13 Jul
2004
13 Jul
'04
4:29 p.m.
Some thoughts about low latency with dual processor or HT capable systems.
Isn't it possible to get low latency by locking everything (= all processes +
interrupts) except for the audio/midi interrupt and the RT process(es) to
one cpu and the latter two to the second, using the cpu and irq affinity calls
?
I thought this way interrupts, bottom halfs and the like are handled per cpu
and don't interfere (lock each other out) between cpus and processes not using
system calls should never be blocked.
If it is that way, this would help at least the MP/HT systems,regardless of
the kernel used.
Or am i thinking wrong here?