New sieve program.... - Page 3

Greenbank · 2005-10-18, 09:49

Thanks for your replies. Now I have a few more questions, follow-up's.

1) Understood that 3 <= k < 2^31. Can you possibly include support for larger k's, let's say up tp 2^63?

I don't control proth_sieve, however I'll let Chuck/Joe_O know. I think the largest you'll get with proth_sieve/jjsieve would be 2^50.

2a) I was able to start jjsieve to factor from 5 by entering p_min=3 and p_max on the command line. But it skips 3. It will be nice if you can add an option to start a completely new sieve so that jjsieve creates the .dat file. Running NewPGen to 1bn on 10 or 20 k's, then combining the output into a large .dat file is no fun

Try entering p_min = 2. It might work.

2b) When I group k's into .dat file is it better to have one .dat for k's divisible by 3 and one .dat for the others? Instead of a large file for all k's.

Large file for all k's, otherwise you are duplicating work.

3) About "-u" and "-full" switches: Does it mean these two are not working? IMO, It will be nice to have them working to shrink the size of .dat from time to time, specially in early stages. By "file for LLR" I mean a file with a list of exponents, one per line, I can add the rest easily.

No idea about this, again I'm not the author of the program, I've just used it quite a bit.

4) We at 15k Search have several projects running, for example:
PJ-1: 1 < k < 300
PJ-2: 15k like k=9225, 355424355, 3611911875 (all divisible by 3, 5 and other small primes, for example 355424355 = 3^4*5*11*13*17*19^2).
PJ-3: low weight, like k=253, 10943321, 355281299 (no one divisible by 3)

So if we use jjsieve we are going to have several .dat files (some ppl may prefer to sieve alone) and with some exceptions we sieve to smaller p than SOB and RieselSieve. I wonder is there any guidance how many k's to group in one .dat file, 4, 10, 16? Is it better to have the number of k's be a power of 2 or not? What's the smallest number of k's when it makes sense to use jjsieve?

The more you have in one file the more efficient it is. It's also much easier to keep track of what ranges you have sieved if everything is in one file.

Interesting though, what do you hope to prove/find out from looking at these k values?

BTW, FYI, the latest version of LLR can test k*2^n+1 too, but at "15k" we only test k*2^n-1.

Well, it's doing a Proth test, the LLR test does not work for k*2^n+1.

Sierpinski (k*2^n+1): Proth Test.
Riesel (k*2^n-1): Lucas-Lehmer-Riesel (LLR)

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Advantage of lattice sieve over line sieve	binu	Factoring	3	2013-04-13 16:32
Program	Primeinator	Information & Answers	5	2009-07-16 21:42
Program for GPU	tribal	Information & Answers	5	2009-03-19 20:54
program to verify factors found by sr(x)sieve?	mdettweiler	Software	16	2009-03-08 02:06
program P-1 for K*2^n-1	jocelynl	15k Search	19	2004-01-11 17:24

2005-10-18, 09:49	#23
Greenbank Jul 2005 2·193 Posts	Thanks for your replies. Now I have a few more questions, follow-up's. 1) Understood that 3 <= k < 2^31. Can you possibly include support for larger k's, let's say up tp 2^63? I don't control proth_sieve, however I'll let Chuck/Joe_O know. I think the largest you'll get with proth_sieve/jjsieve would be 2^50. 2a) I was able to start jjsieve to factor from 5 by entering p_min=3 and p_max on the command line. But it skips 3. It will be nice if you can add an option to start a completely new sieve so that jjsieve creates the .dat file. Running NewPGen to 1bn on 10 or 20 k's, then combining the output into a large .dat file is no fun Try entering p_min = 2. It might work. 2b) When I group k's into .dat file is it better to have one .dat for k's divisible by 3 and one .dat for the others? Instead of a large file for all k's. Large file for all k's, otherwise you are duplicating work. 3) About "-u" and "-full" switches: Does it mean these two are not working? IMO, It will be nice to have them working to shrink the size of .dat from time to time, specially in early stages. By "file for LLR" I mean a file with a list of exponents, one per line, I can add the rest easily. No idea about this, again I'm not the author of the program, I've just used it quite a bit. 4) We at 15k Search have several projects running, for example: PJ-1: 1 < k < 300 PJ-2: 15k like k=9225, 355424355, 3611911875 (all divisible by 3, 5 and other small primes, for example 355424355 = 3^4511131719^2). PJ-3: low weight, like k=253, 10943321, 355281299 (no one divisible by 3) So if we use jjsieve we are going to have several .dat files (some ppl may prefer to sieve alone) and with some exceptions we sieve to smaller p than SOB and RieselSieve. I wonder is there any guidance how many k's to group in one .dat file, 4, 10, 16? Is it better to have the number of k's be a power of 2 or not? What's the smallest number of k's when it makes sense to use jjsieve? The more you have in one file the more efficient it is. It's also much easier to keep track of what ranges you have sieved if everything is in one file. Interesting though, what do you hope to prove/find out from looking at these k values? BTW, FYI, the latest version of LLR can test k2^n+1 too, but at "15k" we only test k2^n-1.* Well, it's doing a Proth test, the LLR test does not work for k2^n+1. Sierpinski (k2^n+1): Proth Test. Riesel (k*2^n-1): Lucas-Lehmer-Riesel (LLR)