As the plain neon qpel_h functions process two rows at a time,
we need to allocate storage for h+8 rows instead of h+7.
By allocating storage for h+8 rows, incrementing the stack
pointer won't end up at the right spot in the end. Store the
intended final stack pointer value in a register x14 which we
store on the stack.
AWS Graviton 3:
put_hevc_qpel_bi_hv4_8_c: 385.7
put_hevc_qpel_bi_hv4_8_neon: 131.0
put_hevc_qpel_bi_hv4_8_i8mm: 92.2
put_hevc_qpel_bi_hv6_8_c: 701.0
put_hevc_qpel_bi_hv6_8_neon: 239.5
put_hevc_qpel_bi_hv6_8_i8mm: 191.0
put_hevc_qpel_bi_hv8_8_c: 1162.0
put_hevc_qpel_bi_hv8_8_neon: 228.0
put_hevc_qpel_bi_hv8_8_i8mm: 225.2
put_hevc_qpel_bi_hv12_8_c: 2305.0
put_hevc_qpel_bi_hv12_8_neon: 558.0
put_hevc_qpel_bi_hv12_8_i8mm: 483.2
put_hevc_qpel_bi_hv16_8_c: 3965.2
put_hevc_qpel_bi_hv16_8_neon: 732.7
put_hevc_qpel_bi_hv16_8_i8mm: 656.5
put_hevc_qpel_bi_hv24_8_c: 8709.7
put_hevc_qpel_bi_hv24_8_neon: 1555.2
put_hevc_qpel_bi_hv24_8_i8mm: 1448.7
put_hevc_qpel_bi_hv32_8_c: 14818.0
put_hevc_qpel_bi_hv32_8_neon: 2763.7
put_hevc_qpel_bi_hv32_8_i8mm: 2468.0
put_hevc_qpel_bi_hv48_8_c: 32855.5
put_hevc_qpel_bi_hv48_8_neon: 6107.2
put_hevc_qpel_bi_hv48_8_i8mm: 5452.7
put_hevc_qpel_bi_hv64_8_c: 57591.5
put_hevc_qpel_bi_hv64_8_neon: 10660.2
put_hevc_qpel_bi_hv64_8_i8mm: 9580.0
Signed-off-by: Martin Storsjö <martin@martin.st>
As the plain neon qpel_h functions process two rows at a time,
we need to allocate storage for h+8 rows instead of h+7.
AWS Graviton 3:
put_hevc_qpel_uni_w_hv4_8_c: 422.2
put_hevc_qpel_uni_w_hv4_8_neon: 140.7
put_hevc_qpel_uni_w_hv4_8_i8mm: 100.7
put_hevc_qpel_uni_w_hv8_8_c: 1208.0
put_hevc_qpel_uni_w_hv8_8_neon: 268.2
put_hevc_qpel_uni_w_hv8_8_i8mm: 261.5
put_hevc_qpel_uni_w_hv16_8_c: 4297.2
put_hevc_qpel_uni_w_hv16_8_neon: 802.2
put_hevc_qpel_uni_w_hv16_8_i8mm: 731.2
put_hevc_qpel_uni_w_hv32_8_c: 15518.5
put_hevc_qpel_uni_w_hv32_8_neon: 3085.2
put_hevc_qpel_uni_w_hv32_8_i8mm: 2783.2
put_hevc_qpel_uni_w_hv64_8_c: 57254.5
put_hevc_qpel_uni_w_hv64_8_neon: 11787.5
put_hevc_qpel_uni_w_hv64_8_i8mm: 10659.0
Signed-off-by: Martin Storsjö <martin@martin.st>
As the plain neon qpel_h functions process two rows at a time,
we need to allocate storage for h+8 rows instead of h+7.
By allocating storage for h+8 rows, incrementing the stack
pointer won't end up at the right spot in the end. Store the
intended final stack pointer value in a register x14 which we
store on the stack.
