mersenneforum.org

mersenneforum.org (https://www.mersenneforum.org/index.php)
-   Riesel Prime Search (https://www.mersenneforum.org/forumdisplay.php?f=59)
-   -   Low weight stats page. (https://www.mersenneforum.org/showthread.php?t=2934)

Thomas11 2005-09-22 08:33

This is just a status update.
The following 23 k have been testet to n=2M now. No new primes found so far...

19370947
59910449
80857169
143316643
162405629
175437131
189030223
203012861
209826493
224371169
243163663
245265883
248690527
260213857
265831619
276278983
290851087
298095191
300871183
308120317
308141737
315940139
326840893

-- Thomas :sleep:

VBCurtis 2005-09-22 18:18

Thanks, Thomas-- I was really hoping you'd weigh in with your research soon. I'll try to get those updates on the stats page soon. How far did you sieve these very low-weight candidates? I have two numbers similar in weight to the candidates you chose; it seems 4T is sufficient (plenty, even) for running to 2 million. Did you sieve deeper, and why?

Kosmaj:
I noticed Joss' work on 403993 after posting that reservation, so I grabbed the sieve file and plan to finish to 500k,which I admit isn't the best time trial for my suggestion. That said, my P3-550 has sieved from 35G to 110G in 20 hrs, so sieving for 1 day on a normal machine is sufficient to prepare LLRing to 500k; I'm still removing candidates at nearly one per hour, so I'll sieve to 150 or 200G before LLRing the 403-500k range.

Citrix 2005-09-23 01:19

I guess you guys are right, Why 100k? I just thought a low n peak estimate would be good, like you use n=5000 for 15k.

Anyway, I will not be working on these low k's because I tested LLR 3.62 and k<70 is twice as fast as compared to k>2000. Hence I see no point in working on k's greater than 70.

Thanks,
Citrix

VBCurtis 2005-09-23 05:45

Gribozavr:
My sieving-depth notes don't make as much sense now as they did in July, but my estimate is at p=1.5T, you're finding a factor about once every 2 hrs 10 min, with the efficiency cutoff at about 2 hrs 20 min. Thus, you've stopped sieving at almost exactly the right point for your chosen k-value.

If you have recent data about how often factors are found, I'd like to hear how good my estimate is. factors are not smoothly distributed, so anything from 2 hrs up to 2 1/2 hrs per factor in NewPGen will make me think my guesswork is accurate.
-Curtis

Citrix 2005-09-23 06:14

[QUOTE=Citrix]I guess you guys are right, Why 100k? I just thought a low n peak estimate would be good, like you use n=5000 for 15k.

Anyway, I will not be working on these low k's because I tested LLR 3.62 and k<70 is twice as fast as compared to k>2000. Hence I see no point in working on k's greater than 70.

Thanks,
Citrix[/QUOTE]

I am instead working on 47*2^n+1, though not part of this project, it has really low weight and each test is as fast as it gets. If anyone is intrested in helping, just PM me.

Citrix

Thomas11 2005-09-23 08:59

VBCurtis:
The values of k I'm working on are all sieved to pmax=12T yet, with nmax=10M.
If you want to do n=0...2M, then pmax=4T should be quite sufficient.
One could also think of using SBFactor to do some P-1 factoring on candidates with n>1M.
I did some trials but found the sieving still more efficient. Maybe SBFactor will get another chance at even larger values of n.

-- Thomas

Kosmaj 2005-09-23 13:23

[B]Citrix[/B]
47*2^n+1 is extensively checked. Have a look [URL=http://www.geocities.jp/turbo_us_p/prime/47/status.html]here[/URL] for details.

Citrix 2005-09-23 17:51

[QUOTE=Kosmaj][B]Citrix[/B]
47*2^n+1 is extensively checked. Have a look [URL=http://www.geocities.jp/turbo_us_p/prime/47/status.html]here[/URL] for details.[/QUOTE]


Thanks for pointing it out to me. I have reserved the k from 1.8M onwards at prothsearch.net and plan to take it to 5M. Since this is the only k I am woking on, I think I can do that alone. Other than that the k is low wt and I really don't care any more about finding small primes. I want to find a prime that makes it to the top 10 list. Since with the few resources I have, the only way to do so is to gamble with a low wt k. Just hoping there is a prime for the k=47. I hope I am lucky.


edit: - I also checked on prothsearch.net, 'Nohara' has not reserved anything on prothsearch.net with n>1.8M , so I guess I am free to work on this k.

Thanks,
Citrix

gribozavr 2005-09-23 19:15

[QUOTE=VBCurtis]If you have recent data about how often factors are found, I'd like to hear how good my estimate is. factors are not smoothly distributed, so anything from 2 hrs up to 2 1/2 hrs per factor in NewPGen will make me think my guesswork is accurate.[/QUOTE]

As far I remember, about 10 factors were found a day. 24 hours / 10 = 2 hours 24 minutes, so you were quite accurate!

Just an update: now I'm LLR'ing n=537154, time per iteration = 7.108 ms.

Templus 2005-09-23 22:23

I'm currently working a bit on k=27791 and k=10813783. I will bring the k=10813783 to 1M, since it has only 550 n's left to test.
k=27791 has been tested until n=214k
k=10813783 has been tested until 315k
Both without new primes.

Keller 2005-09-25 19:16

Working on 23437 and 11955659 ;)


All times are UTC. The time now is 06:58.

Powered by vBulletin® Version 3.8.11
Copyright ©2000 - 2020, Jelsoft Enterprises Ltd.