What are your CRUS plans? - Page 2

gd_barnes · 2010-10-21, 19:37

Quote:

Originally Posted by rogue

I don't understand. Are you saying that there are bases only tested to n=5000?

I have personally tested all untested bases with ck<=10K to n=1000. For those that have had <=35 k's remaining at n=1000, I've tested all of those to n=5000. That is how I am working my way up the bases with x number of k's remaining at n=5K; testing them to n=25K. I'm currently working on the 9 & 10 k'ers. This took only a moderate amount of CPU time. Taking them all to n=25K is what takes a lot of time. My idea is to mostly stabilize the # of bases on the proven/1/2/3 k's remaining list. In the future, for the most part, new bases on that list will have to come from current bases with 4, 5, or 6 k's remaining.

rogue · 2010-10-21, 20:28

Quote:

Originally Posted by gd_barnes

I have personally tested all untested bases with ck<=10K to n=1000. For those that have had <=35 k's remaining at n=1000, I've tested all of those to n=5000. That is how I am working my way up the bases with x number of k's remaining at n=5K; testing them to n=25K. I'm currently working on the 9 & 10 k'ers. This took only a moderate amount of CPU time. Taking them all to n=25K is what takes a lot of time. My idea is to mostly stabilize the # of bases on the proven/1/2/3 k's remaining list. In the future, for the most part, new bases on that list will have to come from current bases with 4, 5, or 6 k's remaining.

IIUC, you have results that have not been posted because there are a number of conjectures for conjectured k < 10000 that have not been started (as far as the rest of us know). I presume then that many that you haven't reserved have more than 10 k's remaining at n=5000. Chances are that you take most (or all) conjectures with k < 1000 up to n = 25000 before I get to them, which I am OK with.

We'll just need modify the script that generates the condensed table to use different colors for conjectures with 4, 5, or 6 k's remaining.

Once we get those conjectures to n=25000, there will be many new opportunities to for participants to get primes into the Top 5000. My problem (like yours) is that there is too much little stuff to do (finding small primes) before I can focus on that.

gd_barnes · 2010-10-23, 05:51

To clarify: I do not consider a base "displayable" unless it has been tested to at least n=10K and so I wouldn't bother to reserve any bases that I've only tested to n=1000 or 5000 unless I planned to take it to n=25K within a couple of weeks. Hence they are still considered "untested" bases unless someone (including myself) have reserved them. I do use my tests to sporadically doublecheck others k's remaining, especially on more complex bases. I don't even keep any primes for the bases that I've tested to n=1000 and many of them I don't plan to reserve in the future. A lot of the effort was just to find some high-weight bases and to know only the # of k's remaining. It takes so little time to just rerun the starting bases script to n=2500 or n=5000 if I want to actually reserve bases only tested to n=1000 so far.

Everything that I've tested to n=1000 or n=5000 takes little CPU time. Most of them take < 3-6 hours CPU time to get that far. As a matter of fact, I've done the entire effort on 2 slow cores of a slow 1.6 Ghz dual-core laptop from 2007. It's only good for very low tests. I just kept it running almost constantly for a little over a month.

No, I'd rather not expand the proven/1k/2k/3k remaining list to include 4/5/6 k's remaining. 1, 2, and 3 k's remaining is enough bases. If people want to find more of such bases, they can test (already tested) existing bases with 4, 5, and 6 k's remaining to a higher search limit. That would be the easiest way to go.

I'm about done with all of the bases that have 9 k's remaining at n=5000 and have started on a couple of bases with 10 k's remaining. Several have 4 k's remaining at n=25K but none have <= 3 k's remaining at this point. There may be just a few more untested ones that can get to <= 3 k's remaining at n=25K but there are unlikely to be much more (untested bases that is). That has been the objective of my effort. A side benefit of it has been that it has virtually completely stabilized the 1k remaining list and so has allowed Ian and a couple of others to test all bases that have 1k remaining at n=25K to n=100K without having to come back around again after more new bases are tested. He is able to add some more 1kers by testing 2kers (and some 3/4kers in the future) but that's about the only way to add new 1kers at this point.

I admit it will be an administrative relief to have most of the "low lying fruit" completed.

Gary

rogue · 2011-01-01, 18:41

Quote:

Originally Posted by rogue

Currently: R928 to n=25000
Currently: Retests for b > 100, both Riesel and Sierpinski

This is what I plan to work on in the future:

Near Future: Retests for Generalized Woodalls for n < 4100
Near Future: Puzzle 10 (from Steven Harvey's website)
Future: S63, finish the team drive to get the conjecture to n=10000
Future: Get R243, S42, S108 to n=25000
Future: Start all remaining conjectures for conjectured k < 1000 and bring them to n=25000

In the past 2 1/2 months I have accomplished quite a bit.