AWS Graviton 3:
put_hevc_qpel_uni_hv4_8_c: 384.2
put_hevc_qpel_uni_hv4_8_neon: 127.5
put_hevc_qpel_uni_hv4_8_i8mm: 85.5
put_hevc_qpel_uni_hv6_8_c: 705.5
put_hevc_qpel_uni_hv6_8_neon: 224.5
put_hevc_qpel_uni_hv6_8_i8mm: 176.2
put_hevc_qpel_uni_hv8_8_c: 1136.5
put_hevc_qpel_uni_hv8_8_neon: 216.5
put_hevc_qpel_uni_hv8_8_i8mm: 214.0
put_hevc_qpel_uni_hv12_8_c: 2259.5
put_hevc_qpel_uni_hv12_8_neon: 498.5
put_hevc_qpel_uni_hv12_8_i8mm: 410.7
put_hevc_qpel_uni_hv16_8_c: 3824.7
put_hevc_qpel_uni_hv16_8_neon: 670.0
put_hevc_qpel_uni_hv16_8_i8mm: 603.7
put_hevc_qpel_uni_hv24_8_c: 8113.5
put_hevc_qpel_uni_hv24_8_neon: 1474.7
put_hevc_qpel_uni_hv24_8_i8mm: 1351.5
put_hevc_qpel_uni_hv32_8_c: 14744.5
put_hevc_qpel_uni_hv32_8_neon: 2599.7
put_hevc_qpel_uni_hv32_8_i8mm: 2266.0
put_hevc_qpel_uni_hv48_8_c: 32800.0
put_hevc_qpel_uni_hv48_8_neon: 5650.0
put_hevc_qpel_uni_hv48_8_i8mm: 5011.7
put_hevc_qpel_uni_hv64_8_c: 57856.2
put_hevc_qpel_uni_hv64_8_neon: 9863.5
put_hevc_qpel_uni_hv64_8_i8mm: 8767.7
Signed-off-by: Martin Storsjö <martin@martin.st>
As the plain neon qpel_h functions process two rows at a time,
we need to allocate storage for h+8 rows instead of h+7.
By allocating storage for h+8 rows, incrementing the stack
pointer won't end up at the right spot in the end. Store the
intended final stack pointer value in a register x14 which we
store on the stack.
AWS Graviton 3:
put_hevc_qpel_hv4_8_c: 386.0
put_hevc_qpel_hv4_8_neon: 125.7
put_hevc_qpel_hv4_8_i8mm: 83.2
put_hevc_qpel_hv6_8_c: 749.0
put_hevc_qpel_hv6_8_neon: 207.0
put_hevc_qpel_hv6_8_i8mm: 166.0
put_hevc_qpel_hv8_8_c: 1305.2
put_hevc_qpel_hv8_8_neon: 216.5
put_hevc_qpel_hv8_8_i8mm: 213.0
put_hevc_qpel_hv12_8_c: 2570.5
put_hevc_qpel_hv12_8_neon: 480.0
put_hevc_qpel_hv12_8_i8mm: 398.2
put_hevc_qpel_hv16_8_c: 4158.7
put_hevc_qpel_hv16_8_neon: 659.7
put_hevc_qpel_hv16_8_i8mm: 593.5
put_hevc_qpel_hv24_8_c: 8626.7
put_hevc_qpel_hv24_8_neon: 1653.5
put_hevc_qpel_hv24_8_i8mm: 1398.7
put_hevc_qpel_hv32_8_c: 14646.0
put_hevc_qpel_hv32_8_neon: 2566.2
put_hevc_qpel_hv32_8_i8mm: 2287.5
put_hevc_qpel_hv48_8_c: 31072.5
put_hevc_qpel_hv48_8_neon: 6228.5
put_hevc_qpel_hv48_8_i8mm: 5291.0
put_hevc_qpel_hv64_8_c: 53847.2
put_hevc_qpel_hv64_8_neon: 9856.7
put_hevc_qpel_hv64_8_i8mm: 8831.0
Signed-off-by: Martin Storsjö <martin@martin.st>
The hv32 and hv64 functions were identical - both loop and
process 16 pixels at a time.
