GSL 0.9 and Tru64 Unix

GSL 0.9 and Tru64 Unix

[Olaf Lenz]

Hello!

I'm working with the GSL on Tru64 Unix on a Compaq Alpha Workstation.
The compilation and installation of GSL 0.9 works smooth and fine, so
I suppose it can be added on the list of "Supported platforms".

The only problem I have is that some tests in the Eigensystem part
fail in make check. The output of running 'make check' in the 'eigen'
directory of the sources is attached to this mail.

This may or may not be a platform specific problem, right now I can't
verify it. Any suggestions?

-- 
Cheers
  Olaf

Email to olenz@TechFak.uni-bielefeld.de ***** PGP encrypted and signed
Email to: pgp-public-keys@keys.pgp.net ************** mail is welcome.
subject: get olenz@TechFak.uni-bielefeld.de for public key. **********

Re: GSL 0.9 and Tru64 Unix

[Brian Gough]

Olaf Lenz writes:
 > Hello!
 > 
 > I'm working with the GSL on Tru64 Unix on a Compaq Alpha Workstation.
 > The compilation and installation of GSL 0.9 works smooth and fine, so
 > I suppose it can be added on the list of "Supported platforms".

Ok, will do.

 > The only problem I have is that some tests in the Eigensystem part
 > fail in make check. The output of running 'make check' in the 'eigen'
 > directory of the sources is attached to this mail.
 > 
 > This may or may not be a platform specific problem, right now I can't
 > verify it. Any suggestions?

Thanks for the bug report.  The code works ok with gcc/x86 but I have
not tested it beyond that.  Can you send me the results of the
following program, compiled with 

  gcc tmp.c -lgsl -lgslcblas -lm

this will give a dump of the raw eigenvalues/eigenvectors

#include <stdlib.h>
#include <stdio.h>
#include <gsl/gsl_test.h>
#include <gsl/gsl_math.h>
#include <gsl/gsl_blas.h>
#include <gsl/gsl_ieee_utils.h>
#include <gsl/gsl_complex_math.h>
#include <gsl/gsl_eigen.h>

gsl_matrix_complex *
create_random_herm_matrix(int size)
{
  int i, j;
  unsigned long k = 1;
  gsl_matrix_complex * m = gsl_matrix_complex_alloc(size, size);
  for(i=0; i<size; i++) {
    for(j=i; j<size; j++) {
      gsl_complex z;
      k = (69069 * k + 1) & 0xffffffffUL;
      GSL_REAL(z) = k / 4294967296.0;
      k = (69069 * k + 1) & 0xffffffffUL;
      GSL_IMAG(z) = (i == j) ? 0 : k / 4294967296.0;
      gsl_matrix_complex_set(m, i, j, z);
      gsl_matrix_complex_set(m, j, i, gsl_complex_conjugate(z));
    }
  }
  return m;
}

#define N 10

int main()
{
  gsl_matrix_complex *m = create_random_herm_matrix (N);

  gsl_matrix_complex * A = gsl_matrix_complex_alloc(N, N);
  gsl_matrix_complex * evec = gsl_matrix_complex_alloc(N, N);
  gsl_vector * eval = gsl_vector_alloc(N);
  gsl_vector * eval2 = gsl_vector_alloc(N);

  gsl_eigen_herm_workspace * w1 = gsl_eigen_herm_alloc (N);
  gsl_eigen_hermv_workspace * w2 = gsl_eigen_hermv_alloc (N);

  printf("m = [\n"); 
  gsl_matrix_complex_fprintf(stdout, m, "%.18e"); 
  printf("]\n");

  gsl_matrix_complex_memcpy(A, m);
  gsl_eigen_herm(A, eval2, w1);

  printf("eval2 = [\n"); 
  gsl_vector_fprintf(stdout, eval2, "%.18e");
  printf("]\n");

  gsl_matrix_complex_memcpy(A, m);
  gsl_eigen_hermv(A, eval, evec, w2);

  printf("eval = [\n"); 
  gsl_vector_fprintf(stdout, eval, "%.18e");
  printf("]\n");

  printf("evec = [\n"); 
  gsl_matrix_complex_fprintf(stdout, evec, "%.18e");
  printf("]\n");
}