I'm making a 3D array in this way:
#define P 1000 int
SPACE3D[P*2][P*2][P*2];
but I get the following compile error:
Error 1 error C2148: total size of array must not exceed 0x7fffffff bytes
Can i do something to avoid this error?
I tried also to initialize the 3D array like this:
int SPACE3D[P*2][P*2][P*2]={0};
In this way I don't have building error, but the program craches.
Any suggestions?
You need to create the array on the heap, using new or malloc(), not on the stack as you are doing. For example:
int* SPACE3D = new int[P*2*P*2*P*2];
Now you'll need to index into it using arithmetic. If you prefer a more natural, and perhaps safer, interface, there are plenty of matrix classes available outside the standard library. Please do not use a vector-of-vector-of-vectors, though, as that is very inefficient if your matrix is actually rectangular (i.e. not jagged/sparse).
You should allocate large chunks of memory on the heap instead of the stack:
int array[2*P][2*P][2*P]; // placed on the stack
int* array = new int[8*P*P*P]; // placed on the heap
vector<int> array(8*P*P*P); // vector uses heap storage
On the stack you are basically limited to some megabytes of memory, while on the heap you have a good share of you computer RAM (e.g. several GB) available.
Just keep in mind: a 3D array of 1000*1000*1000 ints requires 4 GB of memory!
Additionally I would suggest to build a small class for your 3D array:
template<typename K>
class Array3D
{
public:
Array3D(int size) // beware, data is not initialized!
: size_(size), data_(size*size*size) {}
int size() { return size_; }
size_t index(int x, int y, int z) const {
return x + size_*(y + size_*z);
}
K& operator(int x, int y, int z) {
return data_[index(x,y,z)];
}
private:
size_t size_;
vector<K> data_;
};
You can use it like that:
Array3D arr(1000);
arr(12,43,745) = 15;
This snippet is missing a lot of useful features, but it shows the basic principles of how to build a 3D array. Either build upon something similar, or use one of the libraries out there.
In my opinion, in C++ I would always use a data type for arrays (1D, 2D, ...) to store the array memory and the array size together and to provide at least some kind of boundary information.
Lets get at this problem from a non-technical side: as you named your array "space" I will assume you want to depict space. A whole lot of nothingness and a few planets. Maybe a few spaceships. Let it be a lot of planets and a lot of spaceships. At 10 planets with 10 ships each, we are at 110 units.
With your array, you describe 1000x1000x1000 units. Most of them empty. Nothing.
In real life, you don't do it that way either. If someone asks you where the elephants are in your city, you will probably say "there are 4 in the zoo". You will not go "there's none in my room, there's none in the next room, there's none in the room after that, there's none in the room above mine, there is none in the room next to the one above mine, there is none in the kitchen, there is none on the street...".
You mention where elephants are, which implies that there are none in the other places. Doing otherwise would be an enormous waste of time and paper.
So that's the way to go in your program. Make a list of stuff that exists. If you have nothing in your list of existing things that's at [1, 56, 784], then this implies there is nothing. That way, as long as you have a list of existing spaceships, there is no limit to your universe. Not even 1000x1000x1000.
Allocating a 3d array that way allocates it on the stack, and it looks like it's too large.
To see how to allocate it on the heap, see this answer: C++ Multi-dimensional Arrays on the Heap
Specifically, in your case:
#define P 1000
typedef int SPACE2D_t[P*2][P*2];
SPACE2D_t* SPACE3D = new SPACE2D_t[P*2];
SPACE3D[/* from 0 to P*2-1 */][/* from 0 to P*2-1 */][/* from 0 to P*2-1 */] = 42;
delete [] SPACE3D;
