XYYXF Status Update

[swellman]

It’s been a long time since Andrey posted anything here, so it’s time for a 1 < y < x < 151 XYYXF factoring project update.

- There remain 323 composites left to be factored in the project. All have survived at least 18000 curves @B1=110e6 (t55).

- 10 of these are C300+, all of which have survived ECM to t65+. Greg Childers has agreed to run C311_148_133 on the big siever of NFS@Home late in 2022.

- 3 composites are undoubtedly best factored using GNFS, again all of which have completed ECM. One of these, C203_145_119, is currently being sieved on 16e. I hope to get the remaining two GNFS jobs (C203_147_104 and C204_147_118) factored by the end of the year (or at least enqueued in 16e).

For the past several years I’ve been feeding ECM tasks to Yoyo@Home and occasionally running NFS once ECM was completed. Progress has been very slow, even with the power of BOINC. In the past, Andrey would feed the entire list of composites into Yoyo’s queue and wait for it to finish, then repeat at the next t-level. But this approach won’t work anymore, as t60 requires ~42,000 curves @B1=260e6 and Yoyo has a maximum limit of 20,000 curves on a single task (limit is 9,900 curves @B1=850e6 or higher).

Running three rounds of 14,000 curves could work but that approach would take many years to finish. Running two rounds of 20,000 curves plus some local ECM is still daunting.

Lately I’ve been running ECM in smaller batches, trying get a handful of XYYXF composites ready for NFS. Then I’ll repeat the process. Right now I’m working to get the 7 least difficult jobs remaining (SNFS 267-270) through ECM. Hoping to see the ECM finished by late 2022.

If anyone is reading this, please give feedback. XYYXF factoring has been a lonely space!

[xilman]

Originally Posted by swellman

If anyone is reading this, please give feedback. XYYXF factoring has been a lonely space!

I am reading.

I will PM you in a day or few. Rather distracted right now ...

Paul

[VBCurtis]

I'm game to do a little ECM to help you get individual candidates ready for NFS.

"A little" to me is something like 2000 curves at 85e7 or 12e8. I prefer to do fewer curves at higher bounds to "take a shot" at finding a bigger factor, as opposed to finishing off a t60 with B1=26e7. I just received a 5950X that I'll assemble next week, so I'll have a bunch of new compute capacity.

I also have some interest in running CADO with I=17 on small q-values as a way to boost / expand the range of NFS@home jobs. If we can gather, say, 10% of the total relations via I=17 CADO for a job we can reduce the Q-range for NFS@home by 5% or more. I doubt that C311 is big enough to need that, but maybe we try it out this summer on this XYYX-C311 to test tools etc before we rely on a CADO17 / nfs@home combo for a really tough job like a GNFS225+.

[Batalov]

Originally Posted by swellman

It’s been a long time since Andrey posted anything here, so it’s time for a 1 < y < x < 151 XYYXF factoring project update.

Belarus is in a rather dark place for the last two years (maybe more?), so our thoughts are with Andrey as well.
Prior to that, it was widely thought that Belarus was a haven for IT folks/scientists.

Andrey is a Chemistry teacher (professor?) and a computational chemistry scientist.
His birthday is coming up in a few weeks. He has not been active in Facebook for a year or more.

[swellman]

Originally Posted by VBCurtis

I'm game to do a little ECM to help you get individual candidates ready for NFS.

"A little" to me is something like 2000 curves at 85e7 or 12e8. I prefer to do fewer curves at higher bounds to "take a shot" at finding a bigger factor, as opposed to finishing off a t60 with B1=26e7. I just received a 5950X that I'll assemble next week, so I'll have a bunch of new compute capacity.

There are 303 composites which have only completed t55, so any ECM on these at higher levels is welcome. Would you like to make a pass of 1-2k curves @850M through say a dozen composites? Or the whole group of 303? Let me know what interests you and I’ll post a list. Posting all remaining XYYXF numbers in the current project seems bit unwieldy, too big for a code tag I think. Maybe as a flat text file attachment? Or pdf if people prefer.

I also have some interest in running CADO with I=17 on small q-values as a way to boost / expand the range of NFS@home jobs. If we can gather, say, 10% of the total relations via I=17 CADO for a job we can reduce the Q-range for NFS@home by 5% or more. I doubt that C311 is big enough to need that, but maybe we try it out this summer on this XYYX-C311 to test tools etc before we rely on a CADO17 / nfs@home combo for a really tough job like a GNFS225+.

