On Sat, Jul 25, 2015 at 07:00:00PM +0300, Ivan Uskov wrote: > Hello Michael, > > Saturday, July 25, 2015, 6:14:20 PM, you wrote: > > >> I can implement necessary functions to generate annex-b prefixes just into > >> libavcodec/qsvdec_h264.c if Michael will agree this way. > > MN> would this be faster ? > MN> avoid a copy/malloc ? > If it implemented inside libavcodec/qsvdec_h264.c then it will use > about same code like current bsf implementation uses but with using of > private buffer for sps/pps instead public extradata buf using. > > Here can be the trick to write SPS/PPS directly to > QSVContext::input_fifo which implemented recently into qsvdec. > It will not solve memory copying necessity but it will solve possible twice > memory > copying issue when packet re-allocation perform into mp4toannexb filter then > data copies again into input_fifo since qsv decoder did not consume > previous packet in full. What do you think?
does this affect a large memcpy() on a common path ? if it avoids just a rare or small case then its not worth it [...] -- Michael GnuPG fingerprint: 9FF2128B147EF6730BADF133611EC787040B0FAB Awnsering whenever a program halts or runs forever is On a turing machine, in general impossible (turings halting problem). On any real computer, always possible as a real computer has a finite number of states N, and will either halt in less than N cycles or never halt.
signature.asc
Description: Digital signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
