|
|
|
|
IEEE NUMBERS
Fortran | C (typical) | Bytes | Mant. Bits | Exp. Bits | Sig. Digits | Range |
CHARACTER | char | 1 | 7 | - | 2.11 | 127 |
INTEGER*2 | short int | 2 | 15 | - | 4.52 | 32,767 |
INTEGER*4 | long int | 4 | 31 | - | 9.33 | 2.15E+09 |
INTEGER*8 | long long int | 8 | 63 | - | 18.96 | 9.22E+18 |
REAL*4 | float | 4 | (24) | 8 | 7.22 | 1.0E±38 |
REAL*8 (D.P.) | double | 8 | (53) | 11 | 15.95 | 1.0E±308 |
- | long double | 10 | 64 | 15 | 19.27 | 1.0E±4931 |
REAL*16 | - | 16 | 112 | 15 | 33.72 | 1.0E±4931 |
Notes:
- Where the number of mantissa bits is in parentheses, the number includes an implied initial 1 bit. The number of bits actually stored is one less than this number.
- The number of significant digits corresponding to b (mantissa) bits is b / log2 10 = 0.30103 b.
- The range of a floating-point number is approximately 22x-1, where x is the number of exponent bits.
Ranges of floating-point numbers shown in the table are approximate.
- Each number also uses one bit for the sign of the number, so the total number of bits used is (# mantissa bits) + (# exponents bits) + (1 bit for sign).
The exponent is biased, so it does not contain an explicit sign bit.
Contact Information
I may be contacted at:
|
|