dynamic allocation/deallocation of 2D & 3D arrays

You can also allocate one array and compute individual indices. This requires fewer allocator calls and results in both less fragmentation and better cache use.

typedef struct {
  int a;
  int b;
  int* data;
} Int2d;

Int2d arr2d = { 2, 3 };
arr2d.data = malloc(arr2d.a * arr2d.b * sizeof *arr2d.data);

Now arr2d[r][c] becomes arr2d.data[r * arr2d.b + c]. Deallocation is a single free() away. As a bonus you’re sure to always keep your dynamic array sizes with you.

Extrapolating to 3d:

typedef struct {
  int a;
  int b;
  int c;
  int* data;
} Int3d;

Int3d arr3d = { 2, 3, 4 };
arr3d.data = malloc(arr3d.a * arr3d.b * arr3d.c * sizeof *arr3d.data);

//arr3d[r][c][d]
// becomes:
arr3d.data[r * (arr3d.b * arr3d.c) + c * arr3d.c + d];

You should encapsulate these index operations (and the (de-)allocations for that matter) in a separate function or macro.

(The names for r, c, and d could be better—I was going for row, column, and depth. While a, b, and c are the limits of their corresponding dimensions, you might prefer something like n1, n2, n3 there, or even use an array for them.)

Leave a Comment