Date: Mon, 29 Oct 2001 14:02:18 -0700 From: Nate Williams <nate@yogotech.com> To: Julian Elischer <julian@elischer.org> Cc: Nate Williams <nate@yogotech.com>, Poul-Henning Kamp <phk@critter.freebsd.dk>, Peter Wemm <peter@wemm.org>, arch@FreeBSD.ORG Subject: Re: 64 bit times revisited.. Message-ID: <15325.50138.784256.957313@caddis.yogotech.com> In-Reply-To: <Pine.BSF.4.21.0110291246200.26174-100000@InterJet.elischer.org> References: <15325.41600.832823.952280@caddis.yogotech.com> <Pine.BSF.4.21.0110291246200.26174-100000@InterJet.elischer.org>
next in thread | previous in thread | raw e-mail | index | archive | help
> why? Stealing bits means that folks will end up having to roll their own time-handling code, and is just a hack. Been there, done that, will avoid any solution that does it again. :( Nate > On Mon, 29 Oct 2001, Nate Williams wrote: > > > > yes you are right here.. > > > > > > But the two TOP bits of the nanosecond fields > > > are by definition always 0 > > > (you can only have up to 1,000,000,000 nano seconds in a partial second) > > > and 32 bits goes up to 4(American)billion, so the two top bits can safely > > > be used for multiplying the seconds scale by 4. (in UFS timestamps.. > > > you would never write a non-normalised time to disk) > > > also timestamps can't be before 1970 so making it unsigned > > > allows us to go to 2100+ and mutiplying it by for takes us to about 2600.. > > > > All I can say is *yuck*. > > > > > > > > Nate > > > > > > > >ufs has enough room to fix this.. > > > > > >there has been a field defined in the on disk inode for nanosecs > > > > > >in each of the time values... > > > > > >if we take the lowest 8 bits of that field and re-assign it to be > > > > > >the highest 8 bits of the seconds, then we have time accuracy down to > > > > > >microseconds still and we expand file times by a factor > > > > > >of 256 (which is all of recorded history plus some) > > > > > > > > > > > >we just always set those bits to 0 for the next 37 years and we don;t > > > > > >really lose time resolution and we gain compatibility with the future.. > > > > > >nothing these days has nonosecond resolution there anyhow.... > > > > > > > > Simply not true. We have pico second resolution in our product, which > > > > is necessary because we're using *really* fast transports, and need to > > > > do very precise timing. > > > > > > > > (We're not using FreeBSD now, but if we need that kind of resolution in > > > > 2001, I can easily see the need for much higher resolution in the > > > > future.) > > > > > > > > I'm with PHK here (can you believe it?). :) :) > > > > > > > > > > > > Nate > > > > > > > > > > To Unsubscribe: send mail to majordomo@FreeBSD.org with "unsubscribe freebsd-arch" in the body of the message
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?15325.50138.784256.957313>