topic Calling fortran code from C in Intel® Fortran Compiler

Calling fortran code from C

oguren — Sun, 16 May 2010 00:55:07 GMT

Hi,
I need to compute hankel functions with complex argumant in c code. I try to use amos' codes written in fortran. I get some errors. Could you help me please?

$ ifort -r8 -c *.f

$ icc -c testhankel.c

$ icc -o hank testhankel.o zbesh.o zbesj.o d1mach.o i1mach.o zabs_org.o zuoik.o zbknu.o zacon.o zbunk.o zmlt.o zunk1.o zunk2.o zs1s2.o zseri.o zshch.o zsqrt_org.o zdiv.o zuchk.o dgamln.o zlog_org.o zexp_org.o zairy.o zunhj.o zunik.o xerror.o zkscl.o zbinu.o zacai.o zasyi.o zmlri.o zwrsk.o zbuni.o zrati.o zuni1.o zuni2.o -lm

d1mach.o: In function `d1mach_':
d1mach.f:(.text+0xa2): undefined reference to `for_write_seq_lis'
d1mach.f:(.text+0xbb): undefined reference to `for_write_seq_lis_xmit'
d1mach.f:(.text+0xe2): undefined reference to `for_write_seq_lis_xmit'
d1mach.f:(.text+0xff): undefined reference to `for_stop_core'
d1mach.f:(.text+0x12a): undefined reference to `for_stop_core'
i1mach.o: In function `i1mach_':
i1mach.f:(.text+0x4e): undefined reference to `for_write_seq_fmt'
i1mach.f:(.text+0x6b): undefined reference to `for_stop_core'
xerror.o: In function `xerror_':
xerror.f:(.text+0x72): undefined reference to `for_write_seq_fmt'
xerror.f:(.text+0xd2): undefined reference to `for_write_seq_lis'
xerror.f:(.text+0x116): undefined reference to `for_write_seq_fmt'

Calling fortran code from C

TimP — Sun, 16 May 2010 04:36:36 GMT

You have Fortran WRITE in those subroutines. You will need to include the Fortran run-time libraries in the link, or comment out those WRITEs. Should you hit a case which activates those WRITEs, you have a potential conflict if they use the same output stream as your C code.

Calling fortran code from C

oguren — Sun, 16 May 2010 05:24:02 GMT

Thanks for your reply,

I include run-time libraries. It runs but gives wrong result. I'm confusedI write a test program in fortran named test.f and run as

$ifort -r8 -i4 -c test.f zbesh.f zbesj.f d1mach.f i1mach.f zabs_org.f zuoik.f zbknu.f zacon.f zbunk.f zmlt.f zunk1.f zunk2.f zs1s2.f zseri.f zshch.f zsqrt_org.f zdiv.f zuchk.f dgamln.f zlog_org.f zexp_org.f zairy.f zunhj.f zunik.f xerror.f zkscl.f zbinu.f zacai.f zasyi.f zmlri.f zwrsk.f zbuni.f zrati.f zuni1.f zuni2.f

$ifort -o test test.o zbesh.o zbesj.o d1mach.o i1mach.o zabs_org.o zuoik.o zbknu.o zacon.o zbunk.o zmlt.o zunk1.o zunk2.o zs1s2.o zseri.o zshch.o zsqrt_org.o zdiv.o zuchk.o dgamln.o zlog_org.o zexp_org.o zairy.o zunhj.o zunik.o xerror.o zkscl.o zbinu.o zacai.o zasyi.o zmlri.o zwrsk.o zbuni.o zrati.o zuni1.o zuni2.o

$./test

It gives totaly true result as MATLAB besselh function

But when I call same programs from C routine , it gives completely wrong results. my commands like this:

$ icc -c testhankel.c

$ ifort -r8 -i4 -c zbesh.f zbesj.f d1mach.f i1mach.f zabs_org.f zuoik.f zbknu.f zacon.f zbunk.f zmlt.f zunk1.f zunk2.f zs1s2.f zseri.f zshch.f zsqrt_org.f zdiv.f zuchk.f dgamln.f zlog_org.f zexp_org.f zairy.f zunhj.f zunik.f xerror.f zkscl.f zbinu.f zacai.f zasyi.f zmlri.f zwrsk.f zbuni.f zrati.f zuni1.f zuni2.f

$ icc -o hank testhankel.o zbesh.o zbesj.o d1mach.o i1mach.o zabs_org.o zuoik.o zbknu.o zacon.o zbunk.o zmlt.o zunk1.o zunk2.o zs1s2.o zseri.o zshch.o zsqrt_org.o zdiv.o zuchk.o dgamln.o zlog_org.o zexp_org.o zairy.o zunhj.o zunik.o xerror.o zkscl.o zbinu.o zacai.o zasyi.o zmlri.o zwrsk.o zbuni.o zrati.o zuni1.o zuni2.o -L/RS/progs/intel/Compiler/11.1/056/lib/intel64 -lifcore -limf

