Date: Thu, 2 Jul 1998 13:06:24 -0400 (EDT) From: "Ron G. Minnich" <rminnich@Sarnoff.COM> To: Luigi Rizzo <luigi@labinfo.iet.unipi.it> Cc: hackers@FreeBSD.ORG Subject: Re: timeout granularity (was: Re: Console driver...) Message-ID: <Pine.SUN.3.91.980702115006.1162C-100000@terra> In-Reply-To: <199807021418.QAA12618@labinfo.iet.unipi.it>
next in thread | previous in thread | raw e-mail | index | archive | help
> do you have a patch for the above ? I think most things work fine with > HZ>100 but i have heard that NTP has problems with it. well, this is embarassing. I did the patch on linux and the machine crashed and took the hard disk with it ... all i did was changed the constant put into the timer interrupt by dividing it by a constant factor. Then in hardclock I counted that constant factor times as many interrupts before causing a softclock interrupt. That way, I kept accurate time. I made it a compile time constant so that a rebuild of the clock code would allow me to tweak it. The test was to vary the interrupt rate and then run a program that counted to 10e10. I measured the wall-clock time it took the program to complete. at 2500 interrupts per second the degradation was (maybe) 1%. At 10k IPS degradation was at about 20%. This was a 486/25. I expect we now with P200+ have lots of power to accomodate 10k IPS but I have not tried it. ron To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-hackers" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?Pine.SUN.3.91.980702115006.1162C-100000>