Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!usc!apple!agate!violet.berkeley.edu!jerry
From: jerry@violet.berkeley.edu (Jerry Berkman)
Newsgroups: comp.lang.fortran
Subject: Re: Arbitrary Digit Arithmetic
Message-ID: <1991Apr6.003438.24011@agate.berkeley.edu>
Date: 6 Apr 91 00:34:38 GMT
References: <6005@trantor.harris-atd.com>
Sender: usenet@agate.berkeley.edu (USENET Administrator)
Organization: University of California, Berkeley
Lines: 27

In article <6005@trantor.harris-atd.com> gbastin@x102c.ess.harris.com (Gary Bastin 60293) writes:
>I am looking for a way to perform arbitrary number rounding,
>up to the maximum possible for the data type, for arithmetical
>operations.  For example, assume that 4 digit accuracy is desired
>and REAL types are used, giving approximately 6 or 7 digits of
>accuracy, depending upon the platform.  Then, the problem is to
>perform the arithmetic operation, and round the result to 4 digits.

>...has anyone got any F77 source code for such a function that would be
>portable between platforms?
>
>Thanks in advance!
>
>Gary Bastin, WB4YAF      /-/-/      Internet: gbastin@x102c.ess.harris.com
>Mail Stop 102-4826         |        phone: (407) 729-3045
>Harris Corporation GASD    |        
>P.O.B. 94000, Melbourne FL 32902    Speaking from, but not for, Harris! 

Is this what you want?

	round(x,n) = anint(x*10**n)/10**n

Of course the answers can't be exact because you are trying to "round"
to numbers which can not be exactly represented in binary.

	- Jerry Berkman, U.C. Berkeley,
	  jerry@violet.berkeley.edu, (415)642-4804