The hv16 function was near identical, except for the outer loop
(and using sp instead of a separate register).
Given the size of these functions, the extra cost of the outer
loop is negligible, so use the same function for hv16 as well.
This removes over 200 lines of duplicated assembly, and over 4 KB
of binary size.
Signed-off-by: Martin Storsjö <martin@martin.st>
The first horizontal filter can use either i8mm or plain neon
versions, while the second part is a pure neon implementation.
Signed-off-by: Martin Storsjö <martin@martin.st>
The first horizontal filter can use either i8mm or plain neon
versions, while the second part is a pure neon implementation.
Signed-off-by: Martin Storsjö <martin@martin.st>
For widths of 32 pixels and more, loop first horizontally,
then vertically.
Previously, this function would process a 16 pixel wide slice
of the block, looping vertically. After processing the whole
height, it would backtrack and process the next 16 pixel wide
slice.
When doing 8tap filtering horizontally, the function must load
7 more pixels (in practice, 8) following the actual inputs, and
this was done for each slice.
By iterating first horizontally throughout each line, then
vertically, we access data in a more cache friendly order, and
we don't need to reload data unnecessarily.
Keep the original order in put_hevc_\type\()_h12_8_neon; the
only suboptimal case there is for width=24. But specializing
an optimal variant for that would require more code, which
might not be worth it.
For the h16 case, this implementation would give a slowdown,
as it now loads the first 8 pixels separately from the rest, but
for larger widths, it is a gain. Therefore, keep the h16 case
as it was (but remove the outer loop), and create a new specialized
version for horizontal looping with 16 pixels at a time.
Before: Cortex A53 A72 A73 Graviton 3
put_hevc_qpel_h16_8_neon: 710.5 667.7 692.5 211.0
put_hevc_qpel_h32_8_neon: 2791.5 2643.5 2732.0 883.5
put_hevc_qpel_h64_8_neon: 10954.0 10657.0 10874.2 3241.5
After:
put_hevc_qpel_h16_8_neon: 697.5 663.5 705.7 212.5
put_hevc_qpel_h32_8_neon: 2767.2 2684.5 2791.2 920.5
put_hevc_qpel_h64_8_neon: 10559.2 10471.5 10932.2 3051.7
Signed-off-by: Martin Storsjö <martin@martin.st>
This gets rid of a couple instructions, but the actual performance
is almost identical on Cortex A72/A73. On Cortex A53, it is a
handful of cycles faster.
Signed-off-by: Martin Storsjö <martin@martin.st>
Many of the routines within hevcdsp_epel_neon and hevcdsp_qpel_neon
store temporary buffers on the stack. When consuming it,
many of these functions use the stack pointer as incremental pointer
for reading the data (instead of storing it in another register),
which is rather unusual.
Technically, this is fine as long as the pointer remains properly
aligned.
However in the case of ff_hevc_put_hevc_qpel_uni_w_hv64_8_neon_i8mm,
after incrementing sp when reading data (within each 16 pixel
wide stripe) it would then reset the stack pointer back to a lower
value, for reading the next 16 pixel wide stripe, expecting the
data to remain untouched.
This can't be assumed; data on the stack below the stack pointer
can be clobbered (e.g. by a signal handler). Some OS ABIs
allow for a little margin that won't be touched, aka a red zone,
but not all do. The ones that do, guarantee 16 or 128 bytes, not
9 KB.
Convert this function to use a separate pointer register to
iterate through the data, retaining the stack pointer to point
at the bottom of the data we require to remain untouched.
Signed-off-by: Martin Storsjö <martin@martin.st>
MATCH_PER_STREAM_OPT iterates over all options of a given
OptionDef and tests whether they apply to the current stream;
if so, they are set to ost->apad, otherwise, the code errors
out. If no error happens, ost->apad is av_strdup'ed in order
to take ownership of this pointer.
But this means that setting it originally was premature,
as it leads to double-frees when an error happens lateron.
This can simply be reproduced with
ffmpeg -filter_complex anullsrc -apad bar -apad:n baz -f null -
This is a regression since 83ace80bfd.
