Re: minloc function problem - test program

["S. Umar" <umar at compsci dot cas dot vanderbilt dot edu>]

 is attached!

On Tuesday 15 March 2005 01:38 pm, you wrote:
    | On Tue, Mar 15, 2005 at 01:21:15PM -0600, S. Umar wrote:
    | > Hi,
    | > 
    | > I just tried my large microscopic nuclear reactions code and found
    | > only one problem.
    | > 
    | > In a small routine I have a line
    | > 
    | > loc = minloc(e,  mask = .not. selected)
    | > 
    | > where the array selected is initially all False. I am just running
    | > through and finding minimum values. This works with Intel 8.x and
    | > Absoft 9.x.  But with gfortran it is interpreted like a scalar
    | > rather than an array.  Namely, the second dimension argument is
    | > usually "optional" (I think this is f95 addition to f90).  If you
    | > are doing f95 strictly you may have to have the dimension argument,
    | > but if you are compiling an f90 code you don't have to.  Anyway,
    | > the changed version:
    | > 
    | > loc = minloc(e, ns, mask = .not. selected)
    | > 
    | > works fine.
    | > 
    | 
    | Can you supply a small self-contained program exhibiting the
    | possibly incorrect behavior?  It may be helpful if you submit
    | the output from the program after you compile it with Intel
    | 8.x and Absoft 9.x.  There is the possibility that a bug exist
    | in the implementation of minloc.  Finally, submit the bug via
    | bugzilla at gcc.gnu.org or simply report it here.
    | 

-- 
Prof. S. Umar
Dept. of Physics & Astronomy
Vanderbilt University
Nashville, TN 37235
Tel : (615) 322-2459
umar@compsci.cas.vanderbilt.edu
http://compsci.cas.vanderbilt.edu/~umar/resu.htm

Attachment: test.f90
Description: Text document