Marton Balint <cus@passwd.hu>
This also fixes the case where negative chapter ids where input
And fixes the case where remuxing from mkv changed chapter ids
Found-by: Luca Barbato
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
