Fixes: signed integer overflow: -1284837070 - 982101618 cannot be represented in type 'int'
Fixes: 53105/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_AC3_FIXED_fuzzer-4848015827664896
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Since D3D11 was introduced for QSV in FFmpeg 5.0, there is an implied
API/ABI change for user-supplied frames [1], hence update the
description for AV_PIX_FMT_QSV.
[1] https://ffmpeg.org/pipermail/ffmpeg-devel/2021-December/290444.html
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
Should fix fate failures on Windowx x86 targets, where long is 32 bits.
Reviewed-by: Martin Storsjö <martin@martin.st>
Signed-off-by: James Almer <jamrial@gmail.com>
The amount of lines printed is too high for the verbose level, and the debug
level is a better fit for their content.
Signed-off-by: James Almer <jamrial@gmail.com>
Mostly consistent formatting and consistently ordering of
warnings/notes to be next to the description.
Additionally group the AV_DICT_* macros.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This is a more explicit iteration API rather than using the "magic"
av_dict_get(d, "", t, AV_DICT_IGNORE_SUFFIX) which is not really
trivial to grasp what it does when casually reading through code.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
This can be achieved by moving the AVOnce out of the structure
containing the function pointers; the latter can then be made
const.
This also has the advantage of eliminating padding in the structure
(sizeof(AVOnce) is four here) and allowing the AVOnces to be put
into .bss (dependening upon the implementation).
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is possible to avoid the factors array for the power-of-two
tables for which said array is unused by using a different
structure for initialization for power-of-two tables than for
non-power-of-two-tables. This saves 3*15*16B from .data.
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Up until now, av_aes_init() uses a->round_key[0].u8 + t
as dst of memcpy where it is intended for t to greater
than 16 (u8 is an uint8_t[16]); given that round_key itself
is an array, it is actually intended for the dst to be
in a latter round_key member. To do this properly,
just cast a->round_key to unsigned char*.
This fixes the srtp, aes, aes_ctr, mov-3elist-encrypted,
mov-frag-encrypted and mov-tenc-only-encrypted
FATE-tests with (Clang-)UBSan.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The AES code uses av_aes_block, a union consisting of
uint64_t[2], uint32_t[4], uint8_t[4][4] and uint8_t[16].
subshift() performs byte-wise manipulations of two av_aes_blocks,
but when encrypting, it does so with a shift of two bytes;
more precisely, it uses
"av_aes_block *s1 = (av_aes_block *) (s0[0].u8 - s)"
and lateron uses the uint8_t[16] member to access s0.
Yet av_aes_block requires to be suitably aligned for
the uint64_t[2] member, which s0[0].u8 - 2 is certainly
not. This is in violation of 6.3.2.3 (7) of C11. UBSan
reports this in the aes_ctr, mov-3elist-encrypted,
mov-frag-encrypted, mov-tenc-only-encrypted and srtp
tests.
Furthermore, there is another issue here: The pointer points
outside of s0; this works, because all the accesses lateron
use an index >= 3. (Clang-)UBSan reports this as
"runtime error: index -2 out of bounds for type 'uint8_t[16]'".
This commit fixes both of these issues: The latter issue
is fixed by applying an offset of "+ 3" during the cast
and subtracting this from the indices used lateron.
The former issue is solved by not casting to av_aes_block*
at all; instead simply cast to unsigned char*.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Separate the blocks to make the grouping easier to grasp,
add the file properly to the group and fix the file description
incorrectly used as filename, fixing the following doxy warning:
warning: the name 'Colorspace' supplied as the argument in
the \file statement is not an input file
Make it a bit easier to grasp the grouping when not
unnecessarily splitting comment blocks.
Additionally do not try to add lavu_video_stereo3d to itself, resolving
the following doxy warning:
warning: Refusing to add group lavu_video_stereo3d to itself
Make it a bit easier to grasp the grouping when not
unnecessarily splitting comment blocks.
Additionally do not try to add lavu_video_spherical to itself, resolving
the following doxy warning:
warning: Refusing to add group lavu_video_spherical to itself
Make it a bit easier to grasp the grouping when not
unnecessarily splitting comment blocks.
