Rounding should be used on any integer division.
Otherwise e.g. on 15 values of 10 and a single value of 9 the result was still 9 and you have added an error of almost 1 unit.


Ahhh. OK I see now. However, this:

#define div_rnd(x,y)    ((x + (x >> 1)) / y)

does not round before the division. What you meant was:

#define div_rnd(x,y)    ((x + (y >> 1)) / y)

Jon