mersenneforum.org  

Go Back   mersenneforum.org > Prime Search Projects > Sierpinski/Riesel Base 5

Reply
 
Thread Tools
Old 2007-05-08, 20:51   #1
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

5×701 Posts
Default Sieving question

Because I'm obsessive-compulsive, I tend to worry about things that are only important to me. I have a Linux AMD machine that I use for sieving, and a Pentium-D that I use for LLRing. My brain has decided that it is REALLY important that I sieve until the core on my Linux box is sieving less than 5 factors for every one one of my Pentium-D cores LLRs.

Is there a command in sr5sieve that will tell me how many factors are expected in a range without having to mess around with the files to get the answer?
jasong is offline   Reply With Quote
Old 2007-05-08, 21:01   #2
jasong
 
jasong's Avatar
 
"Jason Goatcher"
Mar 2005

5×701 Posts
Default

Never mind. Apparently, my self-applied rule means I should only LLR.
jasong is offline   Reply With Quote
Old 2007-05-10, 23:27   #3
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

22058 Posts
Default

Quote:
Originally Posted by jasong View Post
Is there a command in sr5sieve that will tell me how many factors are expected in a range without having to mess around with the files to get the answer?
The formula sr5sieve uses N*(1-log(P0)/log(P1)), where N is the number of candidates remaining and P0-P1 is the sieve range.

You can start sr5sieve as `sr5sieve -v -p P0 -P P1' to get this estimate, but it is probably easier just to use a calculator.
geoff is offline   Reply With Quote
Old 2007-07-20, 16:09   #4
Rincewind
 
Rincewind's Avatar
 
Oct 2006

11001112 Posts
Default

Another sieving question, but I don't want to start a new topic:
I used an older version of sr5sieve (I think it was 1.0) I think I should read the name before downloading the file. Is this a problem, it is possible that I missed some factors?
Rincewind is offline   Reply With Quote
Old 2007-07-20, 16:27   #5
masser
 
masser's Avatar
 
Jul 2003
wear a mask

59B16 Posts
Default

Geoff can probably tell you what versions were slightly faulty - a good check is to see if you got close to the expected number of factors.

Make sure you download the most recent versions of sr(x)sieve! There have been huge improvements since version 1.0...
masser is offline   Reply With Quote
Old 2007-07-20, 17:53   #6
Rincewind
 
Rincewind's Avatar
 
Oct 2006

103 Posts
Default

I noticed some improvements, my Pentium 1600 MHz (version 1.x) sieves ~44000 p/sec,
my Pentium 800 MHz (version 1.5.15) sieves ~35000 p/sec.
I expected a bigger difference.

Last fiddled with by Rincewind on 2007-07-20 at 17:56
Rincewind is offline   Reply With Quote
Old 2007-07-20, 18:01   #7
masser
 
masser's Avatar
 
Jul 2003
wear a mask

59B16 Posts
Default

Your 800 Mhz machine is probably a P3, while the 1.6 Ghz machine is probably a P4. IIRC, P4's are not as good at sieving as some of the other processors. I would try the latest version on the 1.6 Ghz machine - that would be a better comparison.

Last fiddled with by masser on 2007-07-20 at 18:02
masser is offline   Reply With Quote
Old 2007-07-20, 20:37   #8
Rincewind
 
Rincewind's Avatar
 
Oct 2006

103 Posts
Default

OK, that's new for me, an you're right, its a P3 and a P4. I thought the 800MHz CPU has this performance only because I used the newer version.
I just started to resieve my range (lost 12 percent) with the new version to find every factor.
Rincewind is offline   Reply With Quote
Old 2007-07-21, 00:13   #9
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

13·89 Posts
Default

Quote:
Originally Posted by Rincewind View Post
Another sieving question, but I don't want to start a new topic:
I used an older version of sr5sieve (I think it was 1.0) I think I should read the name before downloading the file. Is this a problem, it is possible that I missed some factors?
All x86 versions before 1.4.0 had a fault in the mulmod code that would start to give incorrect results when p became larger than 2^46, or maybe as low as 2^42 in some rare cases. These errors would become more frequent as p increases. At the current sieve ranges (8000G+) these errors could be starting to have an effect.
geoff is offline   Reply With Quote
Old 2007-07-23, 00:03   #10
geoff
 
geoff's Avatar
 
Mar 2003
New Zealand

48516 Posts
Default

Quote:
Originally Posted by geoff View Post
All x86 versions before 1.4.0 had a fault in the mulmod code that would start to give incorrect results when p became larger than 2^46, or maybe as low as 2^42 in some rare cases.
That was not quite correct, sorry:

For versions before 1.4.0 the lowest known p that gave an incorrect result was about 2^46, which is about 70000G. I can't rule out the possibility of some bad results between 2^42 and 2^46, but I don't think the current sieve ranges at about 8000G would be significantly affected. Sorry if you repeated any work unnecessarily.
geoff is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Dumb sieving question fivemack Software 7 2017-11-27 22:48
Line sieving vs. lattice sieving JHansen NFSNET Discussion 9 2010-06-09 19:25
2^859-1 sieving fivemack Factoring 45 2009-07-01 22:01
A question on lattice sieving joral Factoring 5 2008-04-03 08:01
Sieving Citrix Factoring 3 2003-10-13 14:49

All times are UTC. The time now is 12:07.

Mon Sep 28 12:07:56 UTC 2020 up 18 days, 9:18, 0 users, load averages: 1.99, 2.13, 1.82

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