Additionally do not try to add lavu_video_display to itself, resolving
the following doxy warning:
warning: Refusing to add group lavu_video_display to itself
VSETVLI xd, x0, ...' has rather nonobvious semantics:
- If xd is x0, then it preserves the current vector length.
- If xd is not x0, it sets the vector length to the supported maximum.
Also somewhat confusingly, while VMV.X.S always does its thing
regardless of the selected vector length, VMV.S.X does _nothing_ if the
selected vector length is zero.
So the current code breaks fails to initialise the accumulator if we
are unlucky to have a selected vector length of zero on entry. Fix it
by forcing the vector length to one.
Add an AV_PIX_FMT_NE macro for RGB32FBE/RGB32FLE and also one for
RGBA32FBE/RGBA32FLE for packed 32-bit float RGB samples, and also
packed 32-bit float RGBA samples, respectively.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: Leo Izen <leo.izen@gmail.com>
Currently it is done in several different ways, which
might cause needless dependencies or in case of
tx_float_neon.S is incorrect.
Reviewed-by: Martin Storsjö <martin@martin.st>
Signed-off-by: Reimar Döffinger <Reimar.Doeffinger@gmx.de>
GCC 4.0 not only added a visibility attribute, but also
a pragma to set it for a whole region of code.*
This commit exposes this via macros.
*: See https://gcc.gnu.org/gcc-4.0/changes.html
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
P012, Y212 and XV36 are used for 12bit content in FFmpeg VAAPI, so
these formats should be used in FFmpeg QSV too, however the SDK only
declares support for P016, Y216 and Y416. So this commit fudged mappings
between AV_PIX_FMT_P012 and MFX_FOURCC_P016, AV_PIX_FMT_Y212 and
MFX_FOURCC_Y216, AV_PIX_FMT_XV36 and MFX_FOURCC_Y416.
Signed-off-by: Fei Wang <fei.w.wang@intel.com>
Signed-off-by: Wenbin Chen <wenbin.chen@intel.com>
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
XV30 is used for 10bit 4:4:4 content in FFmpeg VAAPI, so XV30 should be
used for 10bit 4:4:4 content in FFmpeg QSV too because QSV is based on
VAAPI on Linux. However the SDK only declares support for Y410 but does
nothing with the alpha in Y410, so this commit fudged a mapping between
AV_PIX_FMT_XV30 and MFX_FOURCC_Y410.
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
We can't get Shift from bit depth for some formats in the SDK. For
example, bit depth is 10, however Shift is 0 for Y410 (XV30 in FFmpeg).
In order to support these formats in the next commits, this patch
specified Shift for each format
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
On most cases, the vector type (VTYPE) for the RISC-V Vector extension
is supplied as an immediate value, with either of the VSETVLI or
VSETIVLI instructions. There is however a third instruction VSETVL
which takes the vector type from a general purpose register. That is so
the type can be selected at run-time.
This introduces a macro to load a (valid) vector type into a register.
The syntax follows that of VSETVLI and VSETIVLI, with element size,
group multiplier, then tail and mask policies.
Unfortunately, it is common, and will remain so, that the Bit
manipulations are not enabled at compilation time. This is an official
policy for Debian ports in general (though they do not support RISC-V
officially as of yet) to stick to the minimal target baseline, which
does not include the B extension or even its Zbb subset.
For inline helpers (CPOP, REV8), compiler builtins (CTZ, CLZ) or
even plain C code (MIN, MAX, MINU, MAXU), run-time detection seems
impractical. But at least it can work for the byte-swap DSP functions.
The butterflies_fixed function pointer declaration specifies av_restrict
for the first two pointer arguments. So the corresponding function
definitions should honor this declaration.
MSVC emits warning C4113 for this.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Namely to lavu/tests/pixelutils.c. This way, this function will
not be included into actual binaries any more.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Instead use av_pix_fmt_desc_next(). It is still possible
to check its return values by comparing it with the
(currently) expected values and the code does so.
