They don't modify AVCodec, no needs to call it at register. They will be
wasteful if these codecs are unused. Instead, call static data initialization
at codecs' init.
Benchmark:
old: 51281340 decicycles in avcodec_register_all, 1 runs, 0 skips
new: 6738960 decicycles in avcodec_register_all, 1 runs, 0 skips
Reviewed-by: wm4 <nfxjfg@googlemail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
That line has been written by previous job.
Fix tsan warning.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
Also remove pthread_cond_broadcast(progress_cond) on uninit.
Broadcasting it is not required because workers are always
parked when they are not in thread_execute. So it is imposible
that a worker is waiting on progress_cond when uninitialized.
Benchmark:
./ffmpeg -threads $threads -thread_type slice -i 10slices.mp4 -f null null
threads=2:
old: 70.212s 70.525s 70.877s
new: 65.219s 65.377s 65.484s
threads=3:
old: 65.086s 66.306s 66.409s
new: 63.229s 65.026s 65.116s
threads=4:
old: 60.993s 61.482s 62.123s
new: 59.224s 59.441s 59.667s
threads=5:
old: 57.576s 57.860s 58.832s
new: 53.032s 53.948s 54.086s
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
Rework it to improve performance. Now mutex is not shared by workers,
instead each worker has its own mutex and condition variable. This
reduces lock contention between workers. Also use atomic variable for
counter.
The interface also allows execute to run special function on main
thread, requested by Ronald.
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
It is redundant with costable. The first half of sintable is
identical with the second half of costable. The second half
of sintable is negative value of the first half of sintable.
The computation is changed to handle sign of sin values, in
C code and ARM assembly code.
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
Modifying data pointer when skipping samples may make it unaligned.
Workaround for Ticket6349.
This should fix the crash of ticket's testcase and a crash/regression
with avxsynth (reported by Michael Niedermayer).
Also change frame->nb_samples < max to frame->nb_samples <= max.
This improves performance. Benchmark:
./ffmpeg -filter_complex "aevalsrc=0:n=1166,firequalizer=fixed=on" -f null null
old:
25767 decicycles in take_samples, 1023 runs, 1 skips
25422 decicycles in take_samples, 2047 runs, 1 skips
25181 decicycles in take_samples, 4095 runs, 1 skips
24904 decicycles in take_samples, 8191 runs, 1 skips
new:
550 decicycles in take_samples, 1024 runs, 0 skips
548 decicycles in take_samples, 2048 runs, 0 skips
545 decicycles in take_samples, 4096 runs, 0 skips
544 decicycles in take_samples, 8192 runs, 0 skips
Reviewed-by: Nicolas George <george@nsup.org>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
See http://lists.ffmpeg.org/pipermail/ffmpeg-user/2017-April/035975.html
Parsed_filter_X could remain and user can override it with custom one.
Example:
ffplay -f lavfi "nullsrc=s=640x360,
sendcmd='1 drawtext@top reinit text=Hello; 2 drawtext@bottom reinit text=World',
drawtext@top=x=16:y=16:fontsize=20:fontcolor=Red:text='',
drawtext@bottom=x=16:y=340:fontsize=16:fontcolor=Blue:text=''"
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
Because write_packet() fakely writes packets to muxer by queueing
them when muxer hasn't been initialized, it should also increment
frame_number fakely.
This is required because code in do_streamcopy() rely on
frame_number.
Should fix Ticket6227
Reviewed-by: James Almer <jamrial@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
So, all frames and errors are correctly reported in order.
Also limit the numbers of error during draining to prevent infinite loop.
This fix fate failure with THREADS>=4:
make fate-h264-attachment-631 THREADS=4
This also reverts a755b725ec.
Suggested-by: wm4, Ronald S. Bultje, Marton Balint
Reviewed-by: w4 <nfxjfg@googlemail.com>
Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
use fltp when doing s32 -> s32 resampling
because s32p has no simd optimization
benchmark:
old 17.913s
new 7.584s (use fma3)
Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
when set_compensation is called with zero sample_delta,
compensation does not happen (because dst_incr == ideal_dst_incr)
but compensation_distance is set
regression since 01ebb57c03
Found-by: wm4 <nfxjfg@googlemail.com>
Reviewed-by: wm4 <nfxjfg@googlemail.com>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
so tsf option in aresample will have effect
previously tsf/internal_sample_format had no effect
fate is updated
s32p previously used fltp internally
dblp previously used fltp/dblp internally
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
except filter_length == 1
odd filter_length gives worse frequency response,
even when compared with shorter filter_length
also makes build_filter simpler
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
integrate it inside multiple_resample
allow some calculations to be performed outside loop
Suggested-by: Michael Niedermayer <michael@niedermayer.cc>
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
give very bad quality for soxr resampler.