There are quite a few very difficult SNFS jobs on the list, going up to difficulty 323, using diff = x*log(y) minus y*log(GCD(x,y)). And maybe some GNFS 210+ jobs which may be easier to factor than SNFS though I haven’t sought them out. Your suggested strategy sounds beneficial to any large NFS@Home sieving effort, though much ECM would need to be focused on such a candidate first of course.

[swellman]

Originally Posted by Batalov

Belarus is in a rather dark place for the last two years (maybe more?), so our thoughts are with Andrey as well.
Prior to that, it was widely thought that Belarus was a haven for IT folks/scientists.

Andrey is a Chemistry teacher (professor?) and a computational chemistry scientist.
His birthday is coming up in a few weeks. He has not been active in Facebook for a year or more.

He was always a gentleman to me, and any teacher wears a halo in my eyes.

I hope he and his are well.

[VBCurtis]

What I had in mind for ECM was any composite that you were personally taking from whatever level Yoyo did to "ready for NFS@home". If I do something like 1000 @65e7 and 800 @ 12e8 for each job as you get to it, I'll speed your efforts and together we feed the 15e queue a bit faster. Those curve counts are roughly equivalent to 6000@26e7 as far as "how much of a t60".

[ryanp]

Originally Posted by swellman

It’s been a long time since Andrey posted anything here, so it’s time for a 1 < y < x < 151 XYYXF factoring project update.

- There remain 323 composites left to be factored in the project. All have survived at least 18000 curves @B1=110e6 (t55).

Do you have a link to these 323 composites? I can hit them with some ECM at B1=26e7 or perhaps 85e7.

[swellman]

Originally Posted by VBCurtis

What I had in mind for ECM was any composite that you were personally taking from whatever level Yoyo did to "ready for NFS@home". If I do something like 1000 @65e7 and 800 @ 12e8 for each job as you get to it, I'll speed your efforts and together we feed the 15e queue a bit faster. Those curve counts are roughly equivalent to 6000@26e7 as far as "how much of a t60".

I understand now. The 3 least difficult XYYXFs are now being processed by Yoyo in a final pass of 20k @26e7 prior to NFS so I would rather just let those crunch.

But the next batch on my list are 4 composites of SNFS 270. Can you run 1000 @65e7 and 800 @12e8 on these? Your efforts would greatly speed thing up.

C217_146_130
2273736614959600102299788369214860151882556046845135920321491916509566649822233346525551566590304365550486243473007049647802844778464741732324121635941624450635826410351378351500756252260898809587472865957974314446269

C214_139_87
4195620243915888279671778712345331475268709716797825786349176460911223003450368828626222652158705793360957647290504936745524043932410659915735237866241362258201645994955743683282475290187203096154867195374771993153

C222_141_82
205268521474284689311496808503122136985981225012537353959393016862115907865885179677284356635128069465482618285245343279948507532199640274072355307560636501254976877515298927864568411335060366559573631938887842066290994757

C213_146_132
893056777816347198729161218297917289911779059070888373649385608058717811758800548718293627086572394372637058444446373481226457729536964989297099024892946685162728665191417684132376902310624589757129442005764041901

There will more in the near future.

Originally Posted by ryanp

Do you have a link to these 323 composites? I can hit them with some ECM at B1=26e7 or perhaps 85e7.

I will publish the full list of XYYXF numbers in this thread in a simple text attachment, if everyone is cool with that. If that is not palatable, please suggest another way (pdf, etc). I will publish the composite name and then the decimal form of the number, with a single space in between.

If you can keep me advised on # of curves and B1 values run it would be greatly appreciated!

[Batalov]

Tom Womack had a nice reservation/status interface for Hom.Cunninghams. That approach could be refactored for the simple list of the 323-or-bust and bookkeeppiing, maybe

[swellman]

Originally Posted by Batalov

Tom Womack had a nice reservation/status interface for Hom.Cunninghams. That approach could be refactored for the simple list of the 323-or-bust and bookkeeppiing, maybe

Love the idea but no idea how implement it. It is a nice res system. I’ll reach out.

But I’ll put up the flat text file later tonight, regardless.