View Single Post
Old 2010-03-10, 01:08   #7
VBCurtis's Avatar
Feb 2005
Riverside, CA

22×7×132 Posts
Default Sieve speed and efficiency

To give an idea of how large a project this is-
There are roughly 3.7M candidates in the sieve. At 20T, my stock Core2-1866 (64 bit ubuntu) takes about 150s to find a factor. An LLR test of 1.0M on this machine takes ~1500 sec. If every k started at 1.0M in the sieve, this would imply a 200T depth before LLR is equal to factoring. Luckily, the 1.0-1.1M range is only half the k's, meaning half the sieve depth in principle.

This means 100T should be our target to begin LLR at 1.0M, and 150T or more for 1.1 to 1.2M. This is not the optimal sieve depth, where optimal means the depth that produces the shortest total project length to test all candidates. Optimal depth is so large it's depressing- 160T to get started, 320T at 1.1M, 450T or so at 1.2M. By 1.5M, we "should" be at 650T!!

This is what Cruelty referred to when he said we had to stop including k's somewhere.

Where is the blistering sheep when we need him?? Sheep, I hope you are healthy and working somewhere- you disappeared after discussing some life and wellness challenges with me.

Last fiddled with by VBCurtis on 2010-03-10 at 01:18
VBCurtis is offline   Reply With Quote