For calculating Collatz class records, how is the sieve constructed and used?

260 Views Asked by At

I'm looking at the class record algorithm here. I understand how some of the numbers (e.g. 8*k+5) can be skipped because they join with a lower number. But I don't understand exactly how the sieve of 2^25 elements that he refers to is constructed or used.

How exactly would one build the sieve and use it when calculating class records?