CFD Online Discussion Forums - data structure question

Hi
After writing a SIMPLE solver on unstructured collocated grid, I realized I made a bad design decision. The vertices, facets and cell lists are interlinked with pointers, and all the data for a cell is at one place. That means I have u,v,w and other scalar terms like T (temperature or whatever) for a cell at the same place. That is what good c++ code should do ,but now I suspect that it is better to have a field for each variable separately (like u[100],v[100] etc..) so that different operations can be done separately. When the scalar transport equation is solved for u, we don't need T for example, good for cache I suppose..
Before I go on doing that I want to know which method is being used by fastest cfd solvers.
I have the following

Code:

struct Field {

    union {

        struct {

            Vector U;

            Scalar p;

            Scalar T;

            Scalar pp;

        };

        Scalar f[6];

    };

};

and am going to change it in to the following for 100 cells

Code:

Vector U[100];

Scalar T[100];

...

thanks in advance
P.S : Is there a programming sub-forum where this kind of posts belong ?