On Tue, Oct 14, 2008 at 09:21:07AM +0200, Leslie P. Polzer wrote: Thanks for all the replies! The JACK buffer seems very nice, and probably the fastest and simplest approach (though, not the most abstracted). By not requiring locks, that'd make dealing with pthreads much simpler-- just create and then join the threads, no need to deal with mutexes, etc. The data I'm buffering are ALSA MIDI snd_seq_event_t pointers. I'm using C because this is a simple little daemon, and I am a latency freak and I wanted speed. If I were using Python (or some other high-level language) then the threading and buffering would be ridiculously simple, of course. After researching it a bit, it seems that the simplest way with the simplest code would be to use an "asynchronous queue", as it's called in Glib. Just cast the pointer to void, attach it to a message, and push it onto a queue. In the consumer thread, just push it off of the queue and cast it to snd_seq_event_t * again. Glib is very easy to use, but the JACK/pthreads approach seems lighter weight so I might try that first. Thanks again! -ken