On 05/20/2017 04:52 PM, Jon Degenhardt wrote:
On Thursday, 18 May 2017 at 15:14:17 UTC, Andrei Alexandrescu wrote:
The implementation is an improved version of what we now have in the D
standard library. I'll take up the task of updating phobos at a later
Very nice! Is this materially faster than what is currently in Phobos
(PR 4815)? That update was a substantial performance win by itself.
The code per the paper is faster for more data sets. For the NLP data
you tested on, speed should be about the same. -- Andrei