mersenneforum.org Prime95 switched assignment mid test.
 User Name Remember Me? Password
 Register FAQ Search Today's Posts Mark Forums Read

 2018-12-21, 04:45 #1 C7703     Nov 2018 2 Posts Prime95 switched assignment mid test. So I was about 50% of the way through a LL test and suddenly my computer started failing a bunch of Jacobi checks. I restart the test and now my my computer is starting a completely new LL test. Anyone know what's happening? Attached Thumbnails
 2018-12-21, 05:56 #2 GP2     Sep 2003 22·647 Posts Which version of the software were you using? If it's 29.4 or earlier, you might have a hardware issue. If it's beta 29.5, it might be a software issue. In any case, if the Jacobi check failed, your interim results were erroneous, and if it couldn't find a savefile that passed the Jacobi check, the only alternative was to start over. A Jacobi check only has a 50% chance of detecting an error, however. You might want to stop the program and edit your prime.txt file to do more frequent Jacobi checking and keep more backup files. For example: Code: JacobiErrorCheckingInterval=6 JacobiBackupFiles=4 See the undoc.txt file for a description of those parameters. However, as an alternative, you might also want to consider switching to the PRP work type. That has much more rock-solid error checking. You can do so by stopping the program, editing the prime.txt file to change the WorkPreference= line to WorkPreference=150, and restarting the program. Before restarting the program, you might want to delete the line in worktodo.txt that says Test=................,1,2,88284421,-1 if you want to abandon it, otherwise it will still try to complete that assignment before switching to the PRP work type. You could also edit the line to PRP=1,2,88284421,-1 if you'd rather do that exponent as a PRP test instead of as an LL test.
 2018-12-21, 06:29 #3 C7703     Nov 2018 210 Posts I'm using 29.4. Thanks for the help. I will definitely consider a PRP test. If you don't mind me asking, if this is a more "rock solid" test, why do LL tests at all in the first place?
2018-12-21, 08:08   #4
GP2

Sep 2003

22·647 Posts

Quote:
 Originally Posted by C7703 I'm using 29.4. Thanks for the help. I will definitely consider a PRP test. If you don't mind me asking, if this is a more "rock solid" test, why do LL tests at all in the first place?
The rock-solid error checking method was only discovered relatively recently, in late 2017 if I recall correctly.

PRP testing is an old concept, but it is no faster than LL, and has the additional drawback that it can only show that a Mersenne number is probably prime rather than provably prime. So until the error checking method was invented, there was no reason to prefer it.

The project is being cautious about transitioning to PRP in a more sweeping way, but it's been talked about. However LL tests would still be needed for double checks of old results, and also for actually proving a prime when PRP flags it as a probable prime.

2018-12-21, 09:05   #5
axn

Jun 2003

2×2,459 Posts

Quote:
 Originally Posted by GP2 The project is being cautious about transitioning to PRP in a more sweeping way, but it's been talked about. However LL tests would still be needed for double checks of old results, and also for actually proving a prime when PRP flags it as a probable prime.
The main reason is that the capability was added only to recent client versions. Until all users migrate to them, they can't crunch PRP at all.

 Similar Threads Thread Thread Starter Forum Replies Last Post Welton Information & Answers 7 2016-07-29 12:07 nik PrimeNet 19 2014-07-25 08:29 snme2pm1 PrimeNet 6 2014-04-12 03:53 doffdoff Software 12 2012-05-04 19:53 jonthomson Software 17 2007-05-12 13:21

All times are UTC. The time now is 13:42.

Mon Apr 19 13:42:12 UTC 2021 up 11 days, 8:23, 0 users, load averages: 2.72, 2.92, 2.79