Forgotten in eefec06634.
Reviewed-by: Ronald S. Bultje <rsbultje@gmail.com>
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The AVX and SSE2 functions are identical except for the VEX encodings
used since e9abef437f and
8b8492452d.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It's used by other parts of the module that will fail to build otherwise after
the aforementioned removal.
Signed-off-by: James Almer <jamrial@gmail.com>
wingdi.h defines its own PASSTHROUGH and it is included implicitly
by the VC-1 parser (which is mpegvideo-based and therefore includes
a lot of stuff).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
AVCodecParser has several fields which are not really meant
to be accessed by users, but it has no public-private
demarkation line, so these fields are technically public
and can therefore not simply be made private like
20f9727018 did for AVCodec.*
This commit therefore deprecates these fields and
schedules them to become private. All parsers have already
been switched to FFCodecParser, which (for now) is a union
of AVCodecParser and an unnamed clone of AVCodecParser
(new fields can be added at the end of this clone).
*: This is also the reason why split has never been removed despite
not being set for several years now.
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The current code relies on AV_CODEC_ID_NONE being zero, so that
unused codec ids are set to their proper value. This commit adds
a macro to set unset ids to AV_CODEC_ID_NONE.
(The actual rationale for this macro is to simplify
the transition to making the private fields that are
currently public in avcodec.h really private.)
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It only contains declarations for some auxiliary functions for parsing
that parsers that only work with complete packets don't need.
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The decode API can handle outputting delayed frames without relying on the
parser splitting off the ENDOFSEQ marker.
Signed-off-by: James Almer <jamrial@gmail.com>
These functions are currently always called with height either
being equal to the block size or block size+1. height is
a compile-time constant at every callsite. This makes it possible
to split this function into two to avoid the check inside
the function for whether height is odd or even.
The corresponding avg function is only used with height == block size,
so that it does not have a height parameter at all. Removing the
parameter from the put_l2 functions as well therefore simplifies
the C code.
The new functions increase the size of .text from qpel{dsp}.o
by 32B here, yet they save 464B of C code here.
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
SSSE3 is already quite old (introduced 2006 for Intel, 2011 for AMD),
so that the overwhelming majority of our users (particularly those
that actually update their FFmpeg) will be using the SSSE3 versions.
This commit therefore removes the MMX(EXT) functions overridden
by them (which don't abide by the ABI) to get closer to a removal
of emms_c.
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
The only difference between it and the H.264/VC-1 versions is the
bias constant which depends on the shift parameters for RV40.
This value ends up in a register and therefore one can reuse
the H.264 code by setting the registers for RV40 and then jumping
into the relevant H.264 function, making the four new functions
cheap (just 256 bytes in total). This approach uses one jump more
for the no-filter case and one jump less in the one-dimensional
case than an approach using separate functions.
avg_chroma_mc4_c: 167.5 ( 1.00x)
avg_chroma_mc4_mmxext: 48.1 ( 3.48x)
avg_chroma_mc4_ssse3: 31.1 ( 5.39x)
avg_chroma_mc8_c: 325.5 ( 1.00x)
avg_chroma_mc8_mmxext: 103.2 ( 3.15x)
avg_chroma_mc8_ssse3: 33.5 ( 9.71x)
put_chroma_mc4_c: 137.4 ( 1.00x)
put_chroma_mc4_mmx: 44.5 ( 3.09x)
put_chroma_mc4_ssse3: 28.4 ( 4.83x)
put_chroma_mc8_c: 271.4 ( 1.00x)
put_chroma_mc8_mmx: 99.9 ( 2.72x)
put_chroma_mc8_ssse3: 30.6 ( 8.86x)
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
It is only used by mpegvideo decoders (for lowres). It is also only used
for bitdepth == 8, so don't build the bitdepth == 16 function at all any
more.
Reviewed-by: Lynne <dev@lynne.ee>
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
This is intending to fix the case described in https://lists.ffmpeg.org/archives/list/ffmpeg-devel@ffmpeg.org/thread/AAZ7GJPPUJI5SCVTDGJ6QL7UUEP56WOM/
Where FCBParam optim is used uninitialized
a min_err of 1<<30, allows the struct to be never initilialized as all
err (which is int32_t) can be larger than min_err. By increasing min_err
above the int32_t range this is no longer possible
Untested, as i do not have the testcase
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>i
Fixes: use of uninitialized memory
Fixes: 427814450/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MAGICYUV_DEC_fuzzer-646512196065689
Fixes: 445961558/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_UTVIDEO_DEC_fuzzer-5515158672965632
the multi vlc code will otherwise return uninitialized data. Now one can argue that this data should
not be used, but on errors this data can remain ...
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This would happen if one of the extended transfer characteristics is in
use (currently only AVCOL_TRC_V_LOG).
Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
Panasonic cameras record ProRes RAW in V-Log/V-Gamut internally.
Atomos is a brand of recorders which can record uncompressed
non-debayered RAW over HDMI.
All known cameras output linear RAW over HDMI, so mark the transfer function
as linear in that case.
If avctx->ch_layout is unset (as it's allowed and even expeced by the
AV_CODEC_CAP_CHANNEL_CONF flag), the code setting substream masks will fail for
stereo and mono layouts unless a downmix channel was requested.
Fix this by deriving the mask with coded values only.
Fixes issue #20764.
Signed-off-by: James Almer <jamrial@gmail.com>
We have no use for 14-bit pixel formats for now, so remove support for gray14,
which was broken due to the LSB padding issue.
Similarly YUVA at 10/12 bit was broken for the same reason.
Add a shader-based Apple ProRes decoder.
It supports all codec features for profiles up to
the 4444 XQ profile, ie.:
- 4:2:2 and 4:4:4 chroma subsampling
- 10- and 12-bit component depth
- Interlacing
- Alpha
The implementation consists in two shaders: the
VLD kernel does entropy decoding for color/alpha,
and the IDCT kernel performs the inverse transform
on color components.
Benchmarks for a 4k yuv422p10 sample:
- AMD Radeon 6700XT: 178 fps
- Intel i7 Tiger Lake: 37 fps
- NVidia Orin Nano: 70 fps
