Yafu

[Brian Gladman]

Hi Ben

I have just added x64 Windows binaries to SVN at:

yafu\build.vc10\x64\Release

Brian

[bsquared]

Thanks Brian.

Win32 and the x64 binaries are now available for download here.

[RichD]

Has there been any interest in developing a Mac version, namely running on a PowerPC?

I tinkered around with YAFU a bit a while ago and found several minor changes that needed to be incorporated. I got stuck on why all the PPC ASM code (32-bit & 64-bit) was commented out for INNERMUL & PROPCARRY in type.h. It was straight from TomsFastMath libraries. I didn't understand why the code was turned off.

I know Mark R. did some work on this program years ago, but I believe he is too busy right now.

[bsquared]

Originally Posted by RichD

Has there been any interest in developing a Mac version, namely running on a PowerPC?

I tinkered around with YAFU a bit a while ago and found several minor changes that needed to be incorporated. I got stuck on why all the PPC ASM code (32-bit & 64-bit) was commented out for INNERMUL & PROPCARRY in type.h. It was straight from TomsFastMath libraries. I didn't understand why the code was turned off.

I know Mark R. did some work on this program years ago, but I believe he is too busy right now.

I'm definitely interested but I don't have a MAC development platform. So my role in any development is pretty much limited to "support".

As to why all that code was commented out... err, there probably isn't any good reason. I haven't touched the TFM code in years, so I can't remember why I originally did that. But there is a generic C fallback in the TFM library if a specific processors' ASM code is not enabled... does that not work (albeit non-optimally)?

[rogue]

Originally Posted by RichD

I know Mark R. did some work on this program years ago, but I believe he is too busy right now.

I didn't work on yafu, but I did work on GGNFS.

I have actually never compiled yafu.

[RichD]

Originally Posted by rogue

I didn't work on yafu, but I did work on GGNFS.

I have actually never compiled yafu.

Sorry, my mistake.

RichD.

[lorgix]

I eventually figured it out. Thanks.

Now what's up with this;

223!+1 being interpreted as 222!+1

[bsquared]

Originally Posted by lorgix

I eventually figured it out. Thanks.

Now what's up with this;

223!+1 being interpreted as 222!+1

Errr, good question! Just making sure you're paying attention.

I'll look into it, thanks.

[lorgix]

Another strange error appeared when I tried to do ECM on 307!+1.

c15 = 428498317330021 (found in stg1 of curve 1 (thread 0) with sigma = 541093533)
Finished 1 curves using Lenstra ECM method on C615 input, B1 = 2000000, B2 = 400000000

To make things more interesting I got a different output the second time I tried it;

c14 = 16347718253437 (found in stg1 of curve 1 (thread 0) with sigma = 457055413)
Finished 1 curves using Lenstra ECM method on C617 input, B1 = 2000000, B2 = 400000000

P.S. I just tried 151!-1, and surprise surprise.. It was factored as 150!-1

[axn]

Originally Posted by lorgix

Another strange error appeared when I tried to do ECM on 307!+1.

c15 = 428498317330021 (found in stg1 of curve 1 (thread 0) with sigma = 541093533)
Finished 1 curves using Lenstra ECM method on C615 input, B1 = 2000000, B2 = 400000000

To make things more interesting I got a different output the second time I tried it;

c14 = 16347718253437 (found in stg1 of curve 1 (thread 0) with sigma = 457055413)
Finished 1 curves using Lenstra ECM method on C617 input, B1 = 2000000, B2 = 400000000

Well, it is evaluating 306!+1. 428498317330021 = 307 · 1395759991303, and 16347718253437 = 307 · 53249896591. All three are factors of 306!+1 (http://www.factordb.com/index.php?id...00000002850330). BTW, ecm is not deterministic, so you can and will see different outputs from running similar ecm curves on the same number.

[bsquared]

Yes... there is a bug in my factorial function which skips the last term if it is prime.

I will post fixed binaries tonight on sourceforge.

Sorry about that