Question

I’m trying to understand some code that uses pointer arithmetic in a way I’m not used to. At one point in the code I encounter this:

complex<double> **P, *p_row, result=complex<double>(0,0);
P=new complex<double>*[n];
for(i=0;i<n;i++) P[i]=new complex<double>[n];

for(i=0,p_row=*P;i<n;i++,p_row+=n) result+=log(*(p_row+i));

If P is a matrix, this looks to me like adding the logarithms of the diagonal elements of P. But it turns out the last line above is not equivalent to

for(i=0;i<n;i++) result+=log(P[i][i]);

I’ve been searching for an explanation of what’s going on here but I can’t find it. Also, the code in question apparently gives the right result in the end (it’s part of a Monte Carlo). Any ideas?

Answer 1

The code you posted is wrong; it does invoke undefined behavior.

Eg. in the second iteration, p_row is (*P)+n. *P points to an array of size n, therefore, when the code reads *(p_row+i), it reads past the end of the array.

The last line seems to assume the matrix is stored in a single continuous array (eg. row major). However, that means P would be a complex<double>* initialized by P = new complex<double>[n*n];.