| Help on calling Fortran routines from IDL under linux [message #36586] |
Tue, 07 October 2003 23:45  |
isoaga2
Messages: 4
Registered: October 2003
|
Junior Member |
|
|
For the past few days i have been trying to call basic fortran
routines from IDL (IDL 6.0 and Gentoo Linux on a PC). First i tried
using the CALL_EXTERNAL approach using both fortran and C wrappers as
shown in the IDL documentation however the output i was getting from
the sum_array example was incorrect. Here is what i did first:
------------------------------------------------------------ ---------
test_call_external.pro - the IDL calling routine
------------------------------------------------------------ ---------
pro test_call_external
x = [1.0,2.0,3.0]
sum = 0.0
n = n_elements(x)
s = call_external('/home/david/PhD/Fortran/sum_array.so',
'sum_array', $
x, n, sum)
print, s
help, s
end
------------------------------------------------------------ ---------
sum_array.c - the C wrapper as in the IDL documentation
------------------------------------------------------------ ---------
#include <stdio.h>
void sum_array(int argc, void *argv[])
{
extern void sumd_(); /* Fortran Routine */
int *n;
float *s, *f;
f = (float *) argv[0]; /* Array pntr */
n = (int *) argv[1]; /* Get # of elements */
s = (float *) argv[2]; /* Pass back result a parameter */
sumd_(f, n, s); /* Compute Sum */
}
------------------------------------------------------------ ---------
sumd.f - the Fortran routine as in the documentation
------------------------------------------------------------ ---------
SUBROUTINE sumd(array, n, sum)
INTEGER*4 n
REAL*4 array(n), sum
sum=0.0
DO i=1,n
sum = sum + array(i)
ENDDO
RETURN
END
I compiled both the sumd.f and sum_array.c files to objects then
linked them together into a shared object (sum_array.so) using gcc (i
think, i'm a gcc ultra newbie) as follows:
gcc -c sumd.f
gcc -c sum_array.c
gcc -shared -o sum_array.so sumd.o sum_array.o
However, when i compile and run test_call_external i get the output:
-1073746820
S LONG = -1073746820
This, i don't understand, so i figured i'd try the fortran wrapper
like this...
------------------------------------------------------------ ---------
dave.f - the Fortran routine as in the documentation
------------------------------------------------------------ ---------
SUBROUTINE SUM_ARRAY(argc, argv) !Called by IDL
INTEGER*4 argc, argv(*) !Argc and Argv are integers
j = LOC(argc) !Obtains the number of arguments (argc)
!Because argc is passed by VALUE.
CALL SUM_ARRAY1(%VAL(argv(1)), %VAL(argv(2)), %VAL(argv(3)))
RETURN
END
SUBROUTINE SUM_ARRAY1(array, n, sum)
INTEGER*4 n, test
REAL*4 array(n), sum
sum=0.0
DO i=1,n
sum = sum + array(i)
ENDDO
RETURN
END
I compiled this with gcc as (to get the shared object dave.so)
gcc -w -shared -o dave.so dave.f
and then run the following idl program...
------------------------------------------------------------ ---------
test_call_external2.pro - idl calling routine
------------------------------------------------------------ ---------
pro test_call_external
x = [1.0,2.0,3.0]
sum = 0.0
n = n_elements(x)
s = call_external('/home/david/PhD/Fortran/dave.so', 'sum_array__', $
x, n, sum)
print, s
help, s
end
the entry point has 2 trailing underscores due to the gcc compile
process i think, i used "nm dave.so" to find that out. After running
that idl program i get
-1073746824
S LONG = -1073746824
Which is almost identical to the c wrapper approach, anyone have an
idea as to my mistake?
The next thing i did was to try out the dlm approach described by
S.V.H. Haugun on http://www.astro.uio.no/~steinhh/idl/additions.html
using his cool perl script and ftnchek. I figured i test it out with
the example he suggested, ie this fortran routine...
------------------------------------------------------------ ---------
square.for - fortran rountine
------------------------------------------------------------ ---------
DOUBLE PRECISION FUNCTION SQUARE(X)
DOUBLE PRECISION X
SQUARE = X * X
END
Then, trying to follow his instructions for my machine i did:
ftnchek -nocheck -quiet -makedcl square.for
dlmform square.for
gcc -o square.fo -c square.f
gcc -c square.c
gcc -shared -o square.so square.o square.fo
mv square.so /usr/local/rsi/idl/bin/bin.linux.x86/square.so
mv square.dlm /usr/local/rsi/idl/bin/bin.linux.x86/square.dlm
restarted idl and did
dlm_load, 'square'
% Loaded DLM: SQUARE.
but then,
IDL> print, square(4)
% Variable is undefined: SQUARE.
% Execution halted at: $MAIN$
IDL>
i checked the loaded module with
% Execution halted at: $MAIN$
IDL> help, /dlm, name='square'
** SQUARE - Subroutines: SQUARE (loaded)
Version: <unknown>, Build Date: 8 Oct 2003, Source: Perl script
written by S. V. H. Haugan
Path: /usr/local/rsi/idl_6.0/bin/bin.linux.x86/square.so
IDL>
So i'm not sure why that is not working either :(
Anyhelp would be great, Thanx.
|
|
|
|
| Re: Help on calling Fortran routines from IDL under linux [message #36667 is a reply to message #36586] |
Thu, 09 October 2003 02:34  |
Nigel Wade
Messages: 286
Registered: March 1998
|
Senior Member |
|
|
David Green wrote:
> Nigel Wade <nmw@ion.le.ac.uk> wrote in message news:<bm10fa$ef8$1@south.jnrs.ja.net>...
>
>>
>> Hmm, shouldn't that be 'print, sum'? Your C wrapper puts the sum in the 3rd
>> parameter, argv[2].
>>
>> Also, your C wrapper return type is void, so s in IDL will have no value.
>
>
> Hey thanx Nigel, you are right of course, i thouhgt that the s
> variable would contain the answer however this is obviously not the
> case, do you know what it does contain?
>
> Using the fortran wrapper works fine and print, sum gives the correct
> answer. Now to apply this approach to a more complicated function...
It's the return value of the function called.
Since your external routine has void as it's return type it doesn't return
anything, so s will contain garbage.
--
Nigel Wade, System Administrator, Space Plasma Physics Group,
University of Leicester, Leicester, LE1 7RH, UK
E-mail : nmw@ion.le.ac.uk
Phone : +44 (0)116 2523548, Fax : +44 (0)116 2523555
|
|
|
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
