Been doing some testing.  I have a CPU only setup/program that takes an input file of full BTC addresses and randomly searches user defined keyspace.

I can get 1 MKey/s with a list of 15 million addresses (i5-4690 running 4 threads).  I am curious, does/will your Keyhunt program handle that many addresses and if so, does they Key/s drop?

Also, I read your page and I have a suggestion for you.  You say that searching RIPE is 2x faster than full address; why not read the input file of full addresses, convert them to RIPE, and then do the search on RIPE? Then you only have to have the one user option...to search for full address while under the hood, the program converts to RIPE and searches for RIPE.  That's how my program works, C++ based off of VanitySearch (but CPU only).