[QUOTE=rogue;390121]Look at posts 216 and 217 in this thread. You get triggering a numeric overflow. There are ways around it.[/QUOTE]

Whops, sorry, I misread your "I was wrong on that" comment :redface: Ithougt you were referring to the workaround.

Luigi

I have posted pfgw 3.7.9 at sourceforge. The only change is to use gwnum v28.6 which fixes a bug with certain AVX FFTs (see the other thread that discusses that issue).

I'm working on a change to pfgw so that it uses primesieve instead of the custom sieve that it has been using since 3.5.0. I'm looking for some people to help test the changes. People most notably affected will be those using pfgw for:
[LIST=1][*]primorials[*]GFN testing[*]trial factoring[/LIST]
And there are others that I have likely missed. If you want to help me test, please send me your e-mail via PM and I'll send you a binary as soon as it is ready. Right now I only have Windows binaries. If those work correctly, I will commit all of my changes and make the appropriate updates to the makefiles for OS X and linux.

[QUOTE=rogue;405543]I'm working on a change to pfgw so that it uses primesieve instead of the custom sieve that it has been using since 3.5.0. I'm looking for some people to help test the changes. People most notably affected will be those using pfgw for:
[LIST=1][*]primorials[*]GFN testing[*]trial factoring[/LIST]
And there are others that I have likely missed. If you want to help me test, please send me your e-mail via PM and I'll send you a binary as soon as it is ready. Right now I only have Windows binaries. If those work correctly, I will commit all of my changes and make the appropriate updates to the makefiles for OS X and linux.[/QUOTE]

Fail!

I've converted the code over, but it is very slow. The reason is that primesieve uses callback functions, which slow things down dramatically. It's great if you just want to pump out primes in succession, but not so good when you want to access a specific prime or get the "next" or "previous" primes from defined starting point. I will download primegen and do some tests. I the original problem with primegen (missing primes) is fixed, then I might go back to it. I'm getting busy at work so this is low on my things to do.

I can't help wondering if something like trial division mixed with an algorithm like pollard rho might be effective.

I found a bug in the current version of pfgw (3.7.9) in the trivial factoring code. If the smallest factor is above 64000 it could miss it. I have a fix and I've asked someone to test.

Nothing serious, just something I noticed.

I'm running PFGW 3.7.9 with GWNUM 28.6.

Saw that if HELP is clicked, it says "Open WinPFGW Version 3.6.5 with ...."

I've posted pfgw 3.710 (Windows only right now, Mac to soon follow). This fixes the trial division error that I reported a couple of weeks ago.

Note that I am considering discontinuing the distribution of 32-bit executables of pfgw. I'll be able to build them. I just appears that almost nobody is using them.

Cosmetic "bug"

PFGW Version 3.7.10.64BIT.20150809.Win_Dev [GWNUM [U][COLOR=Red][B]28.6[/B][/COLOR][/U]] it should be GWNUM [U][COLOR=Red][B]28.7[/B][/COLOR][/U]?

It is still gwnum v28.6. I did not update gwnum in this release.

I discovered that I had the wrong version of gwnum linked with the Mac build. That is updated. I'll get around to updating gwnum in 3.7.11.