yuyv422
126354 UNITS in yuv2packedX, 16384 runs, 0 skips
16383 UNITS in yuv2packedX, 16382 runs, 2 skips
yvyu422
117669 UNITS in yuv2packedX, 16384 runs, 0 skips
16271 UNITS in yuv2packedX, 16379 runs, 5 skips
uyvy422
117310 UNITS in yuv2packedX, 16384 runs, 0 skips
16226 UNITS in yuv2packedX, 16382 runs, 2 skips
yuyv422
19339 UNITS in yuv2packed2, 16384 runs, 0 skips
3718 UNITS in yuv2packed2, 16383 runs, 1 skips
yvyu422
19438 UNITS in yuv2packed2, 16384 runs, 0 skips
3800 UNITS in yuv2packed2, 16380 runs, 4 skips
uyvy422
19128 UNITS in yuv2packed2, 16384 runs, 0 skips
3721 UNITS in yuv2packed2, 16380 runs, 4 skips
yuyv422
14513 UNITS in yuv2packed1, 32768 runs, 0 skips
949 UNITS in yuv2packed1, 32767 runs, 1 skips
yvyu422
14516 UNITS in yuv2packed1, 32767 runs, 1 skips
943 UNITS in yuv2packed1, 32767 runs, 1 skips
uyvy422
14530 UNITS in yuv2packed1, 32767 runs, 1 skips
941 UNITS in yuv2packed1, 32766 runs, 2 skips
Similarly to the previous changes, we don't need to synchronise
after a memcpy to device memory. On the other hand, we need to
keep synchronising after a copy to host memory, otherwise there's
no guarantee that subsequent host reads will return valid data.
I put this call in by habit, rather than because there was any
actual need. The filter is simply processing frames one after
the other and has no need to synchronise.
malakudi on the devtalk forums noticed a slowdown when using nvenc
with temporal/spatial aq and that the slowdown went away if the
sync call was removed. I also verified that in the basic encoding
case there's an observable speedup.
I also verified that we aren't doing unnecessary sync calls in any
other filter.
The following are the newly added options:
arnr_max_frames, arnr_strength, aq_mode, denoise_noise_level, denoise_block_size,
rc_undershoot_pct, rc_overshoot_pct, minsection_pct, maxsection_pct, frame_parallel,
enable_cdef, enable_global_motion, and intrabc.
Also added macros for compiling for aom 1.0.0 and fixed the default values.
Zhong Li [Mon, 25 Mar 2019 22:10:21 +0000 (06:10 +0800)]
lavc/qsvenc: fix hevc vps extradata issues
cbs trace qsv vps header failed due to some reasons:
1. vps_temporal_id_nesting_flag is not set but spec required it must to
be 1 when vps_max_sub_layers_minus1 is equal to 0.
2. vps_num_hrd_parameters is not set and written.
3. other issues in ff_hevc_encode_nal_vps() (fixed in pervious commit_id: 520226b683).
Thierry Foucu [Wed, 27 Mar 2019 21:11:59 +0000 (14:11 -0700)]
fftools/ffmpeg: Check if we do have also a filter_complex filter.
Right now, the code check for no filter description, but if we use a
filter_complex, the code will use the AVFrame.duration which could be
wrong in case of using fps filter.
How to reproduce the problem:
ffmpeg -f lavfi -i testsrc=duration=1 -vf fps=fps=50 -vsync 1 -f null -
output 50 frames
Dong, Jerry [Thu, 28 Mar 2019 11:40:53 +0000 (19:40 +0800)]
swscale/swscale_unscaled: fixed the issue that when width/height is not 2-multiple, transition of nv12 to u/v planes is not completed.
Signed-off-by: Dong, Jerry <jerry.dong@intel.com> Signed-off-by: Decai Lin <decai.lin@intel.com> Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Zhong Li [Wed, 27 Mar 2019 11:39:41 +0000 (19:39 +0800)]
lavc/qsvenc_h264: remove the privite option trellis
There are many problems of current qsv trellis option:
1. Duplicated with AVCodecContext definition
2. MFX_TRELLIS_XXX is introduced by MSDK API 1.17
Currently Without MSDK API checking thus may cause compilation issue.
3. user is inclined to enable trellis when set "-trellis 1", but
actually it is to disable since MFX_TRELLIS_OFF is equal to 1.
4. It is too complex for user to enable trellis for every frame(I/P/B).
Just simply remove the private option, and switch to the AVCodecContext
definition. Compatibility should not a big problem (except can't exact map)
since the option name is same as AVCodecContext.