I finished the items that I had started at the time and finished the retest of the GWs for n < 4100. With my discovery (to be double-checked) for the counter-example to Puzzle 10, I am now free to start some new things.

Most of my available cores had started on S63 back in November. I hope to finish all of those ranges this month which should bring the S63 drive to completion (presuming kar_bon submits his results).

I see that someone else has take S42 to n=25000, but the other two remain. I won't take those on until S63 is done. I do have a couple of free cores to take on the few remaining conjectures (with conjectured k < 1000) that haven't been started.

I have thought about taking S63 to n=25000, but realized that it would be a significant effort. A "back of the hand" estimate was about 150 GHz CPU years.

MyDogBuster · 2011-01-02, 05:19

I've completed all the 1k conjectures to n=100K. Still doing the 2ker's, with a projected finish of Oct.

I have my complete farm on CRUS. Not searching for Top 5000 at the moment.

Dabbling in starting the remaining conjectures with ck<10K. Also working on bases < 250 with k's not tested to at least n=50K.

Xentar · 2011-01-03, 19:08

I finally build a new machine last week.. now I hope to hit S17 this year

By the way: I wish all of you a happy new year.

Puzzle-Peter · 2011-02-12, 08:30

I'm mainly trying to have fun which is all about not getting bored. So I'll go on as I started, trying to finish some 1kers, starting new bases etc. The only 'real' goal is trying to prove one conjecture from scratch. This means I'll stay on R698 quite a bit longer...

gd_barnes · 2011-02-13, 09:35

Quote:

Originally Posted by Puzzle-Peter

I'm mainly trying to have fun which is all about not getting bored. So I'll go on as I started, trying to finish some 1kers, starting new bases etc. The only 'real' goal is trying to prove one conjecture from scratch. This means I'll stay on R698 quite a bit longer...

Unfortunately you came to the project just a few months too late to have a reasonable chance of proving any base starting from scratch. Between Mark (rogue), Ian, and me, we've tested all of the lower CK bases. I mainly did the ones that I knew from low-limit testing could possibly get down to 1, 2, or 3 k's remaining by n=25K and Mark worked his way up the CK list. Ian took a liking to bases 1000<CK<10000 for quite a while. I did what I did to stabilize the 1k & 2k bases that Ian had been searching to n=100K in order to maximize our number of "easy" proofs and to get a more stabilized list of proven/1k/2k/3k bases.

Don't kill yourself trying to prove R698. You were very fortunate to prime 6 of the 27 (22%) k's remaining at n=25K. For a lower-weight base such as this, more typical is a ~15% reduction for every doubleing of the n-range. On the search for n=50K-100K, which will take 3-4 times as long, you'll be doing well if you find 4 primes (~20%), which would still leave you with 17 k's at a high test limit on a high base. Like most bases here, it is unlikely to be proven in any of our lifetimes. What I tell people is that if a base > 32 has more than about 10 k's remaining at n=25K, it is unlikely to be proven in most of our natural lives. Believe-it-or-not! The high bases are just plain tough.

Although a small base so it doesn't quite fit the above, Riesel base 6 has been the best one we've had that has defied the conventional wisdom on this. It has primed 17 of its remaining 19 k's at n=25K so far. It has a very good chance of proof in the next few years but one of the remaining 2 k's is extremely low weight and may not be found prime for decades. You just never know.

Gary

Puzzle-Peter · 2011-02-13, 16:11

Quote:

Originally Posted by gd_barnes

Don't kill yourself trying to prove R698. You were very fortunate to prime 6 of the 27 (22%) k's remaining at n=25K. For a lower-weight base such as this, more typical is a ~15% reduction for every doubleing of the n-range. On the search for n=50K-100K, which will take 3-4 times as long, you'll be doing well if you find 4 primes (~20%), which would still leave you with 17 k's at a high test limit on a high base. Like most bases here, it is unlikely to be proven in any of our lifetimes. What I tell people is that if a base > 32 has more than about 10 k's remaining at n=25K, it is unlikely to be proven in most of our natural lives. Believe-it-or-not! The high bases are just plain tough.

Thanks for this calculation! Since I don't want to kill myself, I'll just take R698 high enough to become bored, then pass the sieve file to whoever likes to go on with it.

Speaking of sieve files - when the reservation pages don't say "more sieving needed" does it mean LLRing can start immediately or is it just unknown if the sieve limit was high enough? I suppose most of them are leftovers when people sieved a bigger range than they cared to LLR so maybe sieving depth was optimal for the part that has been tested but not for the rest?

