Messages in this thread | | | Date | Fri, 26 Jun 1998 23:27:31 +0200 (MET DST) | From | MOLNAR Ingo <> | Subject | Re: Scheduler fixes |
| |
On Thu, 25 Jun 1998, Pavel Machek wrote:
> > not the case on Intel SMP at least. > > I thought we are talking about generic code. And I believe this would > be possible even on intel: consider 60MHz but clock and each processor > having different internal clock multiplier.
yes this is technically possible, but is not done in RL for obvious reasons. We would break on a couple of other places in the time (and profiling) code. The first time we encounter such a board (which mixes different speed pentiums and stuff) we have to think about this case too.
> > get_slow() executes in ~6-7 microseconds, we do not want to do that ... > > but then again, this whole issue is only a problem on higly loaded > > ISP's > > No. Whole issue is problem as soon as you run qt application (kde*, > qtscape, qweb, ...), which does busy waiting. Because its busy waiting > loop is close, manages to finish in less than jiffie and that means it > is invisible. Authors of qt maybe are not even aware of what they are > doing.
i see. Lets try the rdtsc driven stuff in 2.3. We really cannot do much on CPUs that have 1) only 1 clock source 2) no way to measure microtime in an efficient way.
-- mingo
- To unsubscribe from this list: send the line "unsubscribe linux-kernel" in the body of a message to majordomo@vger.rutgers.edu
| |