netdata does all its calculations using
long double, it stores all values using
a custom-made 32-bit number.
This custom-made number can store in 29 bits values from
with a precision of 0.00001 (yes, it's a floating point number, meaning that higher integer values
have less decimal precision) and 3 bits for flags.
This provides an extremely optimized memory footprint with just 0.0001% max accuracy loss.