Calling fortran code from C

mecej4 — Sun, 16 May 2010 12:17:00 GMT

You did not show any of the code in testhankel.c that gave you "wrong" results, so I content myself with giving you a C sample code that gives the same results as Fortran. I downloaded the AMOSLIB sources from Netlib, made the necessary configuration changes to d1mach.f and i1mach.f, and ran the commands

[bash]ifort -w -c *.f
gcc -c testhankel.c
ifort -nofor-main *.o
./a.out [/bash] with testhankel.c containing

[bash]#include <stdio.h>
extern void zbesh_(double *,double *,double *,int *,int *, int *,double *,double*,int *,int *);
int main(int argc,char *argv[]){
int i,ierr,nz,m=1,n=3,kode=2;
double zr=7e-2,zi=5e-2,cyr[3],cyi[3],fnu=0.0;
zbesh_(&zr,&zi,&fnu,&kode,&m,&n,cyr,cyi,&nz,&ierr); ...
[/bash] and obtained the results

[bash] 5.0982e-01 -1.7590e+00
-4.8843e+00 -6.0601e+00
-1.7481e+02 -4.6887e+01 [/bash] which agree with the output

[bash] nZ = 0 IERR = 0
0.5098e+00 -0.1759e+01
-0.4884e+01 -0.6060e+01
-0.1748e+03 -0.4689e+02 [/bash]
from the equivalent Fortran driver:

[bash] PROGRAM TZBESH
IMPLICIT NONE
DOUBLE PRECISION ZR,ZI,FNU,CYR(3),CYI(3)
integer m,n,nz,ierr,kode,i
ZR=7d-2
ZI=5d-2
FNU=0d0
M=1
N=3
KODE=2
CALL ZBESH(ZR, ZI, FNU, KODE, M, N, CYR, CYI, NZ, IERR)
write(*,*)' nZ = ',nz,' IERR = ',ierr
write(*,'(1x,2E12.4)')(CYR(i),CYI(i),i=1,3)
stop
end [/bash]

Hello, I need help with

Muhammet_M_ — Sat, 04 May 2013 11:37:00 GMT

Hello, I need help with hankel function.

I downloaded amoslib from netlib (zbesh.f and dependencies)

ifort -w -c *.f
gcc -c -lm wire.c
ifort -nofor-main *.o

then ./a.out

gives me very different answer. I am suspecting

that the problem is in i1mach fortran routine. I did not customize d1mach because

i read that it gets values from i1mach.

wire.c

extern void zbesh_ (double*, double*, double*, int*, int*, int*, double*, dou
ble*, int*, int*);

void main (void)
{
double zr = 0.07; double zi = 0.05; double fnu;
int kode, M, N, nz, ierr, i;
double cyi[2], cyr[2];

fnu = 0.0; kode = 1; M = 1; N = 1;

zbesh_(&zr, &zi, &fnu, &kode, &M, &N, cyr, cyi, &nz, &ierr);

for (i=0; i<2; i++)
printf("real %e\t imag %e \n", cyr, cyi);

I have 64 bit laptop (cpu is Intel(R) Core(TM) i5 CPU M 450 @ 2.40GHz)

and my custom i1mach.f file below

      DATA IMACH( 1) /          5 /
      DATA IMACH( 2) /          6 /
      DATA IMACH( 3) /          7 /
      DATA IMACH( 4) /          0 /
      DATA IMACH( 5) /         8 /
      DATA IMACH( 6) /          8 /
      DATA IMACH( 7) /          2 /
      DATA IMACH( 8) /         31 /
      DATA IMACH( 9) / 2147483647 /
      DATA IMACH(10) /          2 /
      DATA IMACH(11) /         24 /
      DATA IMACH(12) /       -126 /
      DATA IMACH(13) /        127 /
      DATA IMACH(14) /         53 /
      DATA IMACH(15) /      -1022 /
      DATA IMACH(16) /       1023 /

what am I missing here?

thanks in advance

Quote:Muhammet M. wrote:

mecej4 — Sat, 04 May 2013 13:24:00 GMT

Muhammet M. wrote:

gives me very different answer. I am suspecting that the problem is in i1mach fortran routine. I did not customize d1mach because I read that it gets values from i1mach.

"Different answer" compared to what? For your argument values, zbesh gives H_ν⁽¹⁾(z) = 0.600799 - 1.635214*I, which agrees with the result from Maple.

You do need to remove comment marks from the lines of d1mach.f that correspond to "IBM PC", since the routine d1mach gets called soon after entry into zbesh, and the version of d1mach provided in the Netlib amos directory does not call i1mach().

Comments to IDZ forum managers: The inline code in my post of 05/16/2010 above was shown correctly in the old version of the forum. Now that this thread has been reactivated, I see that the new forum software has swallowed C include-file names, removed line endings and, in general, made the code unusable. I have tried to edit and restore the code to the point where it is readable in the new forum software, but I have not run the inline code through a compiler to verify the edits.

Thanks mecej4

Muhammet_M_ — Sat, 04 May 2013 22:37:56 GMT

Thanks mecej4

It worked after I uncommented the IBM PC constants.