mersenneforum.org  

Go Back   mersenneforum.org > Great Internet Mersenne Prime Search > Data > Marin's Mersenne-aries

Reply
 
Thread Tools
Old 2003-09-27, 08:49   #1
GP2
 
GP2's Avatar
 
Sep 2003

2,579 Posts
Default

Quote:
Originally posted by Xyzzy
One question... Why would you choose Pfactor over Pminus1?
Well for Pfactor, you simply specify
Code:
Pfactor=exponent,how_far_factored,has_been_LL_tested_once
So
1) it's easy to automatically generate those lines based on information in the data files.
2) This Pfactor line will use the exact same P-1 bounds on your computer as if this exponent had been automatically assigned to you by Primenet for double-checking and your computer did the P-1 test before doing the double-check.

Your computer calculates the probability of finding a factor and sets the P-1 bounds accordingly. If your chances of finding a P-1 factor are 3%, the P-1 test should be allowed to run only 3% as long as the double-check would, so the P-1 bounds are automatically set for you with this in mind.

Naturally, if you specify that the exponent has not already been LL-tested once (Pfactor=p,bits,0), the program will pick much larger P-1 bounds because then it's worthwhile to spend 3% of the time it would take to do two LL tests. However, all of those exponents I listed have by definition already been LL-tested once.


On the other hand, for Pminus1 you must manually specify the bounds yourself:
Code:
Pminus1=exponent,B1,B2,0,0
so you have to come up with your own values of B1 and B2 (or B2=0 for default).

But go for it if that's what you prefer.


Quote:

If having more memory makes it go a bit faster, wouldn't it be better to run the P-1s at the normal bounds so you get more exponents done?
If you're using Primenet, there are no "normal" bounds, they increase gradually as the exponents increase. Again, it's based on the time tradeoff of P-1 factoring vs. double-checking given a certain probability of finding a P-1 factor. Larger exponents take longer to LL-test, so it's worthwhile to spend more time on the P-1 factoring also.


Quote:

Also, I find it to be odd that Pfactor requires a torture test for the FFT in use but Pminus1 does not...
Not sure about this... maybe the Pminus1 bounds you manually specified require a different FFT size than the Pfactor-generated bounds, namely one which your computer already used before?

Last fiddled with by GP2 on 2003-09-27 at 08:53
GP2 is offline   Reply With Quote
Old 2003-09-27, 09:55   #2
Xyzzy
 
Xyzzy's Avatar
 
"Mike"
Aug 2002

758310 Posts
Default

I don't know... I ran a few thousand P-1s during the last month and I never had to do a torture test until I switched to Pfactor...

Which reminds me, I can't seem to find this work I did in any of the status files... I assume since they were not from PrimeNet I'll never get credit for them...

I will use Pfactor since it is, as you illustrated, much simpler...

These 10M exponents need a little less than an hour each to do...
Xyzzy is offline   Reply With Quote
Old 2003-09-27, 10:53   #3
GP2
 
GP2's Avatar
 
Sep 2003

50238 Posts
Default

Quote:
Originally posted by Xyzzy
Which reminds me, I can't seem to find this work I did in any of the status files... I assume since they were not from PrimeNet I'll never get credit for them...
If you submit the P-1 lines from your results.txt to the manual testing page http://www.mersenne.org/ips/manualtests.html then the results appear in the next version of the database files. I've done this.

They don't appear in the hourly cleared.txt files of Primenet and you don't get Primenet credit.

What I did was to create a second directory for occasional manual P-1 tests. I copied everything and then edited prime.int to use:
UsePrimenet=0
That keeps two different results.txt files, only one of which needs to be manually cut and pasted into the manual testing page.

Can you give some examples of which exponents you tested? I can look for them in PMINUS1.TXT and the factors file.
GP2 is offline   Reply With Quote
Old 2003-09-27, 11:00   #4
Xyzzy
 
Xyzzy's Avatar
 
"Mike"
Aug 2002

1D9F16 Posts
Default

Quote:
Originally posted by GP2
Can you give some examples of which exponents you tested? I can look for them in PMINUS1.TXT and the factors file.
I use the manual page also... Here are the last 10 results I submitted... These were submitted two days ago so I know they aren't showing yet...

UID: Team_Prime_Rib/Xyzzy, M1584259 completed P-1, B1=20000, B2=300000, WZ1: 291DDF51
UID: Team_Prime_Rib/Xyzzy, M1584307 completed P-1, B1=20000, B2=300000, WZ1: 291FDF4F
UID: Team_Prime_Rib/Xyzzy, M1584343 completed P-1, B1=20000, B2=300000, WZ1: 2914DF51
UID: Team_Prime_Rib/Xyzzy, M1584367 completed P-1, B1=20000, B2=300000, WZ1: 292CDF47
UID: Team_Prime_Rib/Xyzzy, M1584413 completed P-1, B1=20000, B2=300000, WZ1: 2923DF54
UID: Team_Prime_Rib/Xyzzy, M1584433 completed P-1, B1=20000, B2=300000, WZ1: 2908DF57
UID: Team_Prime_Rib/Xyzzy, M1584437 completed P-1, B1=20000, B2=300000, WZ1: 2914DF4A
UID: Team_Prime_Rib/Xyzzy, M1584487 completed P-1, B1=20000, B2=300000, WZ1: 2910DF49
UID: Team_Prime_Rib/Xyzzy, M1584509 completed P-1, B1=20000, B2=300000, WZ1: 292EDF4F
UID: Team_Prime_Rib/Xyzzy, M1584607 completed P-1, B1=20000, B2=300000, WZ1: 2904DF4C


I know I did several thousand, but I don't have the log files anymore... (I have a hard time keeping track of stuff like that!)

Most of my work is turned in under Team_Prime_Rib/Xyzzy or Xyzzy/Xyzzy...
Xyzzy is offline   Reply With Quote
Old 2003-09-30, 02:52   #5
GP2
 
GP2's Avatar
 
Sep 2003

2,579 Posts
Default

Quote:
Originally posted by Xyzzy
I use the manual page also... Here are the last 10 results I submitted... These were submitted two days ago so I know they aren't showing yet...

UID: Team_Prime_Rib/Xyzzy, M1584259 completed P-1, B1=20000, B2=300000, WZ1: 291DDF51
...

All of these are found in the latest PMINUS1.TXT data file as of September 29 2003. The previous bounds used to be B1=B2=10000.
GP2 is offline   Reply With Quote
Reply

Thread Tools


Similar Threads
Thread Thread Starter Forum Replies Last Post
Pminus1 - Bug report and feature request axn Software 36 2020-06-14 09:48
ivy bridge versus haswell diep Hardware 29 2017-12-06 13:43
Prime95 P-1/Pfactor using to find factors on CRUS bases pepi37 Conjectures 'R Us 23 2017-03-14 20:38
CUDALucas versus Mfaktc/o Brain GPU Computing 26 2011-12-06 08:48
NTT transform at (AMD) GPU versus *lucas diep GPU Computing 11 2011-05-11 20:27

All times are UTC. The time now is 12:49.

Sat Aug 8 12:49:02 UTC 2020 up 22 days, 8:35, 1 user, load averages: 1.62, 2.01, 1.92

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.