The previous implementation targeted DTS Coherent Acoustics, which only
requires nbits == 4 (fft16()). This case was (and still is) linked directly
rather than being indirected through ff_fft_calc_vfp(), but now the full
range from radix-4 up to radix-65536 is available. This benefits other codecs
such as AAC and AC3.
The implementaion is based upon the C version, with each routine larger than
radix-16 calling a hierarchy of smaller FFT functions, then performing a
post-processing pass. This pass benefits a lot from loop unrolling to
counter the long pipelines in the VFP. A relaxed calling standard also
reduces the overhead of the call hierarchy, and avoiding the excessive
inlining performed by GCC probably helps with I-cache utilisation too.
I benchmarked the result by measuring the number of gperftools samples that
hit anywhere in the AAC decoder (starting from aac_decode_frame()) or
specifically in the FFT routines (fft4() to fft512() and pass()) for the
same sample AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
Audio decode 2245.5 53.1 1599.6 43.8 100.0% +40.4%
FFT routines 940.6 22.0 348.1 20.8 100.0% +170.2%
Signed-off-by: Martin Storsjö <martin@martin.st>
The previous implementation targeted DTS Coherent Acoustics, which only
requires mdct_bits == 6. This relatively small size lent itself to
unrolling the loops a small number of times, and encoding offsets
calculated at assembly time within the load/store instructions of each
iteration.
In the more general case (codecs such as AAC and AC3) much larger arrays
are used - mdct_bits == [8, 9, 11]. The old method does not scale for
these cases, so more integer registers are used with non-unrolled versions
of the loops (and with some stack spillage). The postrotation filter loop
is still unrolled by a factor of 2 to permit the double-buffering of some
VFP registers to facilitate overlap of neighbouring iterations.
I benchmarked the result by measuring the number of gperftools samples
that hit anywhere in the AAC decoder (starting from aac_decode_frame())
or specifically in ff_imdct_half_c / ff_imdct_half_vfp, for the same
example AAC stream:
Before After
Mean StdDev Mean StdDev Confidence Change
aac_decode_frame 2368.1 35.8 2117.2 35.3 100.0% +11.8%
ff_imdct_half_* 457.5 22.4 251.2 16.2 100.0% +82.1%
Signed-off-by: Martin Storsjö <martin@martin.st>
Move the GNU as check before the arch specific asm checks since the .dn
check requires gas compatible assembler.
Disable the VC-1 motion compensation NEON asm which is the only part
using that directive. The integrated assembler in the upcoming clang 3.5
does not support .dn/.qn without plans to change that. Too much effort
to implement it while it is rarely used.
http://llvm.org/bugs/show_bug.cgi?id=18199.
Further performance improvements and security fixes by
Vittorio Giovara, Luca Barbato and Diego Biurrun.
Signed-off-by: Vittorio Giovara <vittorio.giovara@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
Signed-off-by: Diego Biurrun <diego@biurrun.de>
Profiling results for overall decode and the output_data function in
particular are as follows:
Before After
Mean StdDev Mean StdDev Confidence Change
6:2 total 339.6 15.1 329.3 16.0 95.8% +3.1% (insignificant)
6:2 function 24.6 6.0 9.9 3.1 100.0% +148.5%
8:2 total 324.5 15.5 323.6 14.3 15.2% +0.3% (insignificant)
8:2 function 20.4 3.9 9.9 3.4 100.0% +104.7%
6:6 total 572.8 20.6 539.9 24.2 100.0% +6.1%
6:6 function 54.5 5.6 16.0 3.8 100.0% +240.9%
8:8 total 741.5 21.2 702.5 18.5 100.0% +5.6%
8:8 function 63.9 7.6 18.4 4.8 100.0% +247.3%
The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.
