[Mono-devel-list] Mono and GSL

[Paolo Molaro]

On 03/24/03 bsuss_ca at yahoo.ca wrote:
> I've been playing around with getting Mono to work
> with the Gnu Scientific Library (GSL).  

Great!

> uncommenting the comment.  If you
> have any clue about how to get this to work, it would
> be much appreciated.
> 
> The complex number is a C struct that looks like:
> >>
> typedef struct
> {
>   double dat[2];
> } gsl_complex;
> <<

Please file a bug report in bugzilla.ximian.it with the test case,
this looks like an error while marshaling back a structure with a fixed
length array when it's a return type.

> 	public struct complex
> 	{
> 		[MarshalAs(UnmanagedType.ByValArray, SizeConst=2)]
> 		public double[] dat;
> 
> 		complex(double re, double im)
> 		{
> 			this.dat=new double[] {re,im};
> 		}
> 	}

Now, I would suggest implementing the mapping between gsl_complex and a
C# struct in this way, instead:

	public struct complex
	{
		public double real;
		public double imm;

		complex(double re, double im) {
			real = re;
			imm = im;
		}
	}

This way, the generated code is much faster (no need to do conversions
when marshaling) and it uses less memory (no need to allocate an array).
Note: I still get the wrong result from gsl_complex_sin(), but at least
it doesn crash:-)
I haven't investigated what the issue really is, since the same test is
executed correctly by the new jit, mini (due out in a few days):
$ mini gsl.exe 
1
5
3
3.85373687744141
The same program translated to C gives:
$ ./a.out
1
5
3
3.85374

lupus

-- 
-----------------------------------------------------------------
lupus at debian.org                                     debian/rules
lupus at ximian.com                             Monkeys do it better