Date: Tue, 8 Nov 2005 12:26:13 +0100 From: Michael Schuh <michael.schuh@gmail.com> To: freebsd-stable@freebsd.org Subject: Re: timecounter and Hz quality in kern RELENG_6 Message-ID: <1dbad3150511080326p62ec11at@mail.gmail.com> In-Reply-To: <1dbad3150511070339k15a99854r@mail.gmail.com> References: <1dbad3150511070339k15a99854r@mail.gmail.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Hello, thanks to Oliver and Michael for explain me the "mechanics" behind timecounter and "Hz quality". I have send the question, why i have to wonder about different IRQ-Counts but same HZ. It is really thaat HZ on both machines is the same. I have only knowed the old behavior. I set the HZ over Options HZ in my Kernelconfig, and then i could see that the IRQCount's on clk are alwas the same value as HZ +- a few tics. So that i have thinked ( and feeled on busy machines) that the IRQ-work is more smooth and they are less holes in responsitivity if the HZ is going to an Value > 1000. So i have alwas set HZ to 2000 on every machine. If i have the answers and the "mechanics" behind timecounter and "Hz quality" right understand, so i have no more modify the HZ Value to become an smooth polling of the interupts. if i be right the smooth behavior comes with the calculation from "Hz quality". Please correct me if i am wrong. thanks to all and best regards michael 2005/11/7, Michael Schuh <michael.schuh@gmail.com>: > Hello, > > i be very surprised about the performance of RELENG_6. > Congratulations to the entire Team for this very good work. > > Now i have 2 Machines installed with 6.0-RC1, and i have seen that on > both machines the Hz is differntly with GENERIC-Kernel. > Machine A is an Sempron 2400+ that runs as 2500+ > (i have tuned the clock to best RAM-Performace) > Machine B is an Duron 700MHz > > On Machine A i got an Hz from 2000 effectively > systat -vmstat 1 show me 2000 IRQ/s on clk > sysctl say's 1000....i think, but not sure > > On Machine B i got an Hz from 1000 effectively > systat-vmstat 1 show me 1000 IRQ/s on clk > > After digging in the source i have found that timec.c have an routine for > computing the so called "Hz quality". > > Can anyone explain me the "mystics" behind Hz quality, > and why or how this quality is computed and what are the > efforts? > > My knowledge is not deep enough to know these details. > > thanks > > best regards > > michael >
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1dbad3150511080326p62ec11at>