Signed-off-by: Martin Storsjö <martin@martin.st>
Profiling results for overall audio decode and the rematrix_channels function
in particular are as follows:
Before After
Mean StdDev Mean StdDev Confidence Change
6:2 total 370.8 17.0 348.8 20.1 99.9% +6.3%
6:2 function 46.4 8.4 45.8 6.6 18.0% +1.2% (insignificant)
8:2 total 343.2 19.0 339.1 15.4 54.7% +1.2% (insignificant)
8:2 function 38.9 3.9 40.2 6.9 52.4% -3.2% (insignificant)
6:6 total 658.4 15.7 604.6 20.8 100.0% +8.9%
6:6 function 109.0 8.7 59.5 5.4 100.0% +83.3%
8:8 total 896.2 24.5 766.4 17.6 100.0% +16.9%
8:8 function 223.4 12.8 93.8 5.0 100.0% +138.3%
The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.
Signed-off-by: Martin Storsjö <martin@martin.st>
Profiling results for overall audio decode and the mlp_filter_channel(_arm)
function in particular are as follows:
Before After
Mean StdDev Mean StdDev Confidence Change
6:2 total 380.4 22.0 370.8 17.0 87.4% +2.6% (insignificant)
6:2 function 60.7 7.2 36.6 8.1 100.0% +65.8%
8:2 total 357.0 17.5 343.2 19.0 97.8% +4.0% (insignificant)
8:2 function 60.3 8.8 37.3 3.8 100.0% +61.8%
6:6 total 717.2 23.2 658.4 15.7 100.0% +8.9%
6:6 function 140.4 12.9 81.5 9.2 100.0% +72.4%
8:8 total 981.9 16.2 896.2 24.5 100.0% +9.6%
8:8 function 193.4 15.0 103.3 11.5 100.0% +87.2%
Experiments with adding preload instructions to this function yielded no
useful benefit, so these have not been included.
The assembly version has also been tested with a fuzz tester to ensure that
any combinations of inputs not exercised by my available test streams still
generate mathematically identical results to the C version.
Signed-off-by: Martin Storsjö <martin@martin.st>
The function is assigned to a function pointer that does not have the
restrict keyword for that parameter.
This fixes compilation for MSVC builds that don't recognize "restrict",
broken since ed9625eb62.
Based on the aarch64 asm. CPU cycle counts on cortex-a9 compared to
gcc 4.8.2:
before: 475 decicycles in get_cabac_noinline, 67106035 runs, 2829 skips
after: 393 decicycles in get_cabac_noinline, 67106474 runs, 2390 skips
Overall speedup is above 2%. Code generated by clang 3.4 is slower on
the same hardware and the relative change is a little larger.
The overread avoidance fix in cbddee1cca
broke the computation for the last row since it prevented the safe
reading from the height+1-th row.
CC: libav-stable@libav.org
The vector dequantization has a test in a loop preventing effective SIMD
implementation. By moving it out of the loop, this loop can be DSPized.
Therefore, modify the current DSP implementation. In particular, the
DSP implementation no longer has to handle null loop sizes.
The decode_hf implementations have following timings:
For x86 Arrandale:
C SSE SSE2 SSE4
win32: 260 162 119 104
win64: 242 N/A 89 72
The arm NEON optimizations follow in a later patch as external asm. The
now unused check for the y modifier in arm inline asm is removed from
configure.
The scaling factor is constant so it is faster to scale the
FIR coefficients in the tables during compilation.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>
The x86 runs short on registers because numerous elements are not static.
In addition, splitting them allows more optimized code, at least for x86.
Arm asm changes by Janne Grunau.
Signed-off-by: Janne Grunau <janne-libav@jannau.net>
It is currently declared as a macro who is set to inlinable functions,
among which a Neon and a default C implementations.
Add a DSP parameter to each inline function, unused except by the
default C implementation which calls a function from the DSP context.
On an Arrandale CPU, gain for an inlined SSE2 function vs. a call:
- Win32: 29 to 26 cycles
- Win64: 25 to 23 cycles
Signed-off-by: Janne Grunau <janne-libav@jannau.net>
The function macro always sets .align 2 before declaring the
function label (since 5c5e1ea3) and always sets the section to
.text (since 278caa6a).
The .align 5 before certain functions, added in fc252eba, were added
before .text and .align were added to the function macro and thus
became useless/unused when the function macro got them.
