Re: hi-res preemption tick

Posted Oct 25, 2007 22:20 UTC (Thu) by goaty (guest, #17783)
Parent article: Fair user scheduling and other scheduler patches

I thought that the cost of reprogramming hi-res timers was too high to do frequently, and
that's why we can't have a nanosleep() call that does what the name says? If hi-res timers can
be used for scheduling, then when can I have a nanosleep(1ms) call that actually sleeps for 1
millisecond and not 7?

Re: hi-res preemption tick

Posted Oct 27, 2007 0:53 UTC (Sat) by nevets (subscriber, #11875) [Link]

When high-resolution timers are enabled, they do just that. Your nanosleep would then get the
resolution of what you asked for. Of course without something like the RT patch, you may hit
other latencies in the system.