On 4/30/2018 8:26 PM, Mark Thompson wrote: > --- > Main change since last time is including the array subscripts. Constants are > also cleaned up a bit, but stay in the cbs header (vp9.h could probably be > taken over for this purpose, but currently it's an unnamespaced header used > by the decoder so I haven't touched it). > > > configure | 2 + > doc/bitstream_filters.texi | 2 +- > libavcodec/Makefile | 1 + > libavcodec/cbs.c | 6 + > libavcodec/cbs.h | 1 + > libavcodec/cbs_internal.h | 1 + > libavcodec/cbs_vp9.c | 679 > +++++++++++++++++++++++++++++++++++ > libavcodec/cbs_vp9.h | 201 +++++++++++ > libavcodec/cbs_vp9_syntax_template.c | 390 ++++++++++++++++++++ > 9 files changed, 1282 insertions(+), 1 deletion(-) > create mode 100644 libavcodec/cbs_vp9.c > create mode 100644 libavcodec/cbs_vp9.h > create mode 100644 libavcodec/cbs_vp9_syntax_template.c
LGTM. No apparent data copy on any of the read methods which is very promising for the AV1 implementation. Only CodedBitstreamType->write_unit() still seems a tad sub-optimal with the temp write_buffer, especially in this module where unlike mpeg2/h2645 you memcpy the data twice. Can't really comment on the actual bitstream parsing code, but trace_headers_bsf doesn't complain about any of the samples i tried. _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
