This speeds up aac_tablegen to a ludicruous degree (~97%), i.e to the point
where it can be argued that runtime initialization can always be done instead of
hard-coded tables. The only cost is essentially a trivial increase in
the stack size.
Even if one does not care about this, the patch also improves accuracy
as detailed below.
Performance:
Benchmark obtained by looping 10^4 times over ff_aac_tableinit.
Sample benchmark (x86-64, Haswell, GNU/Linux):
old:
1295292 decicycles in ff_aac_tableinit, 512 runs, 0 skips
1275981 decicycles in ff_aac_tableinit, 1024 runs, 0 skips
1272932 decicycles in ff_aac_tableinit, 2048 runs, 0 skips
1262164 decicycles in ff_aac_tableinit, 4096 runs, 0 skips
1256720 decicycles in ff_aac_tableinit, 8192 runs, 0 skips
new:
21112 decicycles in ff_aac_tableinit, 511 runs, 1 skips
21269 decicycles in ff_aac_tableinit, 1023 runs, 1 skips
21352 decicycles in ff_aac_tableinit, 2043 runs, 5 skips
21386 decicycles in ff_aac_tableinit, 4080 runs, 16 skips
21299 decicycles in ff_aac_tableinit, 8173 runs, 19 skips
Accuracy:
The previous code was resulting in needless loss of
accuracy due to the pow being called in succession. As an illustration
of this:
ff_aac_pow34sf_tab[3]
old : 0.000000000007598092294225
new : 0.000000000007598091426864
real: 0.000000000007598091778545
truncated to float
old : 0.000000000007598092294225
new : 0.000000000007598091426864
real: 0.000000000007598091426864
showing that the old value was not correctly rounded. This affects a
large number of elements of the array.
Patch tested with FATE.
Reviewed-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Avoid decoding a frame to get the codec parameters while the codec
supports FF_CODEC_CAP_SKIP_FRAME_FILL_PARAM. This is particulary useful
to avoid decoding twice images (once in avformat_find_stream_info and
once when the actual decode is made).
This hugely reduces the echo which was introduced with the previous
commit (though likely because previously everything was broken).
Makes LTP actually worthwhile now.
Signed-off-by: Rostislav Pehlivanov <atomnuker@gmail.com>
Should fix issues with ppc, tested by bug reporter.
Reported-by: John Warburton <john@johnwarburton.net>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
This removes wasteful pow(x, 2.0) that although not terribly important
for speed, is still useless.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
Gain computation for various curves was being done in a needlessly
inaccurate fashion. Of course these are all subjective curves, but when
a curve is advertised to the user, it should be matched as closely as
possible within the limitations of libm. In particular, the constants
kept here were pretty inaccurate for double precision.
Speed improvements are mainly due to the avoidance of pow, the most
notorious of the libm functions in terms of performance. To be fair, it
is the GNU libm that is among the worst, but it is not really GNU libm's fault
since others simply yield a higher error as measured in ULP.
"Magic" constants are also accordingly documented, since they take at
least a minute of thought for a casual reader.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
This is a trivial rewrite of the loops that results in better
prefetching and associated cache efficiency. Essentially, the problem is
that modern prefetching logic is based on finite state Markov memory, a reasonable
assumption that is used elsewhere in CPU's in for instance branch
predictors.
Surrounding loops all iterate forward through the array, making the
predictor think of prefetching in the forward direction, but the
intermediate loop is unnecessarily in the backward direction.
Speedup is nontrivial. Benchmarks obtained by 10^6 iterations within
solve_lls, with START/STOP_TIMER. File is tests/data/fate/flac-16-lpc-cholesky.err.
Hardware: x86-64, Haswell, GNU/Linux.
new:
17291 decicycles in solve_lls, 2096706 runs, 446 skips
17255 decicycles in solve_lls, 4193657 runs, 647 skips
17231 decicycles in solve_lls, 8384997 runs, 3611 skips
17189 decicycles in solve_lls,16771010 runs, 6206 skips
17132 decicycles in solve_lls,33544757 runs, 9675 skips
17092 decicycles in solve_lls,67092404 runs, 16460 skips
17058 decicycles in solve_lls,134188213 runs, 29515 skips
old:
18009 decicycles in solve_lls, 2096665 runs, 487 skips
17805 decicycles in solve_lls, 4193320 runs, 984 skips
17779 decicycles in solve_lls, 8386855 runs, 1753 skips
18289 decicycles in solve_lls,16774280 runs, 2936 skips
18158 decicycles in solve_lls,33548104 runs, 6328 skips
18420 decicycles in solve_lls,67091793 runs, 17071 skips
18310 decicycles in solve_lls,134187219 runs, 30509 skips
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
In some conditions, where the first band was being zeroed
mainly, the wrong global gain scalefactor would be written
to the stream since it's always taken from the first band
regardless of whether it's been marked as zero or not.
So, always make sure it contians something useful.
When both M/S coding and PNS are enabled, scalefactors
and coding books would be mistakenly clobbered when setting
the M/S flag on PNS'd bands. The flag needs to be set to
signal the generation of correlated noise, but the scalefactors,
coefficients and the coding books need to be kept intact.
Commit 14ea4151d7 had a bug in that the
conversion of the uint64_t result to an int (the return signature) would
lead to implementation defined behavior, and in this case simply
returned 0 for NAN. A fix via AND'ing the result with 1 does the trick,
simply by ensuring a 0 or 1 return value.
Patch tested with FATE on x86-64, GNU/Linux by forcing the compatibility
code via an ifdef hack suggested by Michael.
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
lrintf is anyway used, suggesting we only care up to floating precision.
Rurthermore, there is a compat hack in avutil/libm for this function,
and it is used in avcodec/aacps_tablegen.h.
This yields a non-negligible speedup. Sample benchmark:
x86-64, Haswell, GNU/Linux:
old (draw_mandelbrot):
274635709 decicycles in draw_mandelbrot, 256 runs, 0 skips
300287046 decicycles in draw_mandelbrot, 512 runs, 0 skips
371819935 decicycles in draw_mandelbrot, 1024 runs, 0 skips
336663765 decicycles in draw_mandelbrot, 2048 runs, 0 skips
581851016 decicycles in draw_mandelbrot, 4096 runs, 0 skips
new (draw_mandelbrot):
269882717 decicycles in draw_mandelbrot, 256 runs, 0 skips
296359285 decicycles in draw_mandelbrot, 512 runs, 0 skips
370076599 decicycles in draw_mandelbrot, 1024 runs, 0 skips
331478354 decicycles in draw_mandelbrot, 2048 runs, 0 skips
571904318 decicycles in draw_mandelbrot, 4096 runs, 0 skips
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Ganesh Ajjanagadde <gajjanagadde@gmail.com>
This option can be used to select useful frames from an ffconcat file which is
using inpoints and outpoints but where the source files are not intra frame
only.
Reviewed-by: Stefano Sabatini <stefasab@gmail.com>
Signed-off-by: Marton Balint <cus@passwd.hu>
If duration is still AV_NOPTS_VALUE when opening the next file, we can assume
that outpoint is not set.
Reviewed-by: Nicolas George <george@nsup.org>
Signed-off-by: Marton Balint <cus@passwd.hu>
Fixes out of array access
Fixes: 1430e9c43fae47a24c179c7c54f94918/signal_sigsegv_421427_2049_f2192b6829ab6e0eefcb035329c03c60.264
Found-by: Mateusz "j00ru" Jurczyk and Gynvael Coldwind
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