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
ff_check_pixfmt_descriptors() was added in commit
20e99a9c10. At this time,
the values of enum AVPixelFormat were not contiguous;
instead there was a jump from 111 to 291 (or from 115
to 295 depending upon AV_PIX_FMT_ABI_GIT_MASTER).
ff_check_pixfmt_descriptors() accounts for this
by skipping empty descriptors. Yet this issue no longer
exists: There are no holes.
The check for said holes makes GCC believe that the name
can be NULL; because it is used as argument corresponding to
%s in a log statement, it therefore emits a warning
(since d75c4693fe). Therefore
this commit simply removes these checks.
Also move the checks for name before the log statement.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Don't place it as doxy specific for the order field, and generalize it both to
also cover already defined orders and to not make it seem like the user is
required to handle a layout they don't fully support or understand.
Signed-off-by: James Almer <jamrial@gmail.com>
RVV defines a total of 12 different extensions, including:
- 5 different instruction subsets:
- Zve32x: 8-, 16- and 32-bit integers,
- Zve32f: Zve32x plus single precision floats,
- Zve64x: Zve32x plus 64-bit integers,
- Zve64f: Zve32f plus Zve64x,
- Zve64d: Zve64f plus double precision floats.
- 6 different vector lengths:
- Zvl32b (embedded only),
- Zvl64b (embedded only),
- Zvl128b,
- Zvl256b,
- Zvl512b,
- Zvl1024b,
- and the V extension proper: equivalent to Zve64f and Zvl128b.
In total, there are 6 different possible sets of supported instructions
(including the empty set), but for convenience we allocate one bit for
each type sets: up-to-32-bit ints (RVV_I32), floats (RVV_F32),
64-bit ints (RVV_I64) and doubles (RVV_F64).
Whence the vector size is needed, it can be retrieved by reading the
unprivileged read-only vlenb CSR. This should probably be a separate
helper macro if needed at a later point.
This introduces compile-time and run-time CPU detection on RISC-V. In
practice, I doubt that FFmpeg will ever see a RISC-V CPU without all of
I, F and D extensions, and if it does, it probably won't have run-time
detection. So the flags are essentially always set.
But as things stand, checkasm wants them that way. Compare the ARMV8
flag on AArch64. We are nowhere near running short on CPU flag bits.
They are intended as replacements for avcodec_enum_to_chroma_pos()
and avcodec_chroma_pos_to_enum().
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
To support non-aligned buffers during the post-transform step, just iterate
backwards over the array.
This allows using the 15xN-point FFT, with which the speed is 2.1 times
faster than our old libavcodec implementation.
~4x faster than the C version.
The shuffles in the 15pt dim1 are seriously expensive. Not happy with it,
but I'm contempt.
Can be easily converted to pure AVX by removing all vpermpd/vpermps
instructions.
In case new orders are added in the future, existing library users can still
use the layout simply by ignoring everything but the channel count in it, so
make this explicit.
Reviewed-by: Anton Khirnov <anton@khirnov.net>
Signed-off-by: James Almer <jamrial@gmail.com>
Up until now, using NULL as key in av_dict_get() on a non-empty
AVDictionary would crash; using NULL as key in av_dict_set()
would also crash for a non-empty AVDictionary unless AV_DICT_MULTIKEY
was set; in case the dictionary was initially empty or AV_DICT_MULTIKEY
was set, it was even possible for av_dict_set() to succeed when
adding a NULL key, namely when one uses a value != NULL and
the AV_DICT_DONT_STRDUP_VAL flag. Using av_dict_get() on such
an AVDictionary will usually lead to crashes, though.
Fix this by actually checking for key in both functions; error out
if they are NULL.
While just at it, also stop relying on av_strdup(NULL) to return NULL
in av_dict_set().
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Since introducing the various packed formats used by VAAPI (and p012),
we've noticed that there's actually a gap in how
av_find_best_pix_fmt_of_2 works. It doesn't actually assign any value
to having the same bit depth as the source format, when comparing
against formats with a higher bit depth. This usually doesn't matter,
because av_get_padded_bits_per_pixel() will account for it.
