Use bitfields in FormatEntry array to avoid wasting an int for each flag.

Every other pixel had the alpha channel taken from the wrong
row.

This fixes bug 504.

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

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

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

This reverts parts of d6d5ef55, that didn't work right. (The
tests that were added failed on big endian, and the output looked
garbled on little endian as well.)

This is due to the fact that the intermediate scaling values (from
e.g. hScale8To19_c or hScale16To19_c) are stored as int32_t and
thus requires a separate output function, while yuv2gbrp_full_X_c
only interprets it as int16_t.

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

planar rgb formats do not use the table

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

This code path is not implemented and makes not much sense to implement
either.

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

The function doesnt support >8bit currently

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

The new name is more descriptive and will allow defining a separate
public prefix for externally visible library symbols.

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

sws_getCachedContext() and sws_getContext() expect sws_alloc_context()
to return NULL when out of memory, as follows.

    if (!(context = sws_alloc_context()))
        return NULL;

This patch fixes sws_alloc_context() to return NULL in that case.

Signed-off-by: Xi Wang <xi.wang@gmail.com>
Signed-off-by: Derek Buitenhuis <derek.buitenhuis@gmail.com>

SWS_CPU_CAPS are deprecated and slated to removed with libswscale major
version 3. No need to provide a SWS_CPU_CAPS_MMX2 as backward
compatibility define under the same explicit condition.

Otherwise during scaling it will try to interpret input in the wrong way and
that leads to the test results disagreeing on different platforms and with
different optimizations.

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

This is more consistent with the way we handle C #includes and
it simplifies the build system.

This is necessary to allow refactoring some x86util macros with cpuflags.

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

With the input of Kostya and Ronald.

Some compilers (e.g. old gcc) have trouble with these.

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

Some systems, e.g. Minix, have sys/mman.h defining MAP_ANONYMOUS without
providing (working) mmap and friends.  The mmx filter generation code
checks only for MAP_ANONYMOUS, not for availability of mmap itself which
leads to build errors on aforementioned systems.

This changes the conditional compilation to use mmap only if all the
required functions are available.

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

Anonymous structs cannot be forward declared and have no benefit.

This gets rid of the variable-length scratch buffer by filtering 16
pixels at a time and writing directly to the destination.  The extra
loads this requires to load the source values are compensated by not
doing a round-trip to memory before shifting.

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