>Not just zero copy, you don't even need to interrupt at the period >boundary. The critical hardware feature is being able to mmap the >position register. Does any current hardware do this?
old ISA cards could do this :( i did some experiments, oh, maybe 4-5 years ago with my wavefront card using single-sample processing. pretty cool. >If Ingo's RT patches and the high res POSIX timers patch get merged then >Linux can provide the kind of deterministic RT scheduling described in actually, this is the 2nd of two critical features, one h/w, one s/w. --p
