This is fifo: a simple replacement for the shell pipe (|) with increased buffer size

Usage

./fifo [-b bufsize] [-r readsize] -b bufsize: set FIFO buffer size to bufsize bytes -r readsize: read in blocks of readsize bytes

How does it work?

fifo reads from stdin and writes everything it read to stdout immediately, just as the pipe (|) operator. Its only advantage to pipe is that it provides a configurable arbitrary sized fifo buffer. Its disadvantage is that it works in user space, but this is a little price to pay when the pipe would otherwise not provide enough buffering ;)

Why is it needed?

There are many ways of connecting process I/O under UNIX, the most common being the pipe (|) operator. Unfortunately, this operates with a limited buffer size and blocking I/O. Thus, it creates a very tight coupling between the two processes, blocking the feeder process when the consumer can't digest all the data. This is fine in some cases, but not in others.

This tight coupling can be a real pain for multimedia applications such as ffmpeg, ffplay, vlc, mplayer, etc., where the feeder might provide data in big chunks, and the consumer might block for long times while trying to interpret the stream (e.g. figure out the format or the codec).

Usage examples

Actually, fifo does not replace the pipe :), it has to be used with two pipes as follows:

feeder | fifo | consumer

It can be used where applications have some nasty buffering behavior, e.g. between dvbcat (that stalls on reception errors, then bursts after) and ffmpeg (that might load the CPU too much to handle the burst).

dvbcat -P dvb-t "BBC" | fifo | ffmpeg -i - a.avi