However, as many of these formats use padding internally, we find that
av_get_padded_bits_per_pixel() actually returns the same value for the
10 bit, 12 bit, 16 bit flavours, etc. In these tied situations, we end
up just picking the first of the two provided formats, even if the
second one should be preferred because it matches the actual bit depth.
This bug already existed if you tried to compare yuv420p10 against p016
and p010, for example, but it simply hadn't come up before so we never
noticed.
But now, we actually got a situation in the VAAPI VP9 decoder where it
offers both p010 and p012 because Profile 3 could be either depth and
ends up picking p012 for 10 bit content due to the ordering of the
testing.
In addition, in the process of testing the fix, I realised we have the
same gap when it comes to chroma subsampling - we do not favour a
format that has exactly the same subsampling vs one with less
subsampling when all else is equal.
To fix this, I'm introducing a small score penalty if the bit depth or
subsampling doesn't exactly match the source format. This will break
the tie in favour of the format with the exact match, but not offset
any of the other scoring penalties we already have.
I have added a set of tests around these formats which will fail
without this fix.
When appending two values (due to AV_DICT_APPEND), the earlier code
would first zero-allocate a buffer of the required size and then
copy both parts into it via av_strlcat(). This is problematic,
as it leads to quadratic performance in case of frequent enlargements.
Fix this by using av_realloc() (which is hopefully designed to handle
such cases in a better way than simply throwing the buffer we already
have away) and by copying the string via memcpy() (after all, we already
calculated the strlen of both strings).
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
If a key already exists in an AVDictionary and the AV_DICT_APPEND flag
is set, the old entry is at first discarded from the dictionary, but
a pointer to the value is kept. Lateron enough memory to store the
appended string is allocated; should this allocation fail, the old string
is not freed and hence leaks. This commit changes this by moving
creating the combined value to an earlier point in the function,
which also ensures that the AVDictionary is unchanged in case of errors.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
We know that an AVDictionary is not empty if we have just added
an entry to it, so only check for it being empty on the branch
that does not do so.
Reviewed-by: Paul B Mahol <onemda@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This reverts commit 2c8dc7e953.
The loongarch headers have been fixed, so that this wrapper
is no longer necessary.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This reverts commit 6c9a60ada4.
The loongarch headers have been fixed, so that this workaround
is no longer necessary.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
If the target supports the Basic bit-manipulation (Zbb) extension, then
the REV8 instruction is available to reverse byte order.
Note that this instruction only exists at the "XLEN" register size,
so we need to right shift the result down to the data width.
If Zbb is not supported, then this patchset does nothing. Support for
run-time detection is left for the future. Currently, there are no
bits in auxv/ELF HWCAP for Z-extensions, so there are no clean ways to
do this.
This uses the architected RISC-V 64-bit cycle counter from the
RISC-V unprivileged instruction set.
In 64-bit and 128-bit, this is a straightforward CSR read.
In 32-bit mode, the 64-bit value is exposed as two CSRs, which
cannot be read atomically, so a loop is necessary to detect and fix up
the race condition where the bottom half wraps exactly between the two
reads.
AV_PIX_FMT_VUYX is used for 8bit 4:4:4 content in FFmpeg VAAPI, so
AV_PIX_FMT_VUYX should be used for 8bit 4:4:4 content in FFmpeg QSV too
because QSV is based on VAAPI on Linux. However the SDK only declares
support for AYUV and does nothing with the alpha, so this commit fudged
a mapping between AV_PIX_FMT_VUYX and MFX_FOURCC_AYUV.
Reviewed-by: Philip Langdale <philipl@overt.org>
Signed-off-by: Haihao Xiang <haihao.xiang@intel.com>
This matches a similar cap on the number of automatic threads
in libavcodec/pthread_slice.c.
On systems with lots of cores, this fixes a couple fate failures
in 32 bit mode on such machines (where spawning a huge number of
threads runs out of address space).
Signed-off-by: Martin Storsjö <martin@martin.st>
This commit implements an iMDCT in pure assembly.
This is capable of processing any mod-8 transforms, rather than just
power of two, but since power of two is all we have assembly for
currently, that's what's supported.
