On Thu, Aug 25, 2016 at 9:47 AM, Vittorio Giovara <[email protected]> wrote: > On Thu, Aug 25, 2016 at 4:07 AM, Anton Khirnov <[email protected]> wrote: >> Quoting Vittorio Giovara (2016-08-24 16:58:12) >>> also, does it have to be a separate function? If so, what does the old >>> function do that the new one does not? >> >> The doxy says what the difference is -- mainly that this function requires >> alignment. >> We could conceivably add some checks that make it fall back on the >> normal memcpy for unaligned data, but I'm not convinced that using this >> special copy for normal frames is a good idea. > > I can't comment on special accelerated frames, from a user > perspective, a single API that deals with all the copying would be a > better alternative IMO.
The API itself can't know what kind of frame you are copying, and using the special magic for general purpose memory is not ideal in all cases, so ... the user has to tell it somewhere, might as well be a different function for ease of implementation. - Hendrik _______________________________________________ libav-devel mailing list [email protected] https://lists.libav.org/mailman/listinfo/libav-devel
