Added linux support for amf encoder through vulkan.
To use h.264(AMD VCE) encoder on linux amdgru-pro version 19.20+ and
amf-amdgpu-pro package(amdgru-pro contains, but does not install
automatically) are required.
This driver can be installed using amdgpu-pro-install script in
official amd driver archive.
Initialization of amf encoder occurs in this order:
1) trying to initialize through dx11(only windows)
2) trying to initialize through dx9(only windows)
3) trying to initialize through vulkan
Only Vulkan initialization available on linux.
ff_reget_buffer() will attempt to create a writable copy of the frame,
which is not needed when the decoder intends to return a reference to
the same buffer as the previous frame.
Should reduce data copy, hopefully achieving a similar speed up as
a9dacdeea6 without dropping frames.
Reviewed-by: Michael Niedermayer <michael@niedermayer.cc>
Signed-off-by: James Almer <jamrial@gmail.com>
This reverts commit a9dacdeea6.
This patch effectively made the decoder output vfr content out of samples
where cfr is expected.
Addresses ticket #7880.
Signed-off-by: James Almer <jamrial@gmail.com>
wavelets allow significant size expansion and they are also not
very fast.
Fixes: Timeout
Fixes: 16480/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5659892689403904 (108sec -> 17ms)
Fixes: 16480/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_DIRAC_fuzzer-5749422717140992 (big ->15sec)
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: Tomas Härdin <tjoppen@acc.umu.se>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: shift exponent 32 is too large for 32-bit type 'unsigned int'
Fixes: 15764/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ALAC_fuzzer-5102101203517440
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Reviewed-by: James Almer <jamrial@gmail.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Add the support of dehaze filter in existing derain filter source
code. As the processing procedure in FFmpeg is the same for current
derain and dehaze, we reuse the derain filter source code. The
model training and generation scripts are in repo
https://github.com/XueweiMeng/derain_filter.git
Reviewed-by: Steven Liu <lq@onvideo.cn>
Signed-off-by: Xuewei Meng <xwmeng96@gmail.com>
vtctx->cached_hw_frames_ctx is unref'd in videotoolbox_uninit,
but videotoolbox_hevc used ff_videotoolbox_uninit which
doesn't unref cache_hw_frames_ctx.
Signed-off-by: Pavel Koshevoy <pkoshevoy@gmail.com>
See https://trac.ffmpeg.org/ticket/7687
If an encoded frame is returned split over two or more
IL buffers due to the size, then there is a race between
whether get_buffer will fail, return NULL, and a truncated
frame is passed on, or IL will return the remaining part
of the encoded frame.
If get_buffer returns NULL, part of the frame is left behind
in the codec, and will be collected on the next call. That
then leaves a frame stuck in the codec. Repeat enough times
and the codec FIFO is full, and the pipeline stalls.
A performance improvement in the Raspberry Pi firmware means
that the timing has changed, and now frequently drops into the
case where get_buffer returns NULL.
Add code such that should a buffer be received without
OMX_BUFFERFLAG_ENDOFFRAME that get_buffer is called with wait
set, so we wait for the remainder of the frame.
This code has been made conditional on the Pi build in case
other IL implementations don't handle ENDOFFRAME correctly.
Signed-off-by: Dave Stevenson <dave.stevenson@raspberrypi.org>
Signed-off-by: Aman Gupta <aman@tmm1.net>
Signed-off-by: Martin Storsjö <martin@martin.st>
The packet counting based approach caused excessive sdt/pat/pmt for VBR, so
let's use a timestamp based approach instead similar to how we emit PCRs.
SDT/PAT/PMT period should be consistent for both VBR and CBR from now on.
Also change the type of sdt_period and pat_period to AV_OPT_TYPE_DURATION so no
floating point math is necessary.
Fixes ticket #3714.
Signed-off-by: Marton Balint <cus@passwd.hu>
Fixes: signed integer overflow: 2088795537 + 2147254401 cannot be represented in type 'int'
Fixes: signed integer overflow: -1500363496 + -1295351808 cannot be represented in type 'int'
Fixes: signed integer overflow: -79560 * 32640 cannot be represented in type 'int'
Fixes: signed integer overflow: 2088910005 + 2088796058 cannot be represented in type 'int'
Fixes: signed integer overflow: -117258064 - 2088725225 cannot be represented in type 'int'
Fixes: signed integer overflow: 2088725225 - -117258064 cannot be represented in type 'int'
Fixes: 15739/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_ALAC_fuzzer-5630664122040320
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
If all tables are skipped it would be impossible to encode any
"non black" video.
Fixes: Timeout (78sec -> 1ms)
Fixes: 15821/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_SMACKER_fuzzer-5652598838788096
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
This is significantly faster
Fixes: Timeout (1sec after this and the previous commit)
Fixes: 15558/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PPM_fuzzer-5705273643106304
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
None of the keys we support is that long and other keys
lead to decoder failure. None of the values is expected
to be longer, they are all numbers or short keywords.
This simplifies the code
Fixes: Timeout (9sec->43ms)
Fixes: 15177/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PAM_fuzzer-5080556716425216
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: 3e+47 is outside the range of representable values of type 'int'
Fixes: 16057/clusterfuzz-testcase-minimized-ffmpeg_DEMUXER_fuzzer-5691111307214848
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Clipping was added in 2009 to avoid crashes.
The clipped case would produce a 2nd slice with 0 input
thus also producing 0 output.
Subsequent checks will cause decoder failure unless both
slices have the same output length. thus the only way this
would not already fail is if the output from both slices
was 0 bytes.
Fixes: Timeout (134sec -> 241ms)
Fixes: 15599/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_MSZH_fuzzer-5658127116009472
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
The dimensions are always 320x200 they are hardcoded in the demuxer.
Hardcode them instead in the decoder.
Fixes: Timeout (16sec -> 400ms)
Fixes: 15574/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_RL2_fuzzer-5158614072819712
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Reviewed-by: Carl Eugen Hoyos <ceffmpeg@gmail.com>
Signed-off-by: leozhang <leozhang@qiyi.com>
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Current default is 256kbps, which produces inconsistent
results (too high for low-res, too low for hi-res).
Use CRF instead, which will adapt.
Signed-off-by: James Zern <jzern@google.com>
Fix#6591
The content has no rbsp_stop_one_bit for ending the SPS, that
causes the decoding SPS failure, results decoding frame failure as well.
The patch is just adding a retry with complete NALU, copied from the retry in decode_nal_unit()
Signed-off-by: James Almer <jamrial@gmail.com>
