Fourteