From Jun Zhao <mypopydev@gmail.com>:
> Should we relocate this to the decoder? Other codecs typically set this
> parameter in the decoder.
Signed-off-by: Wu Jianhua <toqsxw@outlook.com>
memset tables in the main thread can become a bottleneck for the decoder.
For example, if it takes 1% of the processing time for one core, the maximum achievable FPS will be 100.
Move the memeset to worker threads will fix the issue.
For luma, qp can only change at the CU level, so the qp tab size is related to the CU.
For chroma, considering the joint CbCr, the QP tab size is related to the TU.
The parser stage is not parallelizable.
We need to schedule it as soon as possible to create later stages, which are more parallelizable
clips | before | after | delta
--------------------------------------------|--------|-------|------
RitualDance_1920x1080_60_10_420_37_RA.266 | 342.7 | 365.3 | 6.59%
NovosobornayaSquare_1920x1080.bin | 321.7 | 400 | 24.34%
Tango2_3840x2160_60_10_420_27_LD.266 | 82.3 | 91.7 | 11.42%
RitualDance_1920x1080_60_10_420_32_LD.266 | 323.7 | 319.3 | -1.36%
Chimera_8bit_1080P_1000_frames.vvc | 364 | 411.3 | 12.99%
BQTerrace_1920x1080_60_10_420_22_RA.vvc | 162.7 | 185.7 | 14.14%
Decoder-Side Motion Vector Refinement is about 4~8% CPU usage for some clips
here is the test result for one time
clips | before| after | delta
------------------------------------------|-------|-------|------
RitualDance_1920x1080_60_10_420_37_RA.266 | 338.7 | 354.3 |4.61%
NovosobornayaSquare_1920x1080.bin | 320.3 | 329.3 |2.81%
Tango2_3840x2160_60_10_420_27_LD.266 | 83.3 | 83.7 |0.48%
RitualDance_1920x1080_60_10_420_32_LD.266 | 320.7 | 327.3 |2.06%
Chimera_8bit_1080P_1000_frames.vvc | 360.7 | 381.0 |5.63%
BQTerrace_1920x1080_60_10_420_22_RA.vvc | 161.7 | 163.0 |0.80%
Hardware frames with RGB colorspace will not have a YCbCrMatrixKey.
Currently, it will spam the console with warning if rgb frame is
uploaded.
Signed-off-by: Gnattu OC <gnattuoc@me.com>
Reviewed-by: Marvin Scholz <epirat07@gmail.com>
Signed-off-by: Zhao Zhili <zhilizhao@tencent.com>
x265 added support for alpha starting with build 210.
While doing so, x265_encoder_encode() changed its fifth arg to
an array of pointers to x265_picture. This broke building lavc/libx265.c
This patch simply unbreaks the build and maintains existing single-layer
non-alpha encoding support.
Fixes#11130
Fixes "libavcodec/snowenc.c:718:27: runtime error: left shift of 8509032 by 8 places cannot be represented in type 'int'"
as seen in fate-vsynth2-snow-hpel under ubsan.
Signed-off-by: James Almer <jamrial@gmail.com>
Fixes: two use-of-uninitialized-value
Fixes: 70856/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_VC1IMAGE_fuzzer-5539349918187520
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: use of uninitialized value
Fixes: 70872/clusterfuzz-testcase-minimized-ffmpeg_dem_OBU_fuzzer-6005782487826432
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Fixes: use of uninitialized value
Fixes: 70871/clusterfuzz-testcase-minimized-ffmpeg_dem_SEGAFILM_fuzzer-5883617752973312
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Related: CID1473562 Unchecked return value
Related: CID1473592 Unchecked return value
Sponsored-by: Sovereign Tech Fund
Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Since ffmpeg 6.1 video stats are accidentally hidden from streamcopy progress output.
This patch re-enables video stats (like frames=) in the progress output.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
Encoders may emit a buffering period SEI without a corresponding
SPS/PPS if the SPS/PPS is carried out-of-band, eg with avcc.
During Annex B conversion, this may result in the SPS/PPS being
inserted *after* the buffering period SEI but before the IDR NAL.
Since the buffering period SEI references the SPS, the SPS/PPS
needs to come first.
Signed-off-by: Anton Khirnov <anton@khirnov.net>
binutils 2.43 has stricter validation for labels[1] and results in errors
when building ffmpeg for armv5:
src/libavcodec/arm/mlpdsp_armv5te.S:232: Error: junk at end of line, first unrecognized character is `0'
Remove the leading zero in the "01" label to resolve this error.
[1] https://sourceware.org/git/?p=binutils-gdb.git;a=commit;h=226749d5a6ff0d5c607d6428d6c81e1e7e7a994b
Signed-off-by: Ross Burton <ross.burton@arm.com>
Signed-off-by: Martin Storsjö <martin@martin.st>
These consts are only used in the switch(planar) case located in
avisynth_create_stream_video and nowhere else in the demuxer,
so move them into that function directly.
Signed-off-by: Stephen Hutchinson <qyot27@gmail.com>
As part of this, the mutexes are no longer necessary, and
avisynth_read_close needs to check that avs->avs_library.library
still exists before it attempts to call avisynth_context_destroy
and dlclose.
Signed-off-by: Stephen Hutchinson <qyot27@gmail.com>
The atexit() handler in the avisynth demuxer was added because
there was a conflict in AvxSynth that arose due to their use
of C++ global objects, particularly in relation to having
added a logging function relying on log4cpp.
This conflict was responsible for causing a segfault on exit.
It did not affect Windows with the (at the time) upstream
AviSynth 2.5 and 2.6, nor does it affect AviSynth+.
Unfortunately, none of this was actually shielded by ifdefs
indicating the fact it was only needed for AvxSynth, so four
years ago when AviSynth+ replaced AvxSynth as the handler
for AviSynth scripts on Unix-like OSes, the fact that the
atexit handler was no longer necessary was overlooked.
Signed-off-by: Stephen Hutchinson <qyot27@gmail.com>