vector of vector and computes

[mickey]

Hello,
I have a matrix of int (vector of vector); the problem is that I need to compute the mean of each column (the matrix can be 500 x200); so to do this I did:

foreach (iter_column) {
   float sum = 0;
   iter_line.setFirstLine;
   foreach(iter_line) {
        sum += iter_line[iter_coulmn]. value;   
   }
   mean = sum / nOfLine;
}

foreach (iter_column) {
   float sum = 0;
   iter_line.setFirstLine;
   foreach(iter_line) {
        sum += iter_line[iter_coulmn]. value;   
   }
   mean = sum / nOfLine;
}

To copy to clipboard, switch view to plain text mode 

The problem is that I have to do "two for" one inside other; I thought If is better to do something like to tie a vector<coulum> and do:

class Column {
public:
   vector<double> _value;
};
vector<Column> col; //to fill col properly
foreach (col) {
    int sum = std::accumulate(col.begin(), col.end(), 0);
    mean = sum / col.size();
}

class Column {
public:
   vector<double> _value;
};
vector<Column> col; //to fill col properly
foreach (col) {
    int sum = std::accumulate(col.begin(), col.end(), 0);
    mean = sum / col.size();
}

To copy to clipboard, switch view to plain text mode 

Will be the second way better/faster? Any other ways to improve?

[wysota]

It won't be faster, because std::accumulate will iterate the vector as well... What you can do is to calculate all columns at once and store results in a temporary vector. It won't be much faster though - you have to iterate all columns in all rows ending up with a complexity of O(n x m) anyway. The only benefit you might have is proper use of cpu cache. You could try making things faster by calculating more than one cell at once (by reading/summing all columns in one machine instruction), but you can have overflows then.