Re: [patch] jiffies wraparound [Re: 2.1.125 Show stopper list: Draft]

Richard B. Johnson (root@chaos.analogic.com)
Fri, 23 Oct 1998 08:55:55 -0400 (EDT)


On Fri, 23 Oct 1998, Albert D. Cahalan wrote:
[SNIPPED]
>
> Gee. Someday I will get a 5000-MHz CPU running the scheduler at 100 HZ.
> My 30 Gb/s fiber optic serial ports will work great, eh? Pardon me
> while I puke.

>From the user-mode code, 100Hz represents the rate at which the CPU can
be taken away from you, i.e., 100 times per second you lose. 100Hz may
not be "optimum", but you certainly would not want to increase it very
much.

If you had a 5GHz CPU, you would not worry very much about this because
within each time-slice, you would be able to do 5GHz/now-Mhz more work
than you do now.

When you get a 30Gb/s controller, it will signal the kernel via interrupt
after it has accumulated enough data to make it worthwhile. The context-
switch rate means nothing here. Interrupts don't wait.

Higher switch rates are sometimes are useful if you are doing realtime
process control because the switch rate may define the upper bound of
the dominant pole of the controller's transfer function. Then again
this is not necessarily always the case because realtime functions are
often performed in driver's interrupt service routines rather than
in user-mode code.

Cheers,
Dick Johnson
***** FILE SYSTEM WAS MODIFIED *****
Penguin : Linux version 2.1.123 on an i586 machine (66.15 BogoMips).
Warning : It's hard to remain at the trailing edge of technology.

-
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/