View Single Post
Old 2020-10-22, 20:28   #418
Apr 2020

132 Posts

Originally Posted by VBCurtis View Post
I got 1440@15e7 done. I ran on 30 cores, which makes your curve count... staggering! Our combined effort so far is about 16% of a T60.
I'm using 128 cores - decided to put the CADO 2LP bound experiment on the backburner for now. By the looks of it mine are faster than yours?

I've shifted to B1=42e7; 4e8 uses k=6 and less memory, while 420M uses k=2 and more memory for a 1/3 larger B2 value. About 5-7% longer time per curve but 10% fewer curves needed for T60.
Well spotted! I'll switch over on the machines which have enough memory per core.
charybdis is offline   Reply With Quote