2 ms·
A small minority, but apparently it includes all the floats you’re likely to use. It seems the insight is that you only need 8 bits of exponent in most cases. (
by skybrian 2y ago
A small minority, but apparently it includes all the floats you’re likely to use. It seems the insight is that you only need 8 bits of exponent in most cases. (And single-precision floating point only has 8 bits of exponent.)
Most double-precision floats are never used because they have high exponents.
- pansa2 2y ago> A small minority, but apparently it includes all the floats you’re likely to use. Sorry, I meant a small minority need to be boxed - all the floats you're likely to use can remain unboxed.
- bee_rider 2y agoGot take, but a double with only 8 bits of exponent actually seems kind of nice, you get the extra precision but it can be cast down to a single and you only lose precision; you don’t have values that are outside the range of singles.