It would really benefit if we could somehow use the C code to decide
which function to jump into, but exposing function labels from assebly
into C is anything but easy.
The post-transform loop could probably be improved.
This was somewhat annoying to write, as we must support arbitrary
strides during runtime. There's a fast branch for stride == 4 bytes
and a slower one which uses vgatherdps.
Zen 3 benchmarks for stride == 4 for old (av_imdct_half) vs new (av_tx):
128pt:
2811 decicycles in av_tx (imdct),16775916 runs, 1300 skips
3082 decicycles in av_imdct_half,16776751 runs, 465 skips
256pt:
4920 decicycles in av_tx (imdct),16775820 runs, 1396 skips
5378 decicycles in av_imdct_half,16776411 runs, 805 skips
512pt:
9668 decicycles in av_tx (imdct),16775774 runs, 1442 skips
10626 decicycles in av_imdct_half,16775647 runs, 1569 skips
1024pt:
19812 decicycles in av_tx (imdct),16777144 runs, 72 skips
23036 decicycles in av_imdct_half,16777167 runs, 49 skips
If we want to be able to map between VAAPI and Vulkan (to do Vulkan
filtering), we need to have matching formats on each side.
The mappings here are not exact. In the same way that P010 is still
mapped to full 16 bit formats, P012 has to be mapped that way as well.
Similarly, VUYX has to be mapped to an alpha-equipped format, and XV36
has to be mapped to a fully 16bit alpha-equipped format.
While Vulkan seems to fundamentally lack formats with an undefined,
but physically present, alpha channel, it has have 10X6 and 12X4
formats that you could imagine using for P010, P012 and XV36, but these
formats don't support the STORAGE usage flag. Today, hwcontext_vulkan
requires all formats to be storable because it wants to be able to use
them to create writable images. Until that changes, which might happen,
we have to restrict the set of formats we use.
Finally, when mapping a Vulkan image back to vaapi, I observed that
the VK_FORMAT_R16G16B16A16_UNORM format we have to use for XV36 going
to Vulkan is mapped to Y416 when going to vaapi (which makes sense as
it's the exact matching format) so I had to add an entry for it even
though we don't use it directly.
With the necessary pixel formats defined, we can now expose support for
the remaining 10/12bit combinations that VAAPI can handle.
Specifically, we are adding support for:
* HEVC
** 12bit 420
** 10bit 422
** 12bit 422
** 10bit 444
** 12bit 444
* VP9
** 10bit 444
** 12bit 444
These obviously require actual hardware support to be usable, but where
that exists, it is now enabled.
Note that unlike YUVA/YUVX, the Intel driver does not formally expose
support for the alphaless formats XV30 and XV360, and so we are
implicitly discarding the alpha from the decoder and passing undefined
values for the alpha to the encoder. If a future encoder iteration was
to actually do something with the alpha bits, we would need to use a
formal alpha capable format or the encoder would need to explicitly
accept the alphaless format.
These are the formats we want/need to use when dealing with the Intel
VAAPI decoder for 12bit 4:2:0, 12bit 4:2:2, 10bit 4:4:4 and 12bit 4:4:4
respectively.
As with the already supported Y210 and YUVX (XVUY) formats, they are
based on formats Microsoft picked as their preferred 4:2:2 and 4:4:4
video formats, and Intel ran with it.
P12 and Y212 are simply an extension of 10 bit formats to say 12 bits
will be used, with 4 unused bits instead of 6.
XV30, and XV36, as exotic as they sound, are variants of Y410 and Y412
where the alpha channel is left formally undefined. We prefer these
over the alpha versions because the hardware cannot actually do
anything with the alpha channel and respecting it is just overhead.
Y412/XV46 is a normal looking packed 4 channel format where each
channel is 16bits wide but only the 12msb are used (like P012).
Y410/XV30 packs three 10bit channels in 32bits with 2bits of alpha,
like A/X2RGB10 style formats. This annoying layout forced me to define
the BE version as a bitstream format. It seems like our pixdesc
infrastructure can handle the LE version being byte-defined, but not
when it's reversed. If there's a better way to handle this, please
let me know. Our existing X2 formats all have the 2 bits at the MSB
end, but this format places them at the LSB end and that seems to be
the root of the problem.
