This toggles on Link-Time Optimization for arm and x86-64.
Compile tested on arm/bcm47xx, arm/mvebu, arm/imx6 and x86-64/generic
(Also compile tested on mips/ar71xx where it failed so this change is currently for specific arches.)
Size savings for libffmpeg were 2-3% when measured with libffmpeg-full and libffmpeg-mini.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
This change allows ffmpeg's configure script to determine the necessary parsers based on
selected decoders & encoders. For libffmpeg-audio-dec, this changes from our selecting:
aac, ac3, flac, mpegaudio, opus and aac_latm
to ffmpeg's configure script selecting:
aac_latm, ac3, flac, mpegaudio
for minor space savings and a simpler Makefile to manage on our end.
In libffmpeg-custom, don't select any parsers when choosing decoders/encoders. I'm undecided
on removing the parser selection in its entirety here.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
Removes the x86 optimization disables from the SOFT_FLOAT portion. There are no i386 profiles that do
not use hard float, so these options accomplish nothing.
Target-os=linux rules out dxva2 and vda, so disable statements are unnecessary.
For x86-64 users of libffmpeg-full, optimize around performance instead of size.
Remove patented flag requirement from mp2 and mp3 audio files in libffmpeg-custom per expiration.
Tidies up some of the configure steps where only one option is being changed or cosmetic consistency.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
Some projects, like freeswitch-stable, require avresample in
libffmpeg-full.
Signed-off-by: Sebastian Kemper <sebastian_ml@gmx.net>
Signed-off-by: Ted Hess <thess@kitschensync.net>
Also makes the following changes:
Updates project URL and moves libpthread to a common dependency;
Makes libopus support in libffmpeg-full contingent on selecting the
libopus package, like other external libraries;
Adds note regarding disabling altivec in ppc hard-float;
Reorganizes SIMD sections to be together;
Drops the need to specify decoder=libopus from -full, done by default;
Adds encoder=libopus to custom.
Expand options for libffmpeg-custom:
Build licensing: GPL, GPLv3, LGPLv3
Building for performance or size
Drops version 3 of GPL for -audio-dec, -mini, and -full. Version 3 is
at the discretion of the packager, per:
http://ffmpeg.org/doxygen/trunk/md_LICENSE.html
Add enable-nonfree toggle support to libffmpeg-custom build licensing
to allow inclusion of libfdk-aac.
Use ffmpeg's names for its external libraries in libffmpeg-custom config
Alphabetize same
Correct @BUILD_PATENTED flags for libffmpeg variants
Split libffmpeg-full into hard/soft float:
Hard float configures itself for lame / libx264 using GPL license
Soft float for libshine and LGPL license
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
Signed-off-by: Ted Hess <thess@kitschensync.net>
Update ffmpeg to 2.8.8.
Adds HEVC, Opus, VC-1 and Vorbis to custom and minidlna profiles. This should
more closely match formats media being streamed on the local network via
DLNA/UPNP.
Removes raw aac, ac3 and h264 from minidlna profile. Most players expect
this content to come within a container (mkv, mov, mp4, mpegts, etc...),
so this was unnecessary for the profile's goals.
Remove libopus from being selected by the -custom's audio-dec profile, and as
a depend on -audio-dec. -audio-dec uses ffmpeg's internal Opus decoder, not
the external library.
Don't select specific parsers in libffmpeg-mini profile. The demuxers separate
streams from containers. Parsers break up the streams. These are not needed
for DLNA/UPNP. If a parser is needed, the ffmpeg configure script should grab
it.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
Also no longer disable run-time CPU detection for fpu targets except
arm which is built for each variant (neon, vfp).
This vastly improves performance on x86 which now supports building ASM
code with yasm and makes use of a broad spectrum of extensions
depending on their availability at run-time.
In future, such differentiation should also be introduced for MIPS
to at least conditionally enable the output of MIPS32r2 instructions as
well as DSPr1 and DSPr2 if the target supports that dispite them having
CONFIG_SOFT_FLOAT=y set.
Signed-off-by: Daniel Golle <daniel@makrotopia.org>
Custom builds need to select x264 and/or lame-lib explictly. Automatic
CONFIG_ALL builds of ffmpeg-full include x264 and lame.
Signed-off-by: Ted Hess <thess@kitschensync.net>
Add support for libmp3lame & libx264 to 'full' and optional in 'custom'.
Enable some optimizations according to ARCH and CPU type.
Cleaup options and dependencies.
Signed-off-by: Adrian Panella <ianchi74@outlook.com>
Signed-off-by: Ted Hess <thess@kitschensync.net>
For libffmpeg-audio-dec:
Remove @DEVEL flag from libffmpeg-audio-dec. Reasoning for this is
that choosing this package does not rquire any further action by the
builder in what codecs or formats should be built in.
libffmpeg-custom, the other libffmpeg to require @DEVEL, does need
further interaction.
Add @BUILD_PATENTED as at least one of the audio codecs is likely
patent protected.
Add ffmpeg's native Opus decoder to libffmpeg-audio-dec
Remove libopus dependency to libffmpeg-audio-dec as libopus provides
decode and encode support for Opus audio. Since FFmpeg can decode
Opus audio on its own, and libffmpeg-audio-dec is intended for audio
decoding, libopus is no longer required for Opus audio support.
For libffmpeg-custom:
Add in options for ffmpeg's integrated Opus decoder
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
Some CPUs, such as the au1500 are MIPS32 but not MIPS32r2 and we have to
explicitly disable mips32r2 in the ffmpeg configure script.
Signed-off-by: Bruno Randolf <br1@einfach.org>
Minidlna has no need to mux streams, or be able to read network streams at this point in time, so remove this support from
libffmpeg.
Synchronize the minidlna profile between libffmpeg-ucstom and libffmpeg-mini.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>
As libffmpeg-mini is intended for DLNA purposes, it needs only
to identify media, not interact with input or output devices.
Signed-off-by: Ian Leonard <antonlacon@gmail.com>