Fix this by using a temporary variable instead of directly
setting ost->apad. Also only strdup the string if it actually
is != NULL.
Reviewed-by: Marth64 <marth64@proxyid.net>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
and rename it to FF_INFMT_INIT_CLEANUP. This flag is demuxer-only,
so this is the more appropriate place for it.
This does not preclude adding internal flags common to both
demuxer and muxer in the future.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The average length of the strings used here does not differ much
from the length of the longest string; therefore it makes sense
to use an array big enough for the longest string and not
a pointer to a string. This also moves this array into .rodata
(from .data.rel.ro).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Most of the data in the temporary storage ends up being
returned to the user as AVPacket.data, so it makes sense
to avoid using the AVBPrint for temporary storage altogether
(in particular in light of the fact that the blocks read here
are too big for the small-string optimization anyway) and
read the data directly into AVPacket.data. This also avoids
another memcpy() from a stack buffer to the AVBPrint in ts_image()
(that could always have been avoided with av_bprint_get_buffer()).
These changes also allow to use av_append_packet(), which
greatly simplifies the code; furthermore, one can avoid cleanup
code on error as the packet is already unreferenced generically
on error.
There are two user-visible changes from this patch:
1. Truncated packets are now marked as corrupt.
2. AVPacket.pos is set (it corresponds to the discarded header
line, 80 bytes before the position corresponding to the
actual packet data).
Furthermore, this patch also removes code that triggered
a -Wtautological-constant-out-of-range-compare warning
from Clang (namely a comparison of an unsigned and INT64_MAX
in an assert).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is an implementation detail of other input formats whether
they use raw_codec_id or not. The HLS demuxer should not rely
on this.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The code relies on their presence and would presumably crash
when retrieving in_fmt->name if an encrypted stream with a codec id
without demuxer were encountered.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Due to the bump it is now certain that all devices
that support flushing have the proper internal flag set.
(Notice that the check for LIBAVFORMAT_VERSION was wrong.)
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Besides improving readability this also ensures that
a developer who has the android content protocol enabled
and works on the other parts of the file will not
forget to add necessary inclusions just because of
(indirect) inclusions from the files included only
when said protocol is enabled.
Reviewed-by: Matthieu Bouron <matthieu.bouron@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
(The discrepancy between the definition and the declaration
in protocols.c is actually UB.)
Reviewed-by: Matthieu Bouron <matthieu.bouron@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
An alternative would be to limit all time/duration fields to below 64bit
Fixes: signed integer overflow: -93000000 - 9223372036839000000 cannot be represented in type 'long long'
Fixes: 64546/clusterfuzz-testcase-minimized-ffmpeg_dem_CONCAT_fuzzer-5110813828186112
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 178459578 + 2009763270 cannot be represented in type 'int'
Fixes: 62285/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_OSQ_fuzzer-5013423686287360
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Small cleanup for style, indent, switch case lables.
BTW, the preferred way to ease multiple indentation levels in a
switch statement is to align the switch and its subordinate
case labels in the same column
Signed-off-by: Jun Zhao <barryjzhao@tencent.com>
Fixes: signed integer overflow: 64 + 9223372036854775803 cannot be represented in type 'long long'
Fixes: 51896/clusterfuzz-testcase-minimized-ffmpeg_dem_CAF_fuzzer-6536881135550464
Fixes: 62276/clusterfuzz-testcase-minimized-ffmpeg_dem_CAF_fuzzer-6536881135550464
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: signed integer overflow: 9223372036854775796 + 12 cannot be represented in type 'long long'
Fixes: 51896/clusterfuzz-testcase-minimized-ffmpeg_dem_IFF_fuzzer-4898373660704768
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: 51896/clusterfuzz-testcase-minimized-ffmpeg_dem_DXA_fuzzer-5730576523198464
Fixes: signed integer overflow: 2147483566 + 82 cannot be represented in type 'int'
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Martin Storsjö
Andreas Rheinhardt
James Almer
Michael Niedermayer
Jun Zhao