On Tuesday 06 March 2007 07:38, Matt Mackall wrote: > On Sun, Mar 04, 2007 at 06:06:22PM +1100, Con Kolivas wrote: > > + * This is the time all tasks within the same priority round robin. > > + * Set to a minimum of 6ms. > > */ > > +#define RR_INTERVAL ((6 * HZ / 1001) + 1) > > What happens with small HZ? Like 100? I suppose 10ms is a reasonable > number here. But in general, how well does this scheduler do without a > time source other than jiffies?
Perfectly fine at 100. It only needs to do accounting during a busy tick, and on dynticks we always service a timer interrupt during that busy tick so it works fine there as well. In fact, choosing a lower HZ you are indirectly making a decision that you don't value latency as much as at the higher HZ and 10ms is fine for the scheduling there. It would be possible to service the cpu accounting with any timer that fired at at least 100HZ even if there was some way to turn ticks off fully when the cpu is busy (which there isn't currently a way to do). -- -ck - To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html Please read the FAQ at http://www.tux.org/lkml/
