(Free42) not a bug - COMB

[Werner]

(07-13-2020 12:03 PM)Paul Dale Wrote:  This looks like it would only address the specific example. Other cases would still overflow.

The 34S uses logΓ with a forced rounding to integer: \( ^nC_r = e^{logΓ(n) - logΓ(r) - logΓ(n - r)} \)

This risks cancellation of digits by subtraction.


Pauli

No, it doesn't. As long as the final result is less than 1e6145, it will return it - otherwise it overflows, as it should. There are hundreds of cases like this for which the current COMB overflows. It's nitpicking, perhaps - but to simulate an HP, you have to pay attention to detail :-)
Werner