Bits of integer

[Albert Chan]

This was a PM that I thought wothy of starting a thread: Number of Bits in a Integer

Albert Chan Wrote:Unless you have correct implementation of log2 (or binary logb), do this instead:

Quote:20 DEF FNB(N)=IP(LN(N+.5)/L2)+1 ! BITS OF INTEGER N

It seems it is harder than it look, even with "perfect" log2

>20 def FNB(N) = IP(LOG2(N))+1
>X = 2^35
>X, FNB(X-1), FNB(X), FNB(X+1)
 34359738368      36      36      36

For binary machine, code is trivial, bit_length(x) = logb(x) + 1

For decimal machine, we may be hit with rounding errors. Any ideas ?