Two fundamental problems:
1) You can't guarantee that successive calls to calloc( ) will produce memory segments that are adjacent to each other in memory. So the entire concept of using a single memcpy to copy such memory as one big block is non-conforming. If might happen to work if you are lucky, and then the next time the program will bomb because you weren't lucky. If you want to copy it all as one single block of memory, then you need to allocate it as such. E.g.,
dataDMA = (double **)calloc(sizeof(double *), m); // number of columns
dataDMA[0] = (double *)calloc(sizeof(double), n*m); // entire block of double memory
for (int i = 1; i < m; ++i) // <-- start loop indexing at 1
{
dataDMA[i] = dataDMA[i-1] + n; // next column is n elements over
}
Then, when you are done with this memory you only have two free( ) calls to make:
free(dataDMA[0]);
free(dataDMA);
2) The dataDMA pointer points to the column address array, not to the double data itself. So your memcpy source address is wrong. It should be this:
memcpy(mxGetPr(matDDMA), dataDMA[0], m*n *sizeof(double)); // don't need (void *) here