There are no particular reasons to force the compiler to use the same
register as output and input operand. This forces an extra MOV
instruction if the input value needs to be reused after the swap.
In most cases, this makes no differences, as the compiler will seleect
the same register for both operands either way.
Signed-off-by: Martin Storsjö <martin@martin.st>
There are no particular reasons to force the compiler to use the same
register as output and input operand. This forces an extra MOV
instruction if the input value needs to be reused after the swap.
In most cases, this makes no differences, as the compiler will seleect
the same register for both operands either way.
Signed-off-by: Martin Storsjö <martin@martin.st>
It reduces typing: Before this patch, there were 11 callbacks
that exceeded the 80 char line length limit; now there are zero.
It also allows to remove ONLY_IF_THREADS_ENABLED() in
libavutil/internal.h.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It reduces typing: Before this patch, there were 105 codecs
whose long_name-definition exceeded the 80 char line length
limit. Now there are only nine of them.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It has been deprecated in b4f59beeb4,
but the attribute_deprecated was not set and there was no entry
in APIchanges. This commit adds these and schedules it for removal.
Given that the reason behind the deprecation is exactly the same
as in av_fopen_utf8(), reuse its FF_API_AV_FOPEN_UTF8.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
They are unused since d63443b968.
Furthermore, they were always in the wrong header:
libavutil/internal.h is auto-included almost everywhere, but
FF_SYMVER would only ever be used at a few places, so a proper
header of its own would be appropriate for it.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The AArch64 assembly accesses those symbols directly, without
indirection via e.g. the GOT on ELF. In order for this not to
require text relocations, those symbols need to be resolved fully
at link time, i.e. those symbols can't be interposable.
Normally, so far, this is achieved when linking shared libraries
in two ways; we have a version script (libavcodec/libavcodec.v) which
marks all symbols that don't start with av* as local. Additionally,
we try to add -Wl,-Bsymbolic to the linker options if supported,
making sure that such symbol references are resolved fully at link
time, instead of making them interposable.
When the libavcodec static library is linked into another shared
library, there's no guarantee that it uses similar options (even though
that would be favourable), which would end up requiring text relocations
in the AArch64 assembly.
Explicitly mark the symbols that are accessed from AArch64 assembly
as hidden, so that they are resolved fully at link time even without
the version script and -Wl,-Bsymbolic.
Signed-off-by: Martin Storsjö <martin@martin.st>
As vaapi doesn't actually do anything useful with the alpha channel,
and we have an alphaless format available, let's use that instead.
The changes here are mostly 1:1 switching, but do note the explicit
change in the number of declared channels from 4 to 3 to reflect that
the alpha is being ignored.
This is the alphaless version of VUYA that I introduced recently. After
further discussion and noting that the Intel vaapi driver explicitly
lists XYUV as a support format for encoding and decoding 8bit 444
content, we decided to switch our usage and avoid the overhead of
having a declared alpha channel around.
Note that I am not removing VUYA, as this turned out to have another
use, which was to replace the need for v408enc/dec when dealing with
the format.
The vaapi switching will happen in the next change
The fastest fast Fourier transform in not just the west, but the world,
now for the most popular toy ISA.
On a high level, it follows the design of the AVX2 version closely,
with the exception that the input is slightly less permuted as we don't have
to do lane switching with the input on double 4pt and 8pt.
On a low level, the lack of subadd/addsub instructions REALLY penalizes
any attempt at writing an FFT. That single register matters a lot,
and reloading it simply takes unacceptably long.
In x86 land, vendors would've noticed developers need this.
In ARM land, you get a badly designed complex multiplication instruction
we cannot use, that's not present on 95% of devices. Because only
compilers matter, right?
Future optimization options are very few, perhaps better register
management to use more ld1/st1s.
