Date: Fri, 26 Oct 2001 15:29:30 -0700 From: Bakul Shah <bakul@bitblocks.com> To: Poul-Henning Kamp <phk@critter.freebsd.dk> Cc: Peter Wemm <peter@wemm.org>, arch@FreeBSD.ORG Subject: Re: 64 bit times revisited.. Message-ID: <200110262229.SAA07928@marlborough.cnchost.com> In-Reply-To: Your message of "Fri, 26 Oct 2001 21:46:51 %2B0200." <6790.1004125611@critter.freebsd.dk>
next in thread | previous in thread | raw e-mail | index | archive | help
> The problem is that people tend to think of time as integers > instead of a floating point value. Precisely! So what I am suggesting is to count in the smallest unit that makes sense on a machine. Associate the number of zoptoseconds (or whatever) per tick and add that to your 96 bit kernel time. So adjtime() will change that zs/tick count. When you need to create a file timestamp, divide by appropriate 10^N number to map it to the correct unit. I just can't believe that this operation is so frequent so as to require a micro (1/2^20) optimization. The two are not very differen but there is a lot less of rounding error given the number of decimal clocks in use: 10/100/1000Mhz ethernet, SONET(where 8K frames are sent per second) and so on. Since they (1/2^64 versus a min. unit of zoptosecond) are so similar either is fine with me as far as the kernel time is concerned. I was really more interested in what gets stored in a file inode. For that I would very strongly argue for an integer multiple of a basic time unit of ns instead of timespec or timeval. 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?200110262229.SAA07928>