3 ms·
We have actually done something like that up to the 32bit n_max for sure (but maybe even 64bit?), without any number that would contradict the conjecture. But
by kaba0 3y ago
We have actually done something like that up to the 32bit n_max for sure (but maybe even 64bit?), without any number that would contradict the conjecture.
But yeah, it is not even trivial to say whether it has a bounded max memory, so in case of an arbitrary precision int type, it may not be LBA, but Turing?
- odyssey7 3y agoI guess I was thinking about indexing the visited numbers from a graph-traversal-algorithm-interview perspective instead of a CS theory one. We can actually do it in O(1) space complexity in exchange for higher time complexity. For i in range 1..n, compute n_i, the ith number in the hailstone iteration, and then continue the hailstone iteration n_(i+1)..n_max to see if n_i is equal to any of them. This takes us to quadratic time complexity, O(n * n_max), or constant complexity depending on how you look at it, but it only requires storing a single n_i at a time for cycle detection. But then again, if you actually loop for n_max iterations without halting by reaching 1 or crashing, then you had to reuse a number somewhere, so the explicit cycle detection isn’t really important.