[Solved] Allocate matrix of integer with c

NUMI does not have rows and columns, it is a pointer-to-pointer-to-int, and happens to be pointing at an allocated memory that has room for dim pointers-to-int, not dim * dim ints. This would be equivalent to treating as having been declared int* NUMI[dim]

A call

int* NUMI;
NUMI= malloc( dim*dim*sizeof(int) );

will allocate a dim by dim matrix of integers.

However, please note that with multi-dimensional arrays, say int example[a][b], the size of the allocated area is equivalent to int* example_ = malloc(a*b*sizeof(int)), and the compiler works out the conversion from multidimensional indices to a single-dimension index, i.e. example[c][d] -> example_[c*a+d]

So when you do

int* NUMI;
NUMI= malloc( dim*dim*sizeof(int) );
//...
NUMI[i][i]= value2;

The compiler doesn’t have the information required to convert from the multiple dimension to the single-dimension equivalent.

PROB is similar, pointing to a memory area with room for dim pointers-to-pointers-to-double, not dim * dim * dim doubles.
To get a dim cubed matrix, you’d need

double  *PROB;
PROB = (double *)malloc( dim*dim*dim*sizeof(double) );

and run into the same problem with multi-dimensional indices.

If dim is a compile-time constant, you can declare the multidimensional arrays directly without malloc

double PROB[dim][dim][dim];
int    NUMI[dim][dim];

The loop at the end of the main() should now work as expected.

If you must use malloc, either use malloc as described above:

NUMI= (int *)    malloc( dim*dim*sizeof(int) );
PROB = (double *)malloc( dim*dim*dim*sizeof(double) );

and modify the loop body to

PROB[(ACT * dim * dim) + (ACTSTART * dim ) + i] = value;
NUMI[i + dim * i]= value2;

Alternatively, call malloc in multiple loops as described by Alexey and Paolo.
With my solution, there is one malloc() call per variable, so each variable refer to a single contiguous memory area. With multiple malloc() calls in loops, you have multiple allocated memory areas that is unlikely to be contiguous.