valarray submatrix representation using gslice

How should a function, foo, be defined so that it will accept a valarray (that represents a MxN matrix), so that foo operates on a sub-matrix?

I've read enough to know that I need to use gslice to pick out the sub-matrix. (I don't want to use the mask or indirect because the original matrix and the sub-matrix will be huge.)

As an example of what foo will see..
Say, original valarray has a 35 elements representing a 5 row by 7 col matrix; and the sub-matrix starts in the 2nd row, 2nd column (i.e., offset 8), and whose dimensions are 3 rows by 4 cols.
foo should just print out the sub-matrix values.

Suppose the initial values of the valarray are just set to the index number (i.e., first element is 0). It looks like:

 0  1  2  3  4  5  6
 7  8  9 10 11 12 13
14 15 16 17 18 19 20
21 22 23 24 25 26 27
28 29 30 31 32 33 34

Select allOpen in new window

Then with start offset, S = 8, nRow = 3, nCol = 4, foo should be able to operate on the following sub-matrix:

 8  9 10 11
15 16 17 18
22 23 24 25

Select allOpen in new window

I started with this, but not sure how to use gslice and set up gslice params correctly.
foo should use this gslice, I think.

#include <iostream>
#include <valarray>
using namespace std;

template<class T>
void print(const valarray<T> &v) {
   for(size_t i = 0; i < v.size(); ++i) 
      cout << '\t' << v[i];
   cout << endl;
}

template<class T>
void init(valarray<T>& v)
{
   for(size_t i=0; i < v.size(); ++i)
   {
      v[i] = i;
   }
}

void foo()
{
}

int main()
{
   const int M=7, N=5;
   valarray<int> va(M*N);
   init(va);
   size_t sizes[2]   = {3, 4};
   size_t strides[2] = {7, 1};
   gslice gs;
   foo();
   print(va);
}

Select allOpen in new window

I was told not to use boost or my own class to handle the sub-matrix due to optimization plans.
Thanks.