mersenneforum.org 64-bit gnfs-lasieve*
 Register FAQ Search Today's Posts Mark Forums Read

 2009-03-30, 12:41 #1 mklasson   Feb 2004 2·3·43 Posts 64-bit gnfs-lasieve* The source I got from http://ggnfs.svn.sourceforge.net/viewvc/ggnfs/ doesn't appear to contain much 64-bit asm (only a little snippet inline). Ben seemed to say the other day though that there's a version that does. Does anyone know where I might find it? It's supposed to be linux only, but I'm hoping you all have just been lazy and/or linux-using and getting it to compile under windows won't be terribly difficult... I'm going away on vacation for a week in a couple of days and it sure would be nice to cram two weeks of computation in there.
 2009-03-30, 12:48 #2 jasonp Tribal Bullet     Oct 2004 DD716 Posts There is another version of the Franke/Kleinjung siever that includes a lot of 64-bit optimizations; several here have the source for it. The changes have not been folded into GGNFS and the code very likely doesn't work in Windows. Nobody has had the time or the willpower to fix that.
2009-03-30, 13:00   #3
mklasson

Feb 2004

2·3·43 Posts

Quote:
 Originally Posted by jasonp There is another version of the Franke/Kleinjung siever that includes a lot of 64-bit optimizations; several here have the source for it. The changes have not been folded into GGNFS and the code very likely doesn't work in Windows. Nobody has had the time or the willpower to fix that.
I'm reasonably well endowed with both. Could one of these people share the source so I can determine if sufficiently so? If it's too much work I probably won't be, but running at half speed now is a fairly strong motivator.

2009-03-30, 13:10   #4
bsquared

"Ben"
Feb 2007

3×1,193 Posts

Quote:
 Originally Posted by mklasson I'm reasonably well endowed with both. Could one of these people share the source so I can determine if sufficiently so? If it's too much work I probably won't be, but running at half speed now is a fairly strong motivator.
I hope that you are! I mostly run linux, but would find uses for the optimized version for windows too, if it were available. I'll email you the source, but might wait for Batalov to chime in first because he may have rolled his resume feature and other fixes into the optimized source, which I haven't done yet.

- ben.

2009-03-30, 13:22   #5
mklasson

Feb 2004

2·3·43 Posts

Quote:
 Originally Posted by bsquared I hope that you are! I mostly run linux, but would find uses for the optimized version for windows too, if it were available. I'll email you the source, but might wait for Batalov to chime in first because he may have rolled his resume feature and other fixes into the optimized source, which I haven't done yet.
Excellent! Much obliged.

2009-03-30, 18:06   #6
FactorEyes

Oct 2006
vomit_frame_pointer

16816 Posts

Quote:
 Originally Posted by jasonp There is another version of the Franke/Kleinjung siever that includes a lot of 64-bit optimizations; several here have the source for it. The changes have not been folded into GGNFS ...
I have the source somewhere, but could not build it for the life of me. It seemed to require a tool or two that I had trouble finding on the rpm sites. Some sort of text-processing tool was in the compile chain for the asm.

frmky built Linux 64-bit executables from this code and handed them around a year or so back, to my eternal gratitude.

I'd love Windows executables, but it would also be nice to fold the source into GGNFS in such a way as to make them compilable within Linux, in a straightforward way, which would ease the path to a 64-bit Windows executable.

Is there a licensing issue with the 64-bit code, or is it just too much work?

I would love to understand the Franke lattice-siever code, but it is written at a reading comprehension level somewhat above mine. Dang.

Last fiddled with by FactorEyes on 2009-03-30 at 18:10

2009-03-30, 18:27   #7
mklasson

Feb 2004

1000000102 Posts

Quote:
 Originally Posted by FactorEyes I have the source somewhere, but could not build it for the life of me. It seemed to require a tool or two that I had trouble finding on the rpm sites. Some sort of text-processing tool was in the compile chain for the asm. frmky built Linux 64-bit executables from this code and handed them around a year or so back, to my eternal gratitude.
Oh joy, that sounds promising. Ah well, I'll have a look at least... Ready when you are, Ben.

2009-03-30, 18:58   #8
bsquared

"Ben"
Feb 2007

3·1,193 Posts

Quote:
 Originally Posted by mklasson Oh joy, that sounds promising. Ah well, I'll have a look at least... Ready when you are, Ben.
I sent the source to your gmail account - let me know if you didn't recieve it. It should be modified to at least compile in linux, at least with the 2.6.26 kernel and gcc 4.2.3. But it has not been merged with the latest SVN code AFAIK, and so probably differs somewhat in usability compared with that. On that note, joral had told me last fall that he was going to try to get the optimizations into SVN, but I suspect this is a pretty big task else it would be done already. I certainly don't have the time or knowledge to do so...

Last fiddled with by bsquared on 2009-03-30 at 18:59

 2009-04-01, 01:55 #9 joral     Mar 2008 3716 Posts A couple months ago, my main 64-bit box suffered a catastrophic hard drive failure, so I haven't been able to complete it. It is definitely a good bit of work though. I hope to get back to it soon.
2009-04-01, 18:46   #10
mklasson

Feb 2004

4028 Posts

I got hopeful after seeing there were 64-bit liblasieve*.a libs with the asm already in them.

Unfortunately they don't seem to be in a format msvc accepts:
Code:
2>../../../liblasieve.a : fatal error LNK1107: invalid or corrupt file: cannot read at 0xE064
0xE064 is the end of the file. Does anyone with more cross-cultural experience than me know what the problem might be? It looks very much like a normal .lib, but presumably there's some minor incompatibility somewhere.

Quote:
 Originally Posted by FactorEyes I would love to understand the Franke lattice-siever code, but it is written at a reading comprehension level somewhat above mine. Dang.
Alas, mine too.

The C code didn't seem to require too many changes, but I shouldn't say too much yet... I will be very happy if I can somehow get a working 64-bit .exe out of this, but anything more than that might require a more masochistically inclined soul than mine. On the other hand, the .asm files don't seem quite as terrifying anymore once I learned of m4. So maybe there's hope for me still.

In any case, I'm off for a week come tomorrow. If my prayers are answered a working version will miraculously have appeared here by the time I get back. Pretty please? With sugar on top?

 2009-04-03, 02:12 #11 joral     Mar 2008 5×11 Posts *grumble* looks like I'll have to redo what I had done. Along those lines, could someone please resend me the 64-bit asm code? I lost the original email in my HD crash. This time, I have a place to back it up.

 Similar Threads Thread Thread Starter Forum Replies Last Post pinhodecarlos NFS@Home 0 2014-12-24 19:13 pinhodecarlos NFS@Home 34 2014-04-01 21:27 fivemack Factoring 8 2010-04-27 18:59 Shaopu Lin Factoring 3 2009-11-18 18:42 Andi47 Msieve 5 2009-01-26 18:19

All times are UTC. The time now is 09:13.

Sun Dec 5 09:13:11 UTC 2021 up 135 days, 3:42, 0 users, load averages: 2.72, 2.18, 1.86