When turned on, H264/CAVLC gets ~15% (CVPCMNL1_SVA_C.264) slower for
ultra-high-bitrate files, or ~2.5% (CVFI1_SVA_C.264) for lower-bitrate
files. Other codecs are affected to a lesser extent because they are
less optimized; e.g., VC-1 slows down by less than 1% (all on x86).
The patch generated 3 extra instructions (cmp, cmovae and mov) per
call to get_bits().
The performance penalty on ARM is within the error margin for most
files, up to 4% in extreme cases such as CVPCMNL1_SVA_C.264.
Based on work (for GCI) by Aneesh Dogra <lionaneesh@gmail.com>, and
inspired by patch in Chromium by Chris Evans <cevans@chromium.org>.
* qatar/master:
get_bits: remove A32 variant
avconv: support stream specifiers in -metadata and -map_metadata
wavpack: Fix 32-bit clipping
wavpack: Clip samples after shifting
h264: don't drop B-frames after next keyframe on POC reset.
get_bits: remove useless pointer casts
configure: refactor lists of tests and components into variables
rv40: NEON optimised weak loop filter
mpegts: replace some magic numbers with the existing define
swscale: add unscaled packed 16 bit per component endianess conversion
Conflicts:
libavcodec/get_bits.h
libavcodec/h264.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
The A32 bitstream reader variant is only used on ARMv5 and for
Prores due to the larger bit cache this decoder requires.
In benchmarks on ARMv5 (Marvell Sheeva) with gcc 4.6, the only
statistically significant difference between ALT and A32 is
a 4% advantage for ALT in FLAC decoding. There is thus no (longer)
any reason to keep the A32 reader from this point of view.
This patch adds an option to the ALT reader increasing the bit
cache to 32 bits as required by the Prores decoder. Benchmarking
shows no significant change in speed on Intel i7. Again, the
A32 reader fails to justify its existence.
Signed-off-by: Mans Rullgard <mans@mansr.com>
* qatar/master:
get_bits: remove x86 inline asm in A32 bitstream reader
doc: Remove outdated information about our issue tracker
avidec: Factor out the sync fucntionality.
fate-aac: Expand coverage.
ac3dsp: add x86-optimized versions of ac3dsp.extract_exponents().
ac3dsp: simplify extract_exponents() now that it does not need to do clipping.
ac3enc: clip coefficients after MDCT.
ac3enc: add int32_t array clipping function to DSPUtil, including x86 versions.
swscale: for >8bit scaling, read in native bit-depth.
matroskadec: matroska_read_seek after after EBML_STOP leads to failure.
doxygen: fix usage of @file directive in libavutil/{dict,file}.h
doxygen: Help doxygen parser to understand the DECLARE_ALIGNED and offsetof macros
Conflicts:
doc/issue_tracker.txt
libavformat/avidec.c
libavutil/dict.h
libswscale/swscale.c
libswscale/utils.c
tests/ref/lavfi/pixfmts_scale
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master:
cosmetics: fix some then/than typos
doxygen: Include libavcodec and libavformat examples into the documentation
avutil: elaborate documentation for av_get_random_seed
Add support for aac streams in mp4/mov without extradata.
aes: whitespace cosmetics
adler32: whitespace cosmetics
swscale: fix another yuv range conversion overflow in 16bit scaling.
Fix cpu flags test program
opt-test: Add missing braces to silence compiler warnings.
build: Eliminate obsolete test targets.
udp: Fix a compilation warning
swscale: Unbreak build with --enable-small
base64: add fate test
aes: improve test program and add fate test
adler32: make test program more useful and add fate test
swscale: fix yuv range correction when using 16-bit scaling.
aacenc: Make chan_map const correct
Conflicts:
Makefile
doc/examples/muxing-example.c
libavformat/udp.c
libavutil/random_seed.h
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master: (23 commits)
doc: Check standalone compilation before submitting new components.
Fix standalone compilation of pipe protocol.
Fix standalone compilation of ac3_fixed encoder.
Fix standalone compilation of binkaudio_dct / binkaudio_rdft decoders.
Fix standalone compilation of IMC decoder.
Fix standalone compilation of WTV demuxer.
Fix standalone compilation of MXPEG decoder.
flashsv: K&R cosmetics
matroskaenc: fix memory leak
vc1: make overlap filter for I-frames bit-exact.
vc1dec: use s->start/end_mb_y instead of passing them as function args.
Revert "VC1: merge idct8x8, coeff adjustments and put_pixels."
Replace strncpy() with av_strlcpy().
indeo3: Eliminate use of long.
get_bits: make cache unsigned to eliminate undefined signed overflow.
asfdec: fix assert failure on invalid files
avfilter: check malloc return values.
Not pulled as reason for reindent is not pulled: mpegvideo: reindent.
nutenc: check malloc return values.
Not pulled due to much simpler solution in ffmpeg *: don't av_malloc(0).
...
Conflicts:
doc/developer.texi
libavcodec/Makefile
libavcodec/get_bits.h
libavcodec/mpegvideo.c
libavformat/Makefile
libavutil/log.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
These whitespace changes improve the readability of the get_bits
macros.
Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit fb5c841d5f)
Some of the macros in get_bits.h include a final semicolon,
some do not. This removes these or adds do {} while(0) around
the macros as appropriate and adds semicolons where needed in
calling code.
Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit bf5f9b528b)
Using the libmpeg2 reader causes errors in a multitude of places,
including MPEG and H264 codecs. As the advantage of this reader
is questionable, removing it seems the sensible course of action,
especially considering the simplifications this allows elsewhere
with the bit cache size increasing from 17 to 25 bits as minimum.
Signed-off-by: Mans Rullgard <mans@mansr.com>
(cherry picked from commit 938f72e199)
Some of the macros in get_bits.h include a final semicolon,
some do not. This removes these or adds do {} while(0) around
the macros as appropriate and adds semicolons where needed in
calling code.
Signed-off-by: Mans Rullgard <mans@mansr.com>
Using the libmpeg2 reader causes errors in a multitude of places,
including MPEG and H264 codecs. As the advantage of this reader
is questionable, removing it seems the sensible course of action,
especially considering the simplifications this allows elsewhere
with the bit cache size increasing from 17 to 25 bits as minimum.
Signed-off-by: Mans Rullgard <mans@mansr.com>
Other parts of FFmpeg use NE (native endian) rather than ME (machine).
This makes it consistent.
Originally committed as revision 24169 to svn://svn.ffmpeg.org/ffmpeg/trunk
Passing an explicit filename to this command is only necessary if the
documentation in the @file block refers to a file different from the
one the block resides in.
Originally committed as revision 22921 to svn://svn.ffmpeg.org/ffmpeg/trunk
The limit for get/show_bits_long() to use get/show_bits() directly
should be MIN_CACHE_BITS, not 17.
Originally committed as revision 21951 to svn://svn.ffmpeg.org/ffmpeg/trunk
Any tips on how i can convince gcc that it doesnt need a
mov %eax, %eax
in every get_bits() ?
Originally committed as revision 21433 to svn://svn.ffmpeg.org/ffmpeg/trunk
