On 2014-04-20 21:38:35 +0200, Anton Khirnov wrote: > Initial implementation by Andrew D'Addesio <[email protected]> during > GSoC 2012. > > Completion by Anton Khirnov <[email protected]>, sponsored by the > Mozilla Corporation. > --- > Changelog | 1 + > libavcodec/Makefile | 4 + > libavcodec/allcodecs.c | 2 + > libavcodec/opus.c | 428 +++++++++ > libavcodec/opus.h | 424 +++++++++ > libavcodec/opus_celt.c | 2185 > ++++++++++++++++++++++++++++++++++++++++++++++ > libavcodec/opus_imdct.c | 264 ++++++ > libavcodec/opus_parser.c | 74 ++ > libavcodec/opus_silk.c | 1592 +++++++++++++++++++++++++++++++++ > libavcodec/opusdec.c | 673 ++++++++++++++ > libavcodec/version.h | 2 +- > tests/Makefile | 1 + > tests/fate/opus.mak | 21 + > 13 files changed, 5670 insertions(+), 1 deletion(-) > create mode 100644 libavcodec/opus.c > create mode 100644 libavcodec/opus.h > create mode 100644 libavcodec/opus_celt.c > create mode 100644 libavcodec/opus_imdct.c > create mode 100644 libavcodec/opus_parser.c > create mode 100644 libavcodec/opus_silk.c > create mode 100644 libavcodec/opusdec.c > create mode 100644 tests/fate/opus.mak
care to upload the reference files? Testing opus in ogg makes probably more sense than writing a toy demuxer for the test vector format. First impression is good although there is a lack of obvious optimisation targets (ignoring imdct and resampling which wee not that costly in the two examples I tried). I have a patch for the two TODOs in celt_interleave_hadamard resulting in 5% faster celt decoding http://26-26-54.hardwarebug.org/0 Janne _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