linear_interp is intended for using linear interpolation
between filter bank so quality will be better.
i guess this is misunderstood as 'do not use filter bank,
but directly interpolate linearly between samples'.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
separate dsp.resample to dsp.resample_common and dsp.resample_linear
and choose to call faster resample_common even when linear_interp=on
when c->frac and c->dst_incr_mod are both zero
speed up resampling when exact_rational and linear_interp are both
enabled because exact_rational force c->frac and c->dst_incr_mod to
be zero when soft compensation does not happen
benchmark on exact_rational=on:linear_interp=on
old new
real 8.432s 5.097s
user 7.679s 4.989s
sys 0.125s 0.107s
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
2-channels convolution using complex fft
improves speed significantly
not sure if it should be enabled by default
so disable it by default
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
high phase_count is only useful when dst_incr_mod is non zero
in other word, it is only useful on soft compensation
on init, it will build filter with low phase_count
but when soft compensation is enabled, rebuild filter
with high phase_count
this approach saves lots of memory
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
because exact_rational does not guarantee
that phase_count is even
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
this allow a filter to be written like this:
aformat =
sample_fmts = fltp|flt:
sample_rates = 44100|44800
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
give high quality resampling
as good as with linear_interp=on
as fast as without linear_interp=on
tested visually with ffplay
ffplay -f lavfi "aevalsrc='sin(10000*t*t)', aresample=osr=48000, showcqt=gamma=5"
ffplay -f lavfi "aevalsrc='sin(10000*t*t)', aresample=osr=48000:linear_interp=on, showcqt=gamma=5"
ffplay -f lavfi "aevalsrc='sin(10000*t*t)', aresample=osr=48000:exact_rational=on, showcqt=gamma=5"
slightly speed improvement
for fair comparison with -cpuflags 0
audio.wav is ~ 1 hour 44100 stereo 16bit wav file
ffmpeg -i audio.wav -af aresample=osr=48000 -f null -
old new
real 13.498s 13.121s
user 13.364s 12.987s
sys 0.131s 0.129s
linear_interp=on
old new
real 23.035s 23.050s
user 22.907s 22.917s
sys 0.119s 0.125s
exact_rational=on
real 12.418s
user 12.298s
sys 0.114s
possibility to decrease memory usage if soft compensation is ignored
Signed-off-by: Muhammad Faiz <mfcc64@gmail.com>
correct output pts based on input pts
make seeking possible
output frame one by one on eof
tested with showinfo filter
Suggested-by: Paul B Mahol <onemda@gmail.com>
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
fix default basefreq/endfreq comparison
on platform that does not do comparison
in double type
found on zeranoe 32-bit build, where
default freq range is detected as non-default
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
this makes draw_bar faster
slightly different result with old version
check result (with ~3 minutes audio file):
old:
real 0m49.611s
user 0m49.260s
sys 0m0.073s
new:
real 0m47.606s
user 0m47.378s
sys 0m0.068s
PSNR between old and new:
yuv444p PSNR
y:109.519298 u:107.506485 v:104.746878
average:106.816074 min:99.167305 max:inf
yuv422p PSNR
y:109.519298 u:108.025801 v:104.489734
average:107.279817 min:98.007467 max:inf
yuv420p PSNR
y:109.519298 u:108.363875 v:105.290200
average:108.261511 min:97.461812 max:inf
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
add yuv444p, yuv422p, and yuv420p output format (lower cpu usage
on ffplay playback because it does not do format conversion)
custom size with size/s option (fullhd option is deprecated)
custom layout with bar_h, axis_h, and sono_h option
support rational frame rate (within fps/r/rate option)
relaxed frame rate restriction (support fractional sample step)
support all input sample rates
separate sonogram and bargraph volume (with volume/sono_v and
volume2/bar_v)
timeclamp option alias (timeclamp/tc)
fcount option
gamma option alias (gamma/sono_g and gamma2/bar_g)
support custom frequency range (basefreq and endfreq)
support drawing axis using external image file (axisfile option)
alias for disabling drawing to axis (text/axis)
possibility to optimize it using arch specific asm code
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
faster initialization and less code
slightly different result computationally from previous
coeffclamp option is ignored
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Muhammad Faiz