Learning Forum

[Geoff]

I have writen a compiled UDF in C. I have several static arrays of fixed length for storing boundary ID's and other model parameters. When I define the array dimensions directly as below the code runs fine.

static int boundary_IDs[7] = {30,4,36,37,34,5,6};

When I define the array dimensions using a int variable:

const int N_IDs = 7;
static int boundary_IDs[N_IDs] = {30,4,36,37,34,5,6};

The code compiles but when I run the simulation I get a memory fault and fluent crashes when it tries to access values from the aray within my DEFINE_... macros. Furthermore the memmory fault only ocurs on some machines; the code runs fine on single compute nodes of larger HPC clusters (Linux based) but the problems occur when I run the code on a smaller Linux based server where I do not have access to a full compute node or when I try to test the code on a Windows desktop. 

I have also tried defining N_IDs as a constant :

#define N_IDs 7

but that did not seem to make any difference.

[Rob]

I can't see const int used in any of the examples so wonder if it's an oddity of C, C++, C# etc that will pass the compiler step as it's not technically an illegal statement but may not be recognised by all node or OS configurations. Does integer N_IDs = 7 work?

The #define approach may not store N_IDs as an integer? 

[Geoff]

Part of the problem must have been the loop over the boundary ID array inside my DEFINE_ macro. I changed from a do-while loop to a for loop with a break statement inside and changed the definition of N_IDs back to #define. Now the code appears to run on all plateforms... I have no clue why this change fixed things but at least it works now and that's what matters.