Re: Mark1 - pollard rho implementation (38 minutes for 80 bits solving on CPU)
If this is IDLP you don't need to store full keys, just distance from base start. This alone saves up to 20 bytes. However this works only if the max distance is ensured to never be reached.
Also I'm willing to bet an arm and leg that mmap-ing the DP file will lose big time against an actual database file, when there are lots of DPs, since they'll be spread around the disk, I wonder what your OS will have to say about that, when seeking around and pages start to swap in and out (besides all running apps stalling due to low free memory)
But seems that there's some kind of competition on these forums on who can reinvent a rounder wheel.
You're right about the distance-only storage optimization for DLP.
This implementation isn’t mine. I’m just analyzing the code. I haven't even started it in my OS. This is all theoretical. If the original author wants to optimize further, they’d need to adopt delta encoding (storing walk distances, not full keys), replace mmap with a proper DB for large-scale DP storage, and tighten memory limits to avoid thrashing.
These changes require significant redesign. Until then, it’s stuck with these limitations.
