View Single Post
Old 2018-07-30, 19:00   #25
danaj
 
"Dana Jacobsen"
Feb 2011
Bangkok, TH

16128 Posts
Default

Two Algorithms to Find Primes in Patterns

It isn't clear how efficient their code is for tuplets compared to mine. They ran with 150 cores and parallelize in the inner loop. I ran with 4 cores and parallelize at the outermost level for braindead simplicity (e.g. the sieve is serial but I run N ranges at a time). Extrapolating the time on my 4-thread Macbook to range and cores comes out to nearly the same time as they report, but that's a lot of extrapolation.

Their wheel is significantly larger than mine. I restrict mine based on space and to some extent speed (larger is not always faster, but it really depends on many factors including the depth and speed of primality testing, where I suspect my testing is faster than theirs).

My code just does clusters of {p,p+A,p+B,p+C,...} for any user entered A,B,C,.... Theirs also does things like {p,Ap+A',Bp+B',...} so they can look for Cunningham chains.
danaj is offline   Reply With Quote