Signed-off-by: Zhong Li <zhong.li@intel.com> Reviewed-by: Carl Eugen Hoyos <ceffmpeg@gmail.com> Reviewed-by: Moritz Barsnick <barsnick@gmx.net>
avcodec/hevcdec: Avoid only partly skiping duplicate first slices
Fixes: NULL pointer dereference and out of array access Fixes: 13871/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HEVC_fuzzer-5746167087890432 Fixes: 13845/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_HEVC_fuzzer-5650370728034304
This also fixes the return code for explode mode
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>
The following output formats get about 4.5x speedup:
rgb24
39980 UNITS in yuv2packed1, 32768 runs, 0 skips
8774 UNITS in yuv2packed1, 32768 runs, 0 skips
bgr24
40069 UNITS in yuv2packed1, 32768 runs, 0 skips
8772 UNITS in yuv2packed1, 32766 runs, 2 skips
rgba
39759 UNITS in yuv2packed1, 32768 runs, 0 skips
8681 UNITS in yuv2packed1, 32767 runs, 1 skips
bgra
39729 UNITS in yuv2packed1, 32768 runs, 0 skips
8696 UNITS in yuv2packed1, 32766 runs, 2 skips
argb
39766 UNITS in yuv2packed1, 32768 runs, 0 skips
8672 UNITS in yuv2packed1, 32766 runs, 2 skips
bgra
39784 UNITS in yuv2packed1, 32768 runs, 0 skips
8659 UNITS in yuv2packed1, 32767 runs, 1 skips
The lensfun filter wraps the lensfun library which performs
transformations on videos to correct for lens distortion. Often this
results in areas in the input being mapped to areas that fall outside
the boundaries of the output. The library has a parameter called scale
which is a scale factor applied to the output video. By decreasing it it
is possible to regain the areas of the video which would otherwise have
been lost. There is a special value of 0 which indicates that the
library should automatically determine a scale factor that results in
the output frame being filled (i.e. little or no black/unmapped areas).
This patch adds a corresponding scale option to the lensfun filter which
is passed through to the library. The existing behaviour of using the
automatic value of 0 is retained as the default behaviour, while other
values will be passed through to the library.
Signed-off-by: Daniel Playfair Cal <daniel.playfair.cal@gmail.com>
Fixes: Timeout (longer than i had patience for -> 2sec) Fixes: 13205/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_PROSUMER_fuzzer-5105644481282048 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
avcodec/truemotion2: Fix integer overflow in tm2_null_res_block()
Fixes: signed integer overflow: 1111638592 - -2122219136 cannot be represented in type 'int' Fixes: 13441/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_TRUEMOTION2_fuzzer-5732769815068672 Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Ruiling Song [Fri, 23 Nov 2018 05:39:12 +0000 (13:39 +0800)]
lavu/opencl: replace va_ext.h with standard name
Khronos OpenCL header (https://github.com/KhronosGroup/OpenCL-Headers)
uses cl_va_api_media_sharing_intel.h. And Intel's official OpenCL driver
for Intel GPU (https://github.com/intel/compute-runtime) was compiled
against Khronos OpenCL header. So it's better to align with Khronos.
Signed-off-by: Ruiling Song <ruiling.song@intel.com>
Lou Logan [Thu, 21 Mar 2019 19:26:11 +0000 (11:26 -0800)]
doc/mailing-list-faq: ffmpeg-devel is now subscription only
Nobody is going to check the queue anymore, so users must now subscribe to
send messages to ffmpeg-devel. This will prevent orphaned/ignored messages
from rotting in the abandoned queue. This matches the behavior of ffmpeg-user
and libav-user.
Marton Balint [Fri, 1 Mar 2019 22:48:04 +0000 (23:48 +0100)]
avfilter/af_astats: rework sample loops
The channel loop is now the outer loop for both planar and interleaved. This is
needed by the next patch, and the speed difference is negligable if any.
Marton Balint [Fri, 1 Mar 2019 00:32:00 +0000 (01:32 +0100)]
avfilter/af_astats: add support for selecting measured statistics
set_metadata with many entries is not very efficient, and with small audio
frames the performance loss is noticable. Also with this very simple
calculations (like peak) can be even further optimized.
Unfoturnately there are some small differences in metadata and av_log info
output, so factorizing calculations and output might not worth the hassle.
Optimization found while looking at 13442/clusterfuzz-testcase-minimized-ffmpeg_AV_CODEC_ID_SCPR_fuzzer-5758293933293568
Found-by: continuous fuzzing process https://github.com/google/oss-fuzz/tree/master/projects/ffmpeg Signed-off-by: Michael Niedermayer <michael@niedermayer.cc>
Carl Eugen Hoyos [Thu, 14 Mar 2019 23:52:50 +0000 (00:52 +0100)]
lavf/sdp: Change pointer to configuration from char* to uint8_t*.
This is also what av_base64_encode() expects.
Fixes the following warnings with clang:
libavformat/sdp.c:394:40: warning: implicit conversion from 'int' to 'char' changes value from 254 to -2
libavformat/sdp.c:395:40: warning: implicit conversion from 'int' to 'char' changes value from 205 to -51
libavformat/sdp.c:396:40: warning: implicit conversion from 'int' to 'char' changes value from 186 to -70