Directly branch into the special 64-point deinterleave
subroutine rather than going through the general deinterleave.
64-point transform timings on Zen 3:
Before:
1974 decicycles in av_tx (fft),16776864 runs, 352 skips
After:
1956 decicycles in av_tx (fft),16775378 runs, 1838 skips
x64 always has MMX, MMXEXT, SSE and SSE2 and this means
that some functions for MMX, MMXEXT, SSE and 3dnow are always
overridden by other functions (unless one e.g. explicitly
disables SSE2). So given that the only systems which benefit
from ff_vector_fmul_window_3dnowext are truely ancient 32bit
AMD x86s it is removed.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
x64 always has MMX, MMXEXT, SSE and SSE2 and this means
that some functions for MMX, MMXEXT, SSE and 3dnow are always
overridden by other functions (unless one e.g. explicitly
disables SSE2). So given that the only systems which benefit
from the 8x8 MMX (overridden by MMXEXT) or the 16x16 MMXEXT
(overridden by SSE2) are truely ancient 32bit x86s they are removed.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
When the SLOW_GATHER flag was added to the AVX2 version, this
made FMA3-features not enabled on Zen CPUs.
As FMA3 adds 6-7% across all platforms that support it, in
the interest of saving space, this commit removes the AVX
version and replaces it with an FMA3 version.
The only CPUs affected are Sandy Bridge and Bulldozer, which
have AVX support, but no FMA3 support.
In the future, if there's a demand for it, a version of the
function duplicated for AVX can be added.
This reverts commit 82a68a8771.
Smarter slow ISA penalties makes gathers still useful.
The intention is to use gathers with the final stage of non-ptwo iMDCTs,
where they give benefit.
Its performance loss ranges from either being just as fast as individual loads
(Skylake), a few percent slower (Alderlake), 8% slower (Zen 3), to completely
disasterous (older/other CPUs).
Sadly, gathers never panned out fast on x86, even with the benefit of time and
implementation experience.
This also saves a register, as there's no need to fill out an additional
register mask.
Zen 3 (16384-point transform):
Before: 1561050 decicycles in av_tx (fft), 131072 runs, 0 skips
After: 1449621 decicycles in av_tx (fft), 131072 runs, 0 skips
Alderlake:
2% slower on big transforms (65536), to 1% (131072), to a few percent for smaller
sizes.
Some of these were made possible by moving several common macros to
libavutil/macros.h.
While just at it, also improve the other headers a bit.
Reviewed-by: Martin Storsjö <martin@martin.st>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Only include it if it is needed, namely if __MMX__ is undefined.
X86 is currently the only arch where lavu/cpu.h is basically
automatically included (for internal development): #if ARCH_X86
is true, lavu/internal.h (which is basically included everywhere)
includes lavu/x86/emms.h which can mask missing inclusions
of lavu/cpu.h if the developer works on x86/x64. This has happened
in 8e825ec3ab and also earlier
(see 6d2365882f).
By including said header only if necessary ordinary developer machines
will behave like non-x86 arches, so that missing inclusions of cpu.h
won't go unnoticed any more.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Otherwise nasm writes the full host-specific paths into .o
output, which breaks binary reproducibility.
Signed-off-by: Alexander Kanavin <alex.kanavin@gmail.com>
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This commit does some refactoring to make defining assembly codelets
smaller, and fixes compiler redefinition warnings. It also allows
for other assembly versions to reuse the same boilerplate code as
x86.
Finally, it also adds the out_of_place flag to all assembly codelets.
This changes nothing, as out-of-place operation was assumed to be
available anyway, but this makes it more explicit.
Makes Bulldozer prefer AVX functions rather than AVX2,
which are 64% slower:
AVX: 117653 decicycles in av_tx (fft), 1048535 runs, 41 skips
AVX2: 193385 decicycles in av_tx (fft), 1048561 runs, 15 skips
The only difference between both is that vgatherdpd is used in
the former. We don't want to mark them with the new SLOW_GATHER
flag however, since gathers are still faster on Haswell/Zen 2/3
than plain loads.
This broke builds with --disable-mmx, which also disabled assembly
entirely, but ARCH_X86 was still true, so the init file tried to find
assembly that didn't exist.
Instead of checking for architecture, check if external x86 assembly
is enabled.
This commit rewrites the internal transform code into a constructor
that stitches transforms (codelets).
This allows for transforms to reuse arbitrary parts of other
transforms, and allows transforms to be stacked onto one
another (such as a full iMDCT using a half-iMDCT which in turn
uses an FFT). It also permits for each step to be individually
replaced by assembly or a custom implementation (such as an ASIC).
Changes av_clipf to return amin if a is nan.
Before if a is nan av_clipf_c returned nan and
av_clipf_sse would return amax. Now the both
should behave the same.
This works because nan > amin is false.
The max(nan, amin) will be amin.
