Copy and reuse dictionary cache

[fa1rid]

Simply adding multiprocessing to that step won't do the trick. The bottleneck is likely not the CPU but your hard drive from which the wordlist has to be read in full.

I'm using NVMe which is super fast (around 600MB/s). The CPU is saturated at 100% usage (1 core only).

If the NVMe is being fully utilized, then the hashcat would finish caching 20GB file in less than a minute, but this is not happening and the disk is not fully utilized.