View Single Post
Old 2019-04-06, 23:22   #11
VBCurtis's Avatar
Feb 2005
Riverside, CA

5,279 Posts

Originally Posted by RichD View Post
But who on earth could handle the post-processing?...
I think this particular candidate falls into a no-man's-land, where the lack of possible record factorization acts against a possible XSEDE proposal yet it's too big for any single machine we have access to. I'm willing to spend the $200 to upgrade my 20-core xeon to 128GB, which is likely enough to solve a matrix from a GNFS-215 or GNFS-220 project, but insufficient for RSA-232. Perhaps a pair of infiniband cards would allow me to solve RSA-232 personally in under a year, but that sounds more like a hope than a plan!

However, if we went after RSA-240, we might be able to get XSEDE to grant cluster time for the matrix. That should take on the order of 1000 core-years of sieving, way too much for forumites alone!
In the interests of gaining experience with big jobs, I've volunteered to postprocess a C206 for swellman that Greg is sieving on 16f; fivemack/Greg are doing the same for the C205 from Aliq 276. If we're serious as a group about doing some quixotic team-sieve, I suggest we find a GNFS in 210 to 218 range to team-factor, so that we can collectively iron out details of data management etc. If my time estimates are correct, we can knock one out with 50 to 100 core-years of sieve, and less than a machine-year of postprocessing; if they're not, I'd rather learn I'm wrong by 50 years than 500!

There are cloud-service machines with big memory that could be rented for this matrix, which would well and truly fit a publication of this factorization using "publicly available tools!"

If that goes well, we can decide whether to attack RSA-232 or RSA-240? In the meantime, there's no harm in polyselecting for either one. I have msieve aimed at RSA232 right now, and I'll start another thread for RSA-240 poly select sometime this month.
VBCurtis is online now   Reply With Quote