GNU Scientific Library determinant of complex array help

AI Thread Summary
The discussion centers on calculating the determinant of a complex 6x6 array using the GNU Scientific Library (GSL) in C. The user is facing compilation errors due to type incompatibility when attempting to use GSL functions with a double complex array. The key issue is that GSL requires the use of its own complex number types, necessitating a conversion from the user's existing complex array to a GSL-compatible format. The conversation emphasizes that while the user is reluctant to fully switch to GSL's complex types, doing so is essential for compatibility and functionality. Suggestions include allocating a GSL matrix and populating it with loops to ensure proper type alignment. The user acknowledges the need for this conversion and notes that this approach compiles successfully. An alternative suggestion is made to consider using LAPACK, which can work with standard Fortran arrays.
tosburn3
Messages
6
Reaction score
0
So I suck at programming, but I need to find the determinant of a complex 6x6 array using GSL in C (not GSL complex, complex.h complex). Here is what has failed so far starting with a 6x6 double complex array named mymatrix:

gsl_matrix_complex_view m = gsl_matrix_complex_view_array(mymatrix, 6, 6);
int s;
gsl_permutation *p = gsl_permutation_alloc(6);
gsl_linalg_complex_LU_decomp(&m.matrix, p, &s);
gsl_complex det2 = gsl_linalg_complex_LU_det(&m.matrix, s);
double complex det1 = GSL_REAL(det2)+I*GSL_IMAG(det2);

I refuse to convert my entire program to GSL's complex stuff if you think that would fix it. Here is the error message when compiling in gcc:

lorenz_odd_1.c: In function ‘det’:
lorenz_odd_1.c:122:1: warning: passing argument 1 of ‘gsl_matrix_complex_view_array’ from incompatible pointer type
/usr/local/include/gsl/gsl_matrix_complex_double.h:128:1: note: expected ‘double *’ but argument is of type ‘__complex__ double *’
 
Last edited:
Technology news on Phys.org
Welcome to PF, tosburn3! :smile:

To make any function of GSL work with complex numbers, you need to convert your complex numbers to GSL complex stuff.

Yeah... so I'm suggesting that one.
Sorry.
 
I like Serena said:
Welcome to PF, tosburn3! :smile:

To make any function of GSL work with complex numbers, you need to convert your complex numbers to GSL complex stuff.

Yeah... so I'm suggesting that one.
Sorry.

I don't care if I have to use GSL crap as intermediate steps. I just want to start with a regular complex array and end with a regular complex number. It was my understanding that:

gsl_matrix_complex_view m = gsl_matrix_complex_view_array(mymatrix, 6, 6);

Will take "mymatrix" and make a gsl_matrix_complex called "m", and then at the end I convert back to a regular complex number (quite inefficiently I might add).
 
It does not matter if it's inefficient.
In terms of performance, calculating a determinant is so much more computationally difficult, that a conversion of complex numbers matters not at all.
It's only a little more code.

But... what is your problem then?
 
FWIW here is the error message when compiling with gcc (updated original post for recent version):

lorenz_odd_1.c: In function ‘det’:
lorenz_odd_1.c:122:1: warning: passing argument 1 of ‘gsl_matrix_complex_view_array’ from incompatible pointer type
/usr/local/include/gsl/gsl_matrix_complex_double.h:128:1: note: expected ‘double *’ but argument is of type ‘__complex__ double *’
 
I like Serena said:
It does not matter if it's inefficient.
In terms of performance, calculating a determinant is so much more computationally difficult, that a conversion of complex numbers matters not at all.
It's only a little more code.

But... what is your problem then?

Also, the inefficiency was actually that I computed the determinant once for the real part then again for the imaginary part, which would in fact take a lot longer.
 
Oh okay, so your "mymatrix" is of the wrong type, apparently something like "double complex*".
How did you define "mymatrix"?

Anyway, I strongly recommend that you convert your mymatrix to the type gsl_matrix_complex_view_array() expects.
Just allocate the gsl matrix and fill it with 2 for-loops.
It really does not pay to try to make any shortcuts here.

Generally any shortcuts work against you.
If not now, then later.
 
I like Serena said:
Oh okay, so your "mymatrix" is of the wrong type, apparently something like "double complex*".
How did you define "mymatrix"?

Anyway, I strongly recommend that you convert your mymatrix to the type gsl_matrix_complex_view_array() expects.
Just allocate the gsl matrix and fill it with 2 for-loops.
It really does not pay to try to make any shortcuts here.

Generally any shortcuts work against you.
If not now, then later.

mymatrix is in fact a double complex. I will try that.
 
While I'm still not sure why the other way didn't work, looping through each component seems to work. At least it compiles...

Thanks!
 
  • #10
Good! :wink:
 
  • #11
Alternatively you could use LAPACK, which will work on an ordinary Fortran array.
 
Back
Top