In particular, after an initial fill of the codec's internal buffer, each
receive_frame call is followed by one or more send_packet calls. For
asynchronous codecs, this has the effect of letting the codec work while the
handler thread is waiting for an AVFrame structure to become available or
waiting for more decoded data to be needed. For synchronous codecs, this
makes the send_packet calls use up time that would be spent waiting.
Chris Robinson
9959d661a0