mersenneforum.org I'd like to help out
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

 2005-04-10, 13:10 #45 VBCurtis     "Curtis" Feb 2005 Riverside, CA 3·19·83 Posts Proth appears to cap k at 2^31, and when I run LLR on the output file from NewPGen, it only spits back the first line of the input file. What am I doing wrong? Separate question: I remembered reading something about it being faster to just LLR small n's, rather than sieving them; I pulled the small n out of the sieve, then restarted it. It appears to have restarted sieving at p=2, even though the top line of the file showed p=2 billion (roughly) at last save. Is there some paramater to alter when you pull part of a sieve file to prevent it starting over? Finally, can I force LLR to test primality (PRP, I think, since k>2^n) for the really small cases of n? Say, less than 1500. I tried Proth for this, discovering the apparent 2^31 limit. Am I just better off picking a k smaller than 2^31 and restarting? -Curtis
2005-04-10, 15:02   #46
jocelynl

Sep 2002

2·131 Posts

Quote:
 Originally Posted by VBCurtis Proth appears to cap k at 2^31, and when I run LLR on the output file from NewPGen, it only spits back the first line of the input file. What am I doing wrong?
you must factorize k's larger than 2^32 your NewPGen output should like like this after sieving to p=1000000

Code:
ABC 3*5*11*11*13*331*487*2^\$b-1 //NewPGen:1000000:M:0:2:322
1 2
1 3
1 4
...
You must transform it to run LLR test.
It should look like this

Code:
1000000:M:0:2:258
3803443215 2
3803443215 3
3803443215 4
...
Quote:
 Originally Posted by VBCurtis Separate question: I remembered reading something about it being faster to just LLR small n's, rather than sieving them; I pulled the small n out of the sieve, then restarted it. It appears to have restarted sieving at p=2, even though the top line of the file showed p=2 billion (roughly) at last save. Is there some paramater to alter when you pull part of a sieve file to prevent it starting over? Finally, can I force LLR to test primality (PRP, I think, since k>2^n) for the really small cases of n? Say, less than 1500. I tried Proth for this, discovering the apparent 2^31 limit. Am I just better off picking a k smaller than 2^31 and restarting? -Curtis
the new LLR will trial factor the small n's where k>2^n. You can test these first. Then sieve the rest.

 2005-04-11, 19:58 #47 VBCurtis     "Curtis" Feb 2005 Riverside, CA 3×19×83 Posts Thanks! Everything runs great. It seems my sieve started over because I left a blank line between the top line and the list of exponents; I deleted that space, and it resumed where I left off. I've run to about n=40,000, with roughly 40 primes. I'll post them once I hit n=100k. -Curtis
 2005-07-17, 16:42 #48 Kosmaj     Nov 2003 2·1,811 Posts I did some updates to the table of 15k Primes. It's still incomplete but it's a little bit better than before. I hope Joss will be back soon with a full update. I have two questions: 1) To VBCurtis: k=3803443215, have you checked the 178053-194673 interval and if so have you found any new primes? What limit in terms of n have you reached? 2) To lsoule: k=8331405, have you checked the 45773-186251 interval and if so have you found any new primes? Thanks! Last fiddled with by Kosmaj on 2005-07-17 at 17:07
 2005-07-17, 20:43 #49 VBCurtis     "Curtis" Feb 2005 Riverside, CA 3×19×83 Posts Weird.. I go 30 days without checking the board, and the day I reappear, there's a question for me. I have checked k=3903443215 to n=275k, including the range you asked about, with no further primes to report. I've been focusing my efforts on the low-weight numbers, but plan to eventually run this k to n=500k. I'll post an update on the low-weight numbers shortly. -Curtis
2005-07-18, 04:29   #50
lsoule

Nov 2004
California

23×3×71 Posts

Quote:
 Originally Posted by Kosmaj I did some updates to the table of... 2) To lsoule: k=8331405, have you checked the 45773-186251 interval and if so have you found any new primes? Thanks!
I checked my logs and have not checked the range 47828..181000.
I do have a sieve file (sieved only to 17B with 20k candidates)
if anyone wants it.

thanks,
Larry

 2005-07-18, 14:17 #51 Kosmaj     Nov 2003 1110001001102 Posts Curtis and Larry, Thanks for your quick replies. I added your info and I also added all small primes reported in the Post small primes thread. I updated k=25935 with its 66 small primes n<10k. k=25935 was tested by Thomas Ritschel from about n=100k (?) to 200k, then rediscovered by Larry and tested by me from 200k to 220k. But I just found that two primes for n=46144 and 47396 had been found and reported back in 2000 to Top-5000 by g174 (T.J. Engelsma). k=25935(=3*5*7*13*19) looks promising but for large n's one has to be patient. Please let me know if you find any typos, inconsistencies, etc. Last fiddled with by Kosmaj on 2005-07-18 at 14:19
 2005-08-24, 17:48 #52 Cruelty     May 2005 65816 Posts I will give a try to 736320585... Since this is my first candidate, do you have any suggestions for newpgen/LLR other than those already in the forums? BTW: Is it possible to force CPU affinity in newpgen? Last fiddled with by Cruelty on 2005-08-24 at 18:03
 2005-08-25, 03:23 #53 Kosmaj     Nov 2003 1110001001102 Posts Cruelty No special tips, just be patient and sieve for a while before starting LLR. It's hard to give concrete advice but for example for n=200-500k sieve at least for 48 hours or to 500bn whatever comes first. It is possible to force cpu affinity (I assume it's a dual box, unless you have those cool 4-way?) using Task Manager (in Windows) but I never do that. This can also degrade performance, try it but check the speed. I just found by searching Top-5000 that this k was checked by Herranen back in 1998-99 most likely to n=70k. Looks good, there is a Sophie-Germain prime 736320585*2^6194-1. VBCurtis I'm sorry about that. Joss marked some k's in the low-weight table as being reserved by "Riesel sieve" but obviously he missed this one. Please check their site next time.
2005-08-25, 07:57   #54
Cruelty

May 2005

23·7·29 Posts

Quote:
 Originally Posted by Kosmaj I just found by searching Top-5000 that this k was checked by Herranen back in 1998-99 most likely to n=70k. Looks good, there is a Sophie-Germain prime 736320585*2^6194-1.
Thanks for the info. Next time I will search TOP-5k database before starting sieving

As for the PC I'm using for "15k" it's a Dualcore Athlon 64 @ 2,55GHz

Last fiddled with by Cruelty on 2005-08-25 at 08:00

 2005-08-25, 16:25 #55 ValerieVonck     Mar 2004 Belgium 292 Posts Re-reserving 13236795 from 340k.... Cedric Vonck