PFGW 3.3.6 or PFGW 3.4.2 Please update now!
 

[B]Please update to PFGW 3.3.6 or PFGW 3.4.2 [/B] available [URL="http://openpfgw.svn.sourceforge.net/viewvc/openpfgw/"]here[/URL] as soon as you can. Download the new program, stop your current program with CTRL-c, and write over it with the new program.

Use only machines you own, or have written permision to use. We will not be responsible if you get into trouble. This is always true, but needs to be emphasized now and again. We will not sympathize with you if you break this rule. You must act responsibly.

The primality program PFGW is CPU intensive and will make your machine run [SIZE="6"][COLOR="Red"]hot[/COLOR][/SIZE].
You should probably use a program to monitor temperatures and shut down if your machine gets too [SIZE="6"][COLOR="Red"]hot[/COLOR][/SIZE]. You probably should not use a laptop or all-in-one machine. Please feel free to post any other warnings that I have forgotten. Normally, I would not prefer PFGW over LLR, but PFGW is the first program to have fixed the GWNUM FFT selection problem. If you prefer to use LLR, then please stop running until there is a new LLR. There is now a new LLR Version 3.8.2 using the v26.2 of the gwnum library is available as a development version. Check out the known bugs in GWNUM v26.2 to see if this LLR version is for you: [url]http://www.mersenneforum.org/showpost.php?p=229781&postcount=2[/url]

PFGW 3.4.2 is using GWNUM v 26.4 in which these bugs have been fixed.

Urgent Message
 

When I started to check this out it didn't look too bad. Turns out the test program was faulty. It is really bad. It could take me a month to test all k's just to find out which n need to be retested.

Let's just get everyone on the new PFGW ASAP.

Note that LLR 3.8 is also affected, but phrot is not.

You should be able to run > 100,000 numbers with -F in less than an hour. I don't expect base 5 to show any differences, but I would like to know if you discover any.

I've only done 2 k so far with the "new" 3.3.5 but there are differences. For the lowest +1k they start beyond where we have tested, but for the highest +1k they start almost immediately. I'm now running the highest -1k and will let you know what I find.

Lennart has taken the lead on this. He's compared the FFT sizes for n up to 550K on the +1 side and 500K on the -1 side. Six more on the -1 side have been compared up to 2M. I'll get you detailed stats once we've compiled them. For now, there are 24117 tests on the +1 side that use a different FFT length. We've tested 11449 of them without finding a prime.

All 24117 +1 tests have been completed. No prime was found. We have imported the results as 2nd pass and compared residues. Only 61 tests had different residues (.2429%)