Double complex with gcc and Intel v9 Fortran (PR#8699)

classic Classic list List threaded Threaded
2 messages Options
Reply | Threaded
Open this post in threaded view
|

Double complex with gcc and Intel v9 Fortran (PR#8699)

marquardt.christian
Full_Name: Christian Marquardt
Version: 2.2.1
OS: Linux
Submission from: (NULL) (84.167.229.240)


Hello,

I believe this is a bug in the configuration / installation:

When configuring R-2.2.1 using the Intel v9 Fortran compiler as default Fortran
compiler and g++ as C++ compiler on a Suse 9.3 Linux, the configuration script
finds that the C and Fortran idea of double comples disagree.

I have tried to extract the test used for this from m4/R.m4 (a slightly modified
version of the two test files is added at the end). When compiling

   ifort -c ctest.f
   gcc -c cftest.c

and linking properly,

   ifort -nofor_main cftest.o ctest.o -o runme

running the executable gives the following output:

   ./runme
123.456000 14.710644
0.000000 -0.000006

To me, this inicates that the test should actually be passed successfully.

Unfortunately, I haven't quite understood how to rebuild the configure.ac etc
once R.m4 is changed; otherwise I would try to come up with a patch for R.m4.
But maybe someone else can fix this?

Many thanks,

  Christian.


----<ctest.f>------------------------------------
      subroutine cftest(x)
      complex*16 x(3)
      integer i

c a few tests of constructs that are sometimes missing
      if(x(1) .eq. x(1)) i = 0
      x(1) = x(1)*x(2) + x(3)
      end

----<cftest.c>------------------------------------
#include <math.h>
#define HAVE_F77_UNDERSCORE 1
#ifdef HAVE_F77_UNDERSCORE
# define F77_SYMBOL(x)   x ## _
#else
# define F77_SYMBOL(x)   x
#endif

typedef struct {
        double r;
        double i;
} Rcomplex;

extern void F77_SYMBOL(cftest)(Rcomplex *x);

int main () {
    Rcomplex z[3];

    z[0].r = 3.14159265;
    z[0].i = 2.172;
    z[1].i = 3.14159265;
    z[1].r = 2.172;
    z[2].r = 123.456;
    z[2].i = 0.123456;
    F77_SYMBOL(cftest)(z);
    printf("%f %f\n", z[0].r, z[0].i);
    printf("%f %f\n", z[0].r - 123.456, z[0].i - 14.71065);
    if(fabs(z[0].r - 123.456) < 1e-4 && fabs(z[0].i - 14.71065) < 1e-4)
        exit(0);
    else exit(1);
}

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel
Reply | Threaded
Open this post in threaded view
|

Re: Double complex with gcc and Intel v9 Fortran (PR#8699)

Brian Ripley
On Wed, 22 Mar 2006, [hidden email] wrote:

> Full_Name: Christian Marquardt
> Version: 2.2.1
> OS: Linux
> Submission from: (NULL) (84.167.229.240)
>
>
> Hello,
>
> I believe this is a bug in the configuration / installation:
>
> When configuring R-2.2.1 using the Intel v9 Fortran compiler as default
> Fortran compiler and g++ as C++ compiler on a Suse 9.3 Linux, the
> configuration script finds that the C and Fortran idea of double comples
> disagree.
>
> I have tried to extract the test used for this from m4/R.m4 (a slightly
> modified version of the two test files is added at the end). When
> compiling
>
>   ifort -c ctest.f
>   gcc -c cftest.c
>
> and linking properly,
>
>   ifort -nofor_main cftest.o ctest.o -o runme
>
> running the executable gives the following output:
>
>   ./runme
> 123.456000 14.710644
> 0.000000 -0.000006
>
> To me, this inicates that the test should actually be passed successfully.

No, because R is linked with the C compiler and not the Fortran compiler.
Did you look in config.log and see what it indicated the error was?

You have not shown us anything that indicates that this is a bug in R.

You also haven't told us what architecture this is: I tried mixing gcc
3.4.4 and ifort 9.0 on x86_64 Fedora Core 3, and had no problems.

> Unfortunately, I haven't quite understood how to rebuild the configure.ac etc
> once R.m4 is changed;

I presume you mean to rebuild 'configure': configure.ac does not depend on
R.m4.  Just configure in maintainer mode, and 'make'.

> otherwise I would try to come up with a patch for R.m4.
> But maybe someone else can fix this?

We cannot fix things that we cannot reproduce.

>
> Many thanks,
>
>  Christian.
>
>
> ----<ctest.f>------------------------------------
>      subroutine cftest(x)
>      complex*16 x(3)
>      integer i
>
> c a few tests of constructs that are sometimes missing
>      if(x(1) .eq. x(1)) i = 0
>      x(1) = x(1)*x(2) + x(3)
>      end
>
> ----<cftest.c>------------------------------------
> #include <math.h>
> #define HAVE_F77_UNDERSCORE 1
> #ifdef HAVE_F77_UNDERSCORE
> # define F77_SYMBOL(x)   x ## _
> #else
> # define F77_SYMBOL(x)   x
> #endif
>
> typedef struct {
>        double r;
>        double i;
> } Rcomplex;
>
> extern void F77_SYMBOL(cftest)(Rcomplex *x);
>
> int main () {
>    Rcomplex z[3];
>
>    z[0].r = 3.14159265;
>    z[0].i = 2.172;
>    z[1].i = 3.14159265;
>    z[1].r = 2.172;
>    z[2].r = 123.456;
>    z[2].i = 0.123456;
>    F77_SYMBOL(cftest)(z);
>    printf("%f %f\n", z[0].r, z[0].i);
>    printf("%f %f\n", z[0].r - 123.456, z[0].i - 14.71065);
>    if(fabs(z[0].r - 123.456) < 1e-4 && fabs(z[0].i - 14.71065) < 1e-4)
>        exit(0);
>    else exit(1);
> }
>
> ______________________________________________
> [hidden email] mailing list
> https://stat.ethz.ch/mailman/listinfo/r-devel
>
>

--
Brian D. Ripley,                  [hidden email]
Professor of Applied Statistics,  http://www.stats.ox.ac.uk/~ripley/
University of Oxford,             Tel:  +44 1865 272861 (self)
1 South Parks Road,                     +44 1865 272866 (PA)
Oxford OX1 3TG, UK                Fax:  +44 1865 272595

______________________________________________
[hidden email] mailing list
https://stat.ethz.ch/mailman/listinfo/r-devel