Ok, I've answered my own question... hopefully this might be of use to
someone else.
I've decided to use NAGWare F95, available at www.nag.com. They will
allow you to download and get a trial license (good for a month or
two) for free. After that an academic license is $240 + $15 shipping.
NAGWareF95 does not include %VAL(), so my above example with g77
wouldn't work, but, I've found two different ways of using
call_external with NAGWareF95:
1. use a C wrapper instead of a fortran wrapper.
2. compile a f77 wrapper using g77.
A simple example of each of these follows (both do the same few simple
operations; NOTE that the f95 subroutine names differ by an
underscore):
1. Using a C wrapper:
-------------------file osx_test6.f95---------------------------
SUBROUTINE osx_test6(n1,n2,nl1,nl2,f1,f2,d1,d2,b1,b2)
use f90_kind
IMPLICIT NONE
INTEGER(int16)::n1,n2
INTEGER(int32)::nl1,nl2
REAL(single)::f1(n1),f2(n1)
REAL(double)::d1(n2),d2(n2)
INTEGER(kind=int8)::b1,b2
!add one to each of b1,b2
b1 = b1 + 1
b2 = b2 + 1
!we will copy f1 to f2 and then place the sum of f2 in f1(1)
f2 = f1
f1(1) = SUM(f2)
!we will copy d1 to d2 and then place the sum of d2 in d1(1)
d2 = d1
d1(1) = SUM(d2)
!set nl2 = nl1
nl2 = nl1
return
end
-------------------end of file osx_test6.f95--------------------
---------------------file call_osx_test6.c----------------------
#include <stdio.h>
void call_osx_test6(int argc, void *argv[])
{
extern void osx_test6_(); /* FORTRAN routine */
int *n1,*n2; /* int is equivalent to IDL's standard integer */
long int *nl1,*nl2; /* long int is equivalent to IDL's long integer
*/
float *f1,*f2; /*float is equivalent to IDL's float */
double *d1,*d2; /*double is equivalent to IDL's double */
unsigned char *b1, *b2; /*unsigned char is equivalent to IDL's byte
*/
n1 = (int *) argv[0];
n2 = (int *) argv[1];
nl1 = (long int *) argv[2];
nl2 = (long int *) argv[3];
f1 = (float *) argv[4];
f2 = (float *) argv[5];
d1 = (double *) argv[6];
d2 = (double *) argv[7];
b1 = (unsigned char *) argv[8];
b2 = (unsigned char *) argv[9];
osx_test6_(n1,n2,nl1,nl2,f1,f2,d1,d2,b1,b2);
}
--------------end of file call_osx_test6.c----------------------
To compile:
f95 -PIC -c call_osx_test6.c
f95 -PIC -c osx_test6.f95
f95 -bundle -flat_namespace -dynamic -lm -lc \
-o osx_test6.so call_osx_test6.o osx_test6.o
(there will be a few warnings)
To run an example in IDL:
n1=5
n2=3
nl1=25
nl2=33
f1=findgen(n1)
f2=fltarr(n1)
d1=dindgen(n2)
d2=dblarr(n2)
b1 = 127B
b2 = 255B
print,n1,n2,nl1,nl2
print,f1
print,f2
print,d1
print,d2
print,b1,b2
t = CALL_EXTERNAL( 'osx_test6.so','call_osx_test6', &
n1,n2,nl1,nl2,f1,f2,d1,d2,b1,b2)
print,n1,n2,nl1,nl2
print,f1
print,f2
print,d1
print,d2
print,b1,b2
---------END of C wrapper example
2. compile f77 wrapper with g77, f95 code with f95
-------------------file osx_test7.f95---------------------------
SUBROUTINE osx_test7_(n1,n2,nl1,nl2,f1,f2,d1,d2,b1,b2)
use f90_kind
IMPLICIT NONE
INTEGER(int16)::n1,n2
INTEGER(int32)::nl1,nl2
REAL(single)::f1(n1),f2(n1)
REAL(double)::d1(n2),d2(n2)
INTEGER(kind=int8)::b1,b2
!add one to each of b1,b2
b1 = b1 + 1
b2 = b2 + 1
!we will copy f1 to f2 and then place the sum of f2 in f1(1)
f2 = f1
f1(1) = SUM(f2)
!we will copy d1 to d2 and then place the sum of d2 in d1(1)
d2 = d1
d1(1) = SUM(d2)
!set nl2 = nl1
nl2 = nl1
return
end
------------end of file osx_test7.f95---------------------------
-------------------file call_osx_test6.f------------------------
FUNCTION call_osx_test7(ARGC, ARGV)
IMPLICIT NONE
INTEGER*4 ARGC !Argument count
INTEGER*4 ARGV(*) !Vector of pointers to
argments
INTEGER ARG_CNT
INTEGER*4 call_osx_test7
C The argument count is passed in by value. Get the location of
C this value in memory (a pointer) and convert it into an
C Fortran integer.
ARG_CNT = LOC(ARGC)
C Insure that we got the correct number of arguments
IF(ARG_CNT .ne. 10)THEN
WRITE(*,*)': Incorrect number of arguments'
call_osx_test7 = -1
RETURN
ENDIF
C To convert the pointers to the IDL variables contained in ARGV
C we must use the Fortran function %VAL. This funcion is used
C in the argument list of a Fortran sub-program. Call the
Fortran
C subroutine that will actually perform the desired operations.
C Set the return value to the value of this function.
CALL osx_test7( %val(ARGV(1)), %val(ARGV(2)), %val(ARGV(3)),
& %val(ARGV(4)), %val(ARGV(5)), %val(ARGV(6)), %val(ARGV(7)),
& %val(ARGV(8)), %val(ARGV(9)), %val(ARGV(10)) )
C Thats all, return to IDL.
call_osx_test7 = 1
RETURN
END
--------------end of file call_osx_test6.f----------------------
to compile:
g77 -fPIC -c call_osx_test7.f
f95 -PIC -c osx_test7.f95
g77 -bundle -flat_namespace -dynamic -lm -lc -undefined
suppress \
-o osx_test7.so osx_test7.o
call_osx_test7.o
To run an example in IDL:
n1=5
n2=3
nl1=25
nl2=33
f1=findgen(n1)
f2=fltarr(n1)
d1=dindgen(n2)
d2=dblarr(n2)
b1 = 127B
b2 = 255B
print,n1,n2,nl1,nl2
print,f1
print,f2
print,d1
print,d2
print,b1,b2
t = CALL_EXTERNAL( 'osx_test6.so','call_osx_test6', &
n1,n2,nl1,nl2,f1,f2,d1,d2,b1,b2)
print,n1,n2,nl1,nl2
print,f1
print,f2
print,d1
print,d2
print,b1,b2
---------END of C g77/f95 example
|
comp.lang.idl-pvwave archive
Members
Search
Help
Login
Home