Peter

Mathew · 2011-02-13, 16:54

Most of my sieve files fall into your 3^rd question. I would reserve to n=25K or 50K but sieve to n=100K.

Flatlander · 2011-02-13, 17:09

If they were sieved by me then they should be optimal. You would need to search for my posts on CRUS from a few months ago. (Not too many.

)
iirc These were 25k-50k or 25k-100k, so I'm not sure if they would still be untested.

2010-10-23, 05:51	#14
gd_barnes May 2007 Kansas; USA 101·103 Posts	To clarify: I do not consider a base "displayable" unless it has been tested to at least n=10K and so I wouldn't bother to reserve any bases that I've only tested to n=1000 or 5000 unless I planned to take it to n=25K within a couple of weeks. Hence they are still considered "untested" bases unless someone (including myself) have reserved them. I do use my tests to sporadically doublecheck others k's remaining, especially on more complex bases. I don't even keep any primes for the bases that I've tested to n=1000 and many of them I don't plan to reserve in the future. A lot of the effort was just to find some high-weight bases and to know only the # of k's remaining. It takes so little time to just rerun the starting bases script to n=2500 or n=5000 if I want to actually reserve bases only tested to n=1000 so far. Everything that I've tested to n=1000 or n=5000 takes little CPU time. Most of them take < 3-6 hours CPU time to get that far. As a matter of fact, I've done the entire effort on 2 slow cores of a slow 1.6 Ghz dual-core laptop from 2007. It's only good for very low tests. I just kept it running almost constantly for a little over a month. No, I'd rather not expand the proven/1k/2k/3k remaining list to include 4/5/6 k's remaining. 1, 2, and 3 k's remaining is enough bases. If people want to find more of such bases, they can test (already tested) existing bases with 4, 5, and 6 k's remaining to a higher search limit. That would be the easiest way to go. I'm about done with all of the bases that have 9 k's remaining at n=5000 and have started on a couple of bases with 10 k's remaining. Several have 4 k's remaining at n=25K but none have <= 3 k's remaining at this point. There may be just a few more untested ones that can get to <= 3 k's remaining at n=25K but there are unlikely to be much more (untested bases that is). That has been the objective of my effort. A side benefit of it has been that it has virtually completely stabilized the 1k remaining list and so has allowed Ian and a couple of others to test all bases that have 1k remaining at n=25K to n=100K without having to come back around again after more new bases are tested. He is able to add some more 1kers by testing 2kers (and some 3/4kers in the future) but that's about the only way to add new 1kers at this point. I admit it will be an administrative relief to have most of the "low lying fruit" completed. Gary Last fiddled with by gd_barnes on 2010-10-23 at 06:04

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Plans for the end of the world	Oddball	Lounge	4	2011-04-18 04:06
how high will CRUS go	Mini-Geek	Conjectures 'R Us	1	2010-11-08 20:50
Further Plans	Kosmaj	Riesel Prime Search	6	2009-05-20 01:27
Further Plans	Kosmaj	Riesel Prime Search	6	2006-09-29 22:32
64 bit plans	pyrodave	Software	17	2004-06-05 12:27

2011-01-02, 05:19	#16
MyDogBuster May 2008 Wilmington, DE 2²·23·31 Posts	I've completed all the 1k conjectures to n=100K. Still doing the 2ker's, with a projected finish of Oct. I have my complete farm on CRUS. Not searching for Top 5000 at the moment. Dabbling in starting the remaining conjectures with ck<10K. Also working on bases < 250 with k's not tested to at least n=50K.

2011-01-03, 19:08	#17
Xentar Sep 2006 11×17 Posts	I finally build a new machine last week.. now I hope to hit S17 this year By the way: I wish all of you a happy new year.

2011-02-12, 08:30	#18
Puzzle-Peter Jun 2009 2²×3²×19 Posts	I'm mainly trying to have fun which is all about not getting bored. So I'll go on as I started, trying to finish some 1kers, starting new bases etc. The only 'real' goal is trying to prove one conjecture from scratch. This means I'll stay on R698 quite a bit longer...

2011-02-13, 16:54	#21
Mathew Nov 2009 2·5²·7 Posts	Most of my sieve files fall into your 3^rd question. I would reserve to n=25K or 50K but sieve to n=100K.

2011-02-13, 17:09	#22
Flatlander I quite division it "Chris" Feb 2005 England 31·67 Posts	If they were sieved by me then they should be optimal. You would need to search for my posts on CRUS from a few months ago. (Not too many. ) iirc These were 25k-50k or 25k-100k, so I'm not sure if they would still be untested.