Signed-off-by: James Almer <jamrial@gmail.com>
The field is a standard field, yet we were loading it as if it was
a quadword. This worked for forward transforms by chance, but broke
when the transform was inverse.
checkasm couldn't catch that because we only test forward transforms,
which are identical to inverse transforms but with a different revtab.
This commit adds a pure x86 assembly SIMD version of the FFT in libavutil/tx.
The design of this pure assembly FFT is pretty unconventional.
On the lowest level, instead of splitting the complex numbers into
real and imaginary parts, we keep complex numbers together but split
them in terms of parity. This saves a number of shuffles in each transform,
but more importantly, it splits each transform into two independent
paths, which we process using separate registers in parallel.
This allows us to keep all units saturated and lets us use all available
registers to avoid dependencies.
Moreover, it allows us to double the granularity of our per-load permutation,
skipping many expensive lookups and allowing us to use just 4 loads per register,
rather than 8, or in case FMA3 (and by extension, AVX2), use the vgatherdpd
instruction, which is at least as fast as 4 separate loads on old hardware,
and quite a bit faster on modern CPUs).
Higher up, we go for a bottom-up construction of large transforms, foregoing
the traditional per-transform call-return recursion chains. Instead, we always
start at the bottom-most basis transform (in this case, a 32-point transform),
and continue constructing larger and larger transforms until we return to the
top-most transform.
This way, we only touch the stack 3 times per a complete target transform:
once for the 1/2 length transform and two times for the 1/4 length transform.
The combination algorithm we use is a standard Split-Radix algorithm,
as used in our C code. Although a version with less operations exists
(Steven G. Johnson and Matteo Frigo's "A modified split-radix FFT with fewer
arithmetic operations", IEEE Trans. Signal Process. 55 (1), 111–119 (2007),
which is the one FFTW uses), it only has 2% less operations and requires at least 4x
the binary code (due to it needing 4 different paths to do a single transform).
That version also has other issues which prevent it from being implemented
with SIMD code as efficiently, which makes it lose the marginal gains it offered,
and cannot be performed bottom-up, requiring many recursive call-return chains,
whose overhead adds up.
We go through a lot of effort to minimize load/stores by keeping as much in
registers in between construcring transforms. This saves us around 32 cycles,
on paper, but in reality a lot more due to load/store aliasing (a load from a
memory location cannot be issued while there's a store pending, and there are
only so many (2 for Zen 3) load/store units in a CPU).
Also, we interleave coefficients during the last stage to save on a store+load
per register.
Each of the smallest, basis transforms (4, 8 and 16-point in our case)
has been extremely optimized. Our 8-point transform is barely 20 instructions
in total, beating our old implementation 8-point transform by 1 instruction.
Our 2x8-point transform is 23 instructions, beating our old implementation by
6 instruction and needing 50% less cycles. Our 16-point transform's combination
code takes slightly more instructions than our old implementation, but makes up
for it by requiring a lot less arithmetic operations.
Overall, the transform was optimized for the timings of Zen 3, which at the
time of writing has the most IPC from all documented CPUs. Shuffles were
preferred over arithmetic operations due to their 1/0.5 latency/throughput.
On average, this code is 30% faster than our old libavcodec implementation.
It's able to trade blows with the previously-untouchable FFTW on small transforms,
and due to its tiny size and better prediction, outdoes FFTW on larger transforms
by 11% on the largest currently supported size.
Some files currently rely on libavutil/cpu.h to include it for them;
yet said file won't use include it any more after the currently
deprecated functions are removed, so include attributes.h directly.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Some new warnings regarding use of empty macro parameters has
been added, so adjust some x86inc code to silence those.
Fixes part of ticket #8771
Signed-off-by: James Almer <jamrial@gmail.com>
When clang works in MSVC mode, it does have the _byteswap_ulong
builtin, but one has to include stdlib.h before using it.
Signed-off-by: Martin Storsjö <martin@martin.st>
There are 32 pseudo-instructions for each floating-point comparison
instruction, but only 8 of them are actually valid in legacy-encoded mode.
The remaining 24 requires the use of VEX-encoded (v-prefixed) instructions
and can therefore be disregarded for this purpose.
AVX-512 consists of a plethora of different extensions, but in order to keep
things a bit more manageable we group together the following extensions
under a single baseline cpu flag which should cover SKL-X and future CPUs:
* AVX-512 Foundation (F)
* AVX-512 Conflict Detection Instructions (CD)
* AVX-512 Byte and Word Instructions (BW)
* AVX-512 Doubleword and Quadword Instructions (DQ)
* AVX-512 Vector Length Extensions (VL)
On x86-64 AVX-512 provides 16 additional vector registers, prefer using
those over existing ones since it allows us to avoid using `vzeroupper`
unless more than 16 vector registers are required. They also happen to
be volatile on Windows which means that we don't need to save and restore
existing xmm register contents unless more than 22 vector registers are
required.
Big thanks to Intel for their support.