The last release (1.2.0) introduced a new algorithm to combine the matches found within a password which provided better results, but drastically slowed down estimations. I spent some time profiling where that extra time was going, and optimized most of that slowdown away.
