On Sunday, 17 January 2016 at 03:26:54 UTC, Andrei Alexandrescu
wrote:
On 1/16/16 9:37 PM, Timon Gehr wrote:
Ivan's analysis suggests that even something significantly
larger, like
n/log(n)² might work as an upper bound for k.
I'm not clear on how you got to that boundary. There are a few
implementation details to be minded as well (quickly and
carefully approximating the log etc). Could you please make a
commented pull request against my own? Thanks! -- Andrei
size_t.sizeof * 8 - bsr(n) should be good approximation for
sorting. --Ilya
