Tags: tylerchr/pwnedpass
Tags
Improve performance of pwngen Evidently using `binary.Write` is very slow? It has been replaced with direct use of `dff`. I found that this improved index generation throughput from ~250 MB/s to ~500 MB/s on a 2021 M1 MacBook Pro. This inefficiency was masked because the passwords database is distributed compressed with 7z and decompressing it was already the bottleneck. Using an uncompressed copy of the archive (or one recompressed with Zstandard) it is now possible to generate the index in about a minute.