[nsi]

Dear CFD friends!

I am writing a CFD code solving 3D Euler equations (compressible, inviscid flow). The first problem is, I am not allowed to generate a big matrix to store the conserved variables, if I am having a 1000x1000x1000 Cartesian structured mesh.

Here is what I have tried.

(1) Using "new" to generate a dynamic array;

new.JPG

(2) Using "vector" to generate a dynamic array;

vector.JPG

(3) Generating a static array variable.

static.JPG

But none of them worked.

It makes me think that it is necessary that I change the data structure/organization. But anyway I need a matrix like this to store the conserved variables at each time step. Any idea will be much appreciated!!!

N. Si

[Peralta]

Have you calculated the required RAM to store an array of this size?

[praveen]

All of the methods you tried are bad ways to use multi-d array in c/c++.

One way is to store it as a one-d array and wrap the indexing (I haven't checked the index wrapping, but you see the general idea)

#include<iostream>
  
#define u(i,j,k)  u_[(i) + (j)*imax + (k)*imax*jmax]

using namespace std;

int main()
{
   const int imax = 1000;
   const int jmax = 1000;
   const int kmax = 1000;

   double* u_ = new double[imax*jmax*kmax];
   u(0,0,0) = 100.0;
   cout << u(0,0,0) << endl;
}

Another way with multi-d indexing is

#include<iostream>
using namespace std;

int main()
{
   const int imax = 1000;
   const int jmax = 1000;
   const int kmax = 1000;

   double (*u)[jmax][kmax];

   u = new double[imax][jmax][kmax];
   u[0][0][0] = 100.0;
   cout << u[0][0][0] << endl;
}

[Alisha]

Using vector is a good option. Additionally I would say, as you are into scientific computing, consider using some reliable array library. I personally like blitz++