This restores the original intention, to align the loop entry
points.
Signed-off-by: Martin Storsjö <martin@martin.st>
This file no longer uses the pld instruction at all, all such uses
have been split into hpeldsp_arm.S.
Signed-off-by: Martin Storsjö <martin@martin.st>
For:
ff_vc1_inv_trans_{8,4}x{8,4}_{dc_,}neon
ff_put_pixels8x8_neon
ff_put_vc1_mspel_mc{0,1,2,3}{0,1,2,3}_neon (except for 00)
Based on ARM assembly code in libavcodec/arm by Rob Clark and Mans
Rullgard.
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 508.8 23.4 185.4 9.0 +174.4%
Overall 3068.5 31.7 2752.1 29.4 +11.5%
In combination with the preceding patch:
Before After
Mean StdDev Mean StdDev Change
Overall 2925.6 26.2 2752.1 29.4 +6.3%
Signed-off-by: Martin Storsjö <martin@martin.st>
When building for iOS in thumb mode, gas-preprocessor.pl doesn't
mark unused labels as thumb functions (as it does for other
local labels, where it can figure out that they are functions
due to being referenced in branch instructions). This leads to
linker warnings for some of those local labels, such as:
ld: warning: ARM function not 4-byte aligned: __a_evaluation from
libavcodec/libavcodec.a(simple_idct_arm.o)
Therefore, comment them out since they don't have any function.
They do still have a value in documenting key points in the
assembly source though.
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 1323.0 98.0 746.2 60.6 +77.3%
Overall 15400.0 336.4 14147.5 288.4 +8.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 1389.3 4.2 967.8 35.1 +43.6%
Overall 15577.5 83.2 15400.0 336.4 +1.2%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 868.2 33.5 436.0 27.0 +99.1%
Overall 15973.0 223.2 15577.5 83.2 +2.5%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 2653.0 28.5 1108.8 51.4 +139.3%
Overall 17049.5 408.2 15973.0 223.2 +6.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 366.2 18.3 277.8 13.7 +31.9%
Overall 18420.5 489.1 17049.5 408.2 +8.0%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 1175.0 4.4 366.2 18.3 +220.8%
Overall 19285.5 292.0 18420.5 489.1 +4.7%
Signed-off-by: Martin Storsjö <martin@martin.st>
Before After
Mean StdDev Mean StdDev Change
This function 9295.0 114.9 4853.2 83.5 +91.5%
Overall 23699.8 397.6 19285.5 292.0 +22.9%
Signed-off-by: Martin Storsjö <martin@martin.st>
A few of the h264qpel neon functions are shared with other
hpeldsp functions in this file.
This fixes standalone compilation of the h264 decoder on arm.
Signed-off-by: Martin Storsjö <martin@martin.st>
This way, the special IDCT permutations are no longer needed. This
is similar to how H264 does it, and removes the dsputil dependency
imposed by the scantable code.
Also remove the unused type == 0 cases from the plain C version
of the idct.
Signed-off-by: Martin Storsjö <martin@martin.st>
The non-intra-pcm branch in hl_decode_mb (simple, 8bpp) goes from 700
to 672 cycles, and the complete loop of decode_mb_cabac and hl_decode_mb
(in the decode_slice loop) goes from 1759 to 1733 cycles on the clip
tested (cathedral), i.e. almost 30 cycles per mb faster.
Signed-off-by: Martin Storsjö <martin@martin.st>
This makes the plain-armv6 version use the same registers as the
armv6t2 version above.
This fixes fate-vp8 on plain-armv6 devices.
Signed-off-by: Martin Storsjö <martin@martin.st>
The sh4 optimizations are removed, because the code is
100% identical to the C code, so it is unlikely to
provide any real practical benefit.
Signed-off-by: Diego Biurrun <diego@biurrun.de>
Signed-off-by: Ronald S. Bultje <rsbultje@gmail.com>
Signed-off-by: Luca Barbato <lu_zero@gentoo.org>
