calling with calllib a c function wich returns a pointer of pointers

After first easy test with calllib, I'm working on the next step, quite close of what I need in real situation. I worte a C function wich returns an pointer of pointer. It's something wich looks like a bit as a computation of a vandermonde matrix.

Here is the test_pp.c

#include <stdio.h>
#include <stdlib.h>
#include <math.h>
#include "test_pp.h"

double** almost_vandermonde( double* a, int na, int ns )
// na is the length of a
// ns is the number of lines we want in the matrix
{
    double** sortie = malloc( ns * sizeof(double) ) ;
    for (int i = 0 ; i < ns ; i++ )
    {
	sortie[i] = malloc( na * sizeof(double)) ;
	for (int j = 0 ; j < na ; j++ )
	{
	    sortie[i][j] = pow(a[j],i) ;
	}
    }
    return sortie ;
}

and the test_pp.h

#include <stdio.h>
#include <stdlib.h>
#include <math.h>

double** almost_vandermonde( double* a, int na, int ns ) ;

I compile it using gcc

gcc -c test_pp.c -o test_pp.o
gcc -o test_pp.so -shared test_pp.o

Now I try it in matlab

loadlibrary("test_pp","test_pp.h")
b = libpointer('doublePtr',[2 3 4 ]);
na = int32(3) ; 
ns = int32(4) ;
s = calllib("test_pp","almost_vandermonde",b,na,ns)

And now s is a libpointer of datatype doublePtr. I would like to have a pointer of pointers. If I try to access to the value in s, by setting a size, it seems that I can only have access to a small, basically the first array with only the one inside ( [2 3 4]^ 0 ).

[James Tursa]

Can you tell us what the real situation is? I am not a fan of creating an array of allocated pointers to mimic a 2D matrix. It is simpler to just allocate a single block of memory and address into that. I might propose you so this instead, but it may not be appropriate for what you are "really" doing, so I would like to know that first.

[Tommy Vasek]

You're right, I find a easier way for the real situation. Actually, I have a time serie (financial price), and my function compute three indicators (hurst, volatility, residual), so I have three new time series. Actually, I find easier with matlab to change my C code, so now, if my new time series have length L, instead of having an 2D array of shape (3,L), I have an array 1D of shape 3xL.