All timings below are in cycles:
A53:
Length | C | New (lavu) | Old (lavc) | FFTW
------ |-------------|-------------|-------------|-----
4 | 842 | 420 | 1210 | 1460
8 | 1538 | 1020 | 1850 | 2520
16 | 3717 | 1900 | 3700 | 3990
32 | 9156 | 4070 | 8289 | 8860
64 | 21160 | 9931 | 18600 | 19625
128 | 49180 | 23278 | 41922 | 41922
256 | 112073 | 53876 | 93202 | 101092
512 | 252864 | 122884 | 205897 | 207868
1024 | 560512 | 278322 | 458071 | 453053
2048 | 1295402 | 775835 | 1038205 | 1020265
4096 | 3281263 | 2021221 | 2409718 | 2577554
8192 | 8577845 | 4780526 | 5673041 | 6802722
Apple M1
New - Total for len 512 reps 2097152 = 1.459141 s
Old - Total for len 512 reps 2097152 = 2.251344 s
FFTW - Total for len 512 reps 2097152 = 1.868429 s
New - Total for len 1024 reps 4194304 = 6.490080 s
Old - Total for len 1024 reps 4194304 = 9.604949 s
FFTW - Total for len 1024 reps 4194304 = 7.889281 s
New - Total for len 16384 reps 262144 = 10.374001 s
Old - Total for len 16384 reps 262144 = 15.266713 s
FFTW - Total for len 16384 reps 262144 = 12.341745 s
New - Total for len 65536 reps 8192 = 1.769812 s
Old - Total for len 65536 reps 8192 = 4.209413 s
FFTW - Total for len 65536 reps 8192 = 3.012365 s
New - Total for len 131072 reps 4096 = 1.942836 s
Old - Segfaults
FFTW - Total for len 131072 reps 4096 = 3.713713 s
Thanks to wbs for some simplifications, assembler fixes and a review
and to jannau for giving it a look.
_Float16 support was available on arm/aarch64 for a while, and with gcc
12 was enabled on x86 as long as SSE2 is supported.
If the target arch supports f16c, gcc emits fairly efficient assembly,
taking advantage of it. This is the case on x86-64-v3 or higher.
Same goes on arm, which has native float16 support.
On x86, without f16c, it emulates it in software using sse2 instructions.
This has shown to perform rather poorly:
_Float16 full SSE2 emulation:
frame=50074 fps=848 q=-0.0 size=N/A time=00:33:22.96 bitrate=N/A speed=33.9x
_Float16 f16c accelerated (Zen2, --cpu=znver2):
frame=50636 fps=1965 q=-0.0 Lsize=N/A time=00:33:45.40 bitrate=N/A speed=78.6x
classic half2float full software implementation:
frame=49926 fps=1605 q=-0.0 Lsize=N/A time=00:33:17.00 bitrate=N/A speed=64.2x
Hence an additional check was introduced, that only enables use of
_Float16 on x86 if f16c is being utilized.
On aarch64, a similar uplift in performance is seen:
RPi4 half2float full software implementation:
frame= 6088 fps=126 q=-0.0 Lsize=N/A time=00:04:03.48 bitrate=N/A speed=5.06x
RPi4 _Float16:
frame= 6103 fps=158 q=-0.0 Lsize=N/A time=00:04:04.08 bitrate=N/A speed=6.32x
Since arm/aarch64 always natively support 16 bit floats, it can always
be considered fast there.
I'm not aware of any additional platforms that currently support
_Float16. And if there are, they should be considered non-fast until
proven fast.
IEEE-754 differentiates two different kind of NaNs.
Quiet and Signaling ones. They are differentiated by the MSB of the
mantissa.
For whatever reason, actual hardware conversion of half to single always
sets the signaling bit to 1 if the mantissa is != 0, and to 0 if it's 0.
So our code has to follow suite or fate-testing hardware float16 will be
impossible.
Convert the input from a scatter to a gather instead,
which is faster and better for SIMD.
Also, add a pre-shuffled exptab version to avoid
gathering there at all. This doubles the exptab size,
but the speedup makes it worth it. In SIMD, the
exptab will likely be purged to a higher cache
anyway because of the FFT in the middle, and
the amount of loads stays identical.
For a 960-point inverse MDCT, the speedup is 10%.
This makes it possible to write sane and fast SIMD
versions of inverse MDCTs.
