This adds a hand-optimized assembly version for get_cabac much like the
existing one, but it works if the table offsets are RIP-relative.
Compared to the non-RIP-relative version this adds 2 lea instructions
and it needs one extra register.
There is a surprisingly large performance improvement over the c version (more
so than the generated assembly seems to suggest) just in get_cabac, I measured
roughly 40% faster for get_cabac on a K8. However, overall the difference is
not that big, I measured roughly 5% on a test clip on a K8 and a Core2.
Hopefully it still compiles on x86 32bit...
v2: incorporated feedback from Loren Merritt to avoid rip-relative movs
for every table, and got rid of unnecessary @GOTPCREL.
v3: apply similar fixes to the the decode_significance functions, and use
same macro arguments for non-pic case.
v4: prettify inline asm arguments, add a non-fast-cmov version (as I expect
the c code to be faster otherwise since both cmov and sbb suck hard on a
Prescott, even can't construct the mask with a 64bit shift as that's just as
terrible - it's quite difficult to find usable instructions on that chip...).
This is tested to work but not on a P4, in theory it _should_ be fast there.
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master:
avcodec: add a cook parser to get subpacket duration
FATE: allow lavf tests to alter input parameters
FATE: replace the acodec-pcm_s24daud test with an enc_dec_pcm checksum test
FATE: replace the acodec-g726 test with 4 new encode/decode tests
FATE: replace current g722 encoding tests with an encode/decode test
FATE: add a pattern rule for generating asynth wav files
FATE: optionally write a WAVE header in audiogen
avutil: add audio fifo buffer
Conflicts:
doc/APIchanges
libavcodec/version.h
libavutil/avutil.h
tests/Makefile
tests/codec-regression.sh
tests/fate/voice.mak
tests/lavf-regression.sh
tests/ref/acodec/g722
tests/ref/acodec/g726
tests/ref/acodec/pcm_s24daud
tests/ref/lavf/dv_fmt
tests/ref/lavf/gxf
tests/ref/lavf/mxf
tests/ref/lavf/mxf_d10
tests/ref/seek/lavf_dv
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master:
dv: Initialize encoder tables during encoder init.
dv: Replace some magic numbers by the appropriate #define.
FATE: pass the decoded output format and audio source file to enc_dec_pcm
FATE: specify the input format when decoding in enc_dec_pcm()
x86inc: support AVX abstraction for 2-operand instructions
configure: detect PGI compiler and set suitable flags
avconv: check for an incompatible changing channel layout
avio: make AVIOContext.av_class pointer to const
nutdec: add malloc check and fix const to non-const conversion warnings
Conflicts:
ffmpeg.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
In normal picture decoding this does not need to be checked but as
error concealment is run in the case of errors the availability of
references is less certain. This may be fixed differently at some
point so that all references are always filled in before the EC
code, in which case this should then be changed to an assert()
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* hexene/stagefright:
libstagefright: avoid memory leak
libstagefright: support more output pixel formats
libstagefright: avoid potential deadlock on output MediaBuffer
libstagefright: explicitly set positive timestamps as stagefright expects them so
Merge branches 'stagefright' and 'stagefright-test' into stagefright-test
Merged-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master:
ppc: drop unused function dct_quantize_altivec()
mpegaudiodec: Do not discard mp_decode_frame() return value.
matroska: do not set invalid default duration if frame rate is zero
mkv: use av_reduce instead of av_d2q for framerate estimation
mkv: report average framerate as minimal as well
avcodec_string: Favor AVCodecContext.codec over the default codec.
cook: Make constants passed to AV_BE2NE32C() unsigned to avoid signed overflow.
Conflicts:
libavcodec/cook.c
libavcodec/ppc/mpegvideo_altivec.c
libavcodec/utils.c
libavformat/matroskadec.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>
Maintain an output queue of AVFrames instead of MediaBuffers
so that the latter can be released early. This avoids a potential deadlock
between the stagefright decoder::read() and Stagefright_decode_frame()
This fixes crashes with frame threads caused by inconsistent context parameters.
Fixes Ticket1207
Found-by: John Villamil
Signed-off-by: Michael Niedermayer <michaelni@gmx.at>
* qatar/master:
aacenc: Fix issues with huge values of bit_rate.
dv_tablegen: Drop unnecessary av_unused attribute from dv_vlc_map_tableinit().
proresenc: multithreaded quantiser search
riff: use bps instead of bits_per_coded_sample in the WAVEFORMATEXTENSIBLE header
avconv: only set the "channels" option when it exists for the specified input format
avplay: update get_buffer to be inline with avconv
aacdec: More robust output configuration.
faac: Fix multi-channel ordering
faac: Add .channel_layouts
rtmp: Support 'rtmp_playpath', an option which overrides the stream identifier
rtmp: Support 'rtmp_app', an option which overrides the name of application
avutil: add better documentation for AVSampleFormat
Conflicts:
libavcodec/aac.h
libavcodec/aacdec.c
libavcodec/aacenc.c
Merged-by: Michael Niedermayer <michaelni@gmx.at>