On Thu, Jul 22, 1999 at 03:21:46PM +0200, Rogier Wolff wrote:
>
> I hope I misunderstood you.
>
> HZ is not only exported through /proc . It is an external interface,
> which applications can depend on. No matter what. They can depend on
> "read" reading in a piece from a file, they can expect "sleep (1)" to
> last a second, and they can expect HZ to be the value it was when the
> application was compiled.
>
Oops, mailing list temporal skew. Please ignore my comment about
sysconf(), I see that glibc doesn't make good use of it. And the
methodology of sysconf() is somewhat flawed in this case, as it's designed
in terms of 'conservative' vs. 'liberal' values (e.g. path name lengths).
Still, it's worth thinking about how to change HZ. A fixed 100HZ timer
tick is nice in terms of low overhead, but is kind of a pain for low
latency work.
I've got a piece of hardware in front of me with eight timer interrupts.
It's begging for periodic-tick on one plus earliest-timer on another.
Be nice if someone adds such support to Linux before I get around to
porting it :-)
miket
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/