Jul 27, 2022 · The binlog dataset is not publicly available, only the micro-benchmark is public. Spoiler alert: I'm going with FNV-1a 64-bit hash size and an "optimization": …

I also quickly tried the index approach in a very crude way, and it seems promising: I got from 30s on my benchmark (using the current multi-threaded code) to 17s (single-threaded).

Overall: adds RoboCopy task support for duplicate file copy detection. This is something very prevalent in some solutions I've been working on and would be a good win if we had it on the dupe detec...