Bug in iteration count in Prime95?
 

I am doing a PRP test on M332290099 (again after restarting after I gave up on it after a ton of Gerbicz errors and then went back to it but the save files were bad so the test restarted).

I noticed that it said "Iteration: 6645800/3322900[B]98[/B]" (SEE ATTACHED IMAGE.)

I thought "weird, I never noticed in all this time that second number is 1 less than the exponent". But then I went home and checked on one of my home computers and the second number is definitely the exponent and not 1 less than the exponent.

Did I run into a bug?

[QUOTE=JuanTutors;591998]Did I run into a bug?[/QUOTE]No.
LL does p-2 iterations. [URL="https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test"]https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test[/URL]
PRP does p-1 iterations. [URL]https://en.wikipedia.org/wiki/Fermat_primality_test[/URL]

[QUOTE=kriesel;592005]No.
LL does p-2 iterations. [URL="https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test"]https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test[/URL]
PRP does p-1 iterations. [URL]https://en.wikipedia.org/wiki/Fermat_primality_test[/URL][/QUOTE]

Yes, BUT here are my iterations at home for M332316293 and M332292529 (See image):

"Iteration: 203663120/3322925[b]29[/b]"
"Iteration: 168611800/3323162[b]93[/b]"

I see the same thing on my desktop. The number of iterations it states is the exponent.

[STRIKE]You are running [U][B]two[/B][/U] workers. Each worker has its own exponent. Hint: read your output lines careful.
Probably (accidentally or intentionally) you selected from the menus "windows/show all workers in the same window".[/STRIKE]
Scratch that. I misunderstood your question. Going for the morning coffee... :blush:
Are you running same version of p95, and same test type?

[QUOTE=kriesel;592005]LL does p-2 iterations. [URL="https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test"]https://en.wikipedia.org/wiki/Lucas%E2%80%93Lehmer_primality_test[/URL][/QUOTE]

Yes but prime95 has a bug where it starts numbering LL iterations at 3.

[QUOTE=LaurV;592032]
Are you running same version of p95, and same test type?[/QUOTE]
All running PRP. At home, I confirmed that my laptop running 2 instances that I took a picture of above is running 30.3 b6. I also confirmed that my desktop upstairs (not pictured) is running 30.5 b2 for some reason. The one at work that as one less iteration that I showed above is running 30.3 b6 according to mersenne.org, which I assume is accurate. I never checked a 4th computer.

[QUOTE=Prime95;592039]Yes but prime95 has a bug where it starts numbering LL iterations at 3.[/QUOTE]

I checked to see if I am running LL tests accidentally, but these are all PRP tests. I am sure you can check my username through mersenne.org to confirm.

Looking at the code (which has not changed among all the versions you are using), the only way I see for the iteration counter to be one less than the exponent is to set

PRPStraightForward=1

in prime.txt. This disables Gerbicz error checking and proofs.

[QUOTE=Prime95;592075]Looking at the code (which has not changed among all the versions you are using), the only way I see for the iteration counter to be one less than the exponent is to set

PRPStraightForward=1

in prime.txt. This disables Gerbicz error checking and proofs.[/QUOTE]

I will look to see if that is set when I get to that computer on Monday. However, I did not set that option manually. I didn't know that option existed, and the computer in question is the one that has been having problems, so I would not have set it up to avoid Gerbicz error checking.

You may need to send me prime.txt, local.txt, worktodo.txt, and a savefile.

[QUOTE=Prime95;592079]You may need to send me prime.txt, local.txt, worktodo.txt, and a savefile.[/QUOTE]

Will do when I get to work on Monday. Please let me know how to do that. This is a restarted test as well because the first four save files were detected as bad.

[QUOTE=JuanTutors;592080] Please let me know how to do that. [/QUOTE]

Dropbox, email, whatever works for you. The files will be in the same folder as prime95.exe.

[QUOTE=Prime95;592084]Dropbox, email, whatever works for you. The files will be in the same folder as prime95.exe.[/QUOTE]

Will do on Monday. In the mean time, I think I want to restart all the tests on that computer. After I send the files, I will delete Prime95 with all its files and re-download the program. Which version of the program should I download?

The official version is fine for running PRP tests.

[QUOTE=Prime95;592099]The official version is fine for running PRP tests.[/QUOTE]

Got it. I sent the files and then deleted and re-downloaded 30.3 b6 and restarted the exponent as well.

I've looked at the files and think either

1) A corrupt save file caused the iteration counter to go off by one (switched to non-Gerbicz error checking)
2) The PRP test was continued from an older version of prime95 that did not support Gerbicz error checking.

I think I have a fix in place to catch case number 1.

[QUOTE=Prime95;592563]I've looked at the files and think either

1) A corrupt save file caused the iteration counter to go off by one (switched to non-Gerbicz error checking)
2) The PRP test was continued from an older version of prime95 that did not support Gerbicz error checking.

I think I have a fix in place to catch case number 1.[/QUOTE]
If it's just a corrupted file, I think I may just continue with the restart. It's so early in the test and the computer was being iffy to begin with. Honestly, after 2 issues with the same exponent, if it helps to run just to check for a bug, I am totally willing to do that. Otherwise, I'll stick to the restart.

btw, just stumbled about this thread, remember it as a very old bug, already found in version 25.8 (from 2008) under Windows XP 32-bit ...
so has nothing to do with GEC, corrupted files, special entries from undoc.txt ...