Le quartidi 24 floréal, an CCXXIV, Nicolas George a écrit : > Le quartidi 24 floréal, an CCXXIV, Jan Sebechlebsky a écrit : > > I can change it to array, the advantage of using linked list was that number > > of bitstream > > filters used is not known before arguments are parsed, this way the filters > > were initialized and as the the filters were parsed. With array the argument > > will be parsed twice - first pass just to count and allocate the array, > > second to initialize the filters. But it's not big deal to do it that way. > > There are several instances of similar code, the array is often reallocated. > We have macros to reallocate an array with double size efficiently. But a > quick pre-parsing to count the number of filters is also a legitimate idea. > > > Should I wait with this patch until your changes are pushed? > > You can apply it on a branch of your working tree to continue your task. > > > Can you explain this little bit more? > > Recursive filtering is annoying to debug (lots of nested stack frames, hard > to put break points) and does not play well with threads. > > > Should I rewrite it it non-recursive > > way (process packets with first filter, accumulare resulting packets, > > process with next one and so on)? > > The code in avconv.c has a nice way of dealing with the issue, I encourage > to look at it. > > We briefly had an API to apply several bitstream filters sequentially, until > it was NIHed by the fork. I would support re-adding one.
I had not noticed you replied in private. Please remember to keep the mailing-list in copy, and then no need to send to me directly since I am no the list. Regards, -- Nicolas George
signature.asc
Description: Digital signature
_______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
