![]() |
3 лет назад | |
---|---|---|
.. | ||
tests | 3 лет назад | |
Makefile.am | 5 лет назад | |
README.md | 3 лет назад | |
storage_number.c | 3 лет назад | |
storage_number.h | 3 лет назад |
Although 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 -167772150000000.0
to 167772150000000.0
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.