$ 2^{bit_precision} = 10^{decimal_digits} $ $ bit_precision = log_{2}(10) {decimal_digits} $ $ decimal_digits = log_{10}(2) {bit_precision}$
Type | Sign | Exponent | Significand | Total | Exponent Bias | Bits Precision | Number of Decimal Digits | Wiki |
---|---|---|---|---|---|---|---|---|
f16 Half (IEEE 754-2008) | 1 | 5 | 10 (11) | 16 | 15 | 11 | ~3.3 | https://en.wikipedia.org/wiki/Half-precision_floating-point_format |
f32 Single | 1 | 8 | 23 (24) | 32 | 127 | 24 | ~7.2 | https://en.wikipedia.org/wiki/Single-precision_floating-point_format |