mersenneforum.org  

Go Back   mersenneforum.org > Other Stuff > Archived Projects > 3*2^n-1 Search

 
 
Thread Tools
Old 2006-02-10, 15:13   #1
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

2·32·5·43 Posts
Default Jumping to 1 million digit LLR tests

In about month from now Thomas and I will have finished the sieving for 321search to the optimal level of division of 236 trillion. This leave only LLR work from "n" = 2.5-5 million -- That's 0.75-1.5 million digits.

The purpose of this poll is to guage interest in jumping from our current testing window to tests that have at least 1 million decimal digits.

Tests at 1 million digits only take half as long as they currently do. So an average 321-LLR file will increase from 5 days to 8 days. (This will be going upto 6 days soon anyway)

The primes are rarer at 1 million digits but with sufficient cruching and a little bit of luck we could find something huge in a reasonable amount of time.
paulunderwood is online now  
Old 2006-02-14, 12:56   #2
fatphil
 
fatphil's Avatar
 
May 2003

E716 Posts
Default

For these continuous sequences, I've never liked the idea of leaving gaps.
However, I don't contribute to 321, so my view is that of an outsider.
fatphil is offline  
Old 2006-02-14, 19:50   #3
Thomas11
 
Thomas11's Avatar
 
Feb 2003

22×32×53 Posts
Default

I agree with Phil: Eventually we should not leave any gaps.

Nevertheless it could be attractive to some of our testers -- in particular those with the ultra-fast P4 machines -- to do some tests on million-digits candidates and having the chance of finding the very first non-Mersenne Riesel prime really soon.

Actually, we already have quite a bunch of LLR blocks on the "choose your work" list (well, currently only a few are left from our current range...) and let the people decide by themselves which block they're choosing. So, why not add at least a few million-digit tests to that reservations list and see, if someone takes them?

And, just in case that no-one else will do sub-million-digit tests anymore, I would still continue to close the gap from the lower end...
Thomas11 is offline  
Old 2006-03-29, 00:21   #4
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

500010 Posts
Default

How many years do you expect the 2.5-5 million range for n to last the group?

These searches are mildly frustrating to me, in that the efficiency-maximizing person within me wants to sieve higher than I am likely to test, so I don't regret not sieving higher n; then I find myself compelled to LLR the entire range, so I don't "waste a sieve". Do you feel like when 5 million is reached, you'll be tempted to move to a different k? If not, how large an n-range would you contemplate for a new sieve?

-Curtis
VBCurtis is offline  
Old 2006-03-29, 00:29   #5
VBCurtis
 
VBCurtis's Avatar
 
"Curtis"
Feb 2005
Riverside, CA

23×54 Posts
Default

OK, I checked the project status, and your forecast of n=3M at end of 2006 seems feasible. With the growth in CPU speeds, 4M at end of 2007 is a ballpark, 5 M end 2008/sometime in 2009 is also reasonable.
That allows for ~3 years of sieving to get 5M-? ready. Hrmmm..
-Curtis
VBCurtis is offline  
Old 2006-03-29, 01:18   #6
paulunderwood
 
paulunderwood's Avatar
 
Sep 2002
Database er0rr

2×32×5×43 Posts
Default

Quote:
Originally Posted by VBCurtis
How many years do you expect the 2.5-5 million range for n to last the group?
I can anwser this partially using rough figures. We have 134137 candidates above 2.7M which on average take 25 32bit-Athlon-1GHz hours to LLR test and we still have about 6000 individual candidate primes left below 2.7M to test yet. So roughly:

140000*25/(24*365) ~ 400 32-bit Athlon 1GHz years.

Note we have faster computers than 1GHz Athlon's in more way than one with respect to LLR.

If we have 40 GHz working on 321 then it'll take about 10 years. I guess we've got more like 60GHz at the moment...

Quote:
These searches are mildly frustrating to me, in that the efficiency-maximizing person within me wants to sieve higher than I am likely to test, so I don't regret not sieving higher n; then I find myself compelled to LLR the entire range, so I don't "waste a sieve". Do you feel like when 5 million is reached, you'll be tempted to move to a different k? If not, how large an n-range would you contemplate for a new sieve?
Sieving is a small but significant part of the process -- about 5-10%. For us the primary goal is to find a prime with more than a million digits. We have sieved deeply enough for it to become more efficient use to use LLR on the average candidate size of 4 million bits for our range with a maximum of 5 million bits. Who knows? our primary goal might be satiated next year. or never

Quote:
OK, I checked the project status, and your forecast of n=3M at end of 2006 seems feasible. With the growth in CPU speeds, 4M at end of 2007 is a ballpark, 5 M end 2008/sometime in 2009 is also reasonable.
That allows for ~3 years of sieving to get 5M-? ready. Hrmmm..
See my comment above. At the current rate I'd say we should just make 3M by the end of this year. 4M would take about 3 further years and then about 5-6 more to 5M. But by then we will have much faster computers etc... but the I sincerely hope that I'll not have to continue the organization beyond 5M

Last fiddled with by paulunderwood on 2006-03-29 at 04:28
paulunderwood is online now  
 

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Million digit moonshot MooMoo2 Twin Prime Search 9 2017-12-23 17:36
711 LL-tests in the 100 million digits range? ET_ PrimeNet 9 2010-03-28 20:21
Jumping to 1 million digit LLR tests paulunderwood 3*2^n-1 Search 3 2006-11-14 14:03
k = 2 thru 31 Ten Million Digit numbers TTn 15k Search 4 2004-08-21 18:20
100 MILLION DIGIT NUMBER lpmurray Software 8 2004-05-31 19:22

All times are UTC. The time now is 10:31.


Sun Oct 24 10:31:38 UTC 2021 up 93 days, 5 hrs, 0 users, load averages: 1.23, 1.13, 0.99

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2021, Jelsoft Enterprises Ltd.

This forum has received and complied with 0 (zero) government requests for information.

Permission is granted to copy, distribute and/or modify this document under the terms of the GNU Free Documentation License, Version 1.2 or any later version published by the Free Software Foundation.
A copy of the license is included in the FAQ.