Path: utzoo!utgpu!news-server.csri.toronto.edu!rpi!zaphod.mps.ohio-state.edu!swrinde!elroy.jpl.nasa.gov!ucla-cs!ucla-ma!euphemia!pmontgom
From: pmontgom@euphemia.math.ucla.edu (Peter Montgomery)
Newsgroups: comp.lang.fortran
Subject: Re: MAX function, but carry sign
Message-ID: <1991Apr12.194835.9162@math.ucla.edu>
Date: 12 Apr 91 19:48:35 GMT
References: <282@venice.water.ca.gov> <1991Apr12.151728.12916@craycos.com>
Sender: news@math.ucla.edu
Organization: UCLA Mathematics Dept.
Lines: 33

In article <1991Apr12.151728.12916@craycos.com> jrbd@craycos.com (James Davies) writes:
>In article <282@venice.water.ca.gov> rfinch@caldwr.water.ca.gov (Ralph Finch) writes:

>>I wish to use the max function on absolute values, but carry the sign
>>of the original values. That is:

>>a=max(abs(a1),abs(a2)) ! but a has sign of either a1 or a2

>If you want the result to carry the sign of the maximum value 
>(which is the most likely case), then you could use the sign of the sum
>of the two, i.e.

>	a = sign(max(abs(a1),abs(a2)), a1+a2)

>This will give you the positive value if they are the same, but the sign
>of the larger absolute value if they are different.

	Another way uses the Fortran Extended MERGE function, which
adds a much-needed conditional expression construct to Fortran.  As I recall, 
MERGE(x, y, bool) returns x if bool is TRUE, y if bool is FALSE.  Then

	a = MERGE(a1, a2, ABS(a1).ge.ABS(a2)) .

This makes it obvious to a reader that a = a1 or a = a2.
Depending upon your system, it may be preferable to replace the last
test by the mathematically equivalent  a1**2 .ge. a2**2  
(assuming no overflow/underflow will occur) or by  (a1+a2)*(a1-a2).ge.0.0 .

--
        Peter L. Montgomery 
        pmontgom@MATH.UCLA.EDU 
        Department of Mathematics, UCLA, Los Angeles, CA 90024-1555
If I spent as much time on my dissertation as I do reading news, I'd graduate.