Is a safe accumulator really this complicated?
- by Martin
I'm trying to write an accumulator that is well behaved given unconstrained inputs. This seems to not be trivial and requires some pretty strict planning. Is it really this hard?
int naive_accumulator(unsigned int max,
unsigned int *accumulator,
unsigned int amount) {
if(*accumulator + amount >= max) return 1; // could overflow
*accumulator += max; // could overflow
return 0;
}
int safe_accumulator(unsigned int max,
unsigned int *accumulator,
unsigned int amount) {
// if amount >= max, then certainly *accumulator + amount >= max
if(amount >= max) {
return 1;
}
// based on the comparison above, max - amount is defined
// but *accumulator + amount might not be
if(*accumulator >= max - amount) {
return 1;
}
// based on the comparison above, *accumulator + amount is defined
// and *accumulator + amount < max
*accumulator += amount;
return 0;
}
