This fixes a case where multichannel bitrate isn't accurately
targetted by psy model alone, never achieving the target bitrate.
Signed-off-by: Martin Storsjö <martin@martin.st>

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

The C code treats the data as arrays of uint16_t so strides must not
be in bytes but in pixels.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

Always use the special filter for the first and last 3 columns (only).

Changes made in 64ed3976 slowed the filter to just under 3/4 of what it
was.  This commit restores the speed while maintaining identical output.

For reference, on my Athlon64:
1733222 decicycles in old
2358563 decicycles in new
1727558 decicycles in this
Signed-off-by: Anton Khirnov <anton@khirnov.net>

This fixes https://bugzilla.libav.org/show_bug.cgi?id=514.
Signed-off-by: Martin Storsjö <martin@martin.st>

If this is explicitly disabled for win32/mingw, it should also
be disabled for cygwin, for consistency and for the same reasons
as for win32/mingw.
Signed-off-by: Martin Storsjö <martin@martin.st>

The header is not (anymore) MMX-specific.

Also remove some pointless comments.

The assembler may insert nopl instructions for cpunop, which are not
(universally) supported on i586 CPUs.

The goal is to make the capapility slightly more general and have it
cover the availability of the nopl instruction in addition to cmov.

These platforms do not have any notion of PIC. On some compilers,
enabling pic produces a number of warnings.

This avoids trying to produce PIC loads in the ARM assembly - there
are no relocation types in PE/COFF that correspond to
BFD_RELOC_32_PCREL (R_ARM_REL32 in ELF).

As a side-effect, this avoids enabling PIC on mingw64, getting rid
of the warnings about PIC not having any effect on that platform.
Signed-off-by: Martin Storsjö <martin@martin.st>

This allows it to be overridden, either by the user on the command
line, or by other sections of the configure script.
Signed-off-by: Martin Storsjö <martin@martin.st>

Cells starting at a position aligned to 8 pixels but wider than
4 blocks are copied with 3 blocks per loop. This creates problems on the
next loop iterations since the routine copying 2 blocks requires the
same alignment on some architectures like ARM NEON.

From 253 to 51 cycles on Arrandale and Win64.
44 cycles on SandyBridge.
Signed-off-by: Anton Khirnov <anton@khirnov.net>

delta is in samples, not bytes. Also the sample format is not guaranteed
to be planar.

CC:libav-stable@libav.org

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

This makes linking succeed for tools that include config.h but
don't link to libavutil.
Signed-off-by: Martin Storsjö <martin@martin.st>

The indirection was used to abstract from the different -regression.sh
scripts. Now that only one remains, this indirection is pointless.

Prevent spurios EIO on EOF.

CC:libav-stable@libav.org

Notwithstanding the return value the frame should be freed if
is not going to be used.
Reported-by: Matthew Wolenetz <wolenetz@chromium.org>

Fix the mispresented EINVAL for EOF on partial files, among the other.

0.1 seconds at 25 fps is exactly 2.5 frames, which may be rounded either
to 2 or 3 on different platforms.

The scale filter depends on sws, resample on lavr, and movie on lavf and
lavc.

Fixes Bug 502.

They are now used for -ss/-t

It is incompatible with refcounted frames and since it's been deprecated
for a long time now, fixing it is not worth the effort.

They are not guaranteed to be aligned.
Fixes Bug 503.

CC:libav-stable@libav.org