zlacker

TFA neglected to mention that floats have an "infinity" value, quite a bit larger than any 64 bit busy-beaver.

(And before anyone says it's not a number, call `isnan` with an infinity and get back to me :)

>>repsil+(OP)
Correct.

>>IEEE75+Z2
Username checks out. There is an other article about fp32 vs fp16 the front page for you as well!

>>repsil+(OP)
Fixed:-)

>>repsil+(OP)
I don't believe IEEE 754 specifies which infinities it encodes as its positive and negative infinity. I'd tend to treat it as the surreal equivalence class {0,1,2…|}, but it might be any of the others.

>>SAI_Pe+9c
IEEE 754 is very clear that the infinities are the endpoints of the extended real line (and hence also the extended integers, which matches your assumption).

>>stephe+Se
Thanks, clearly it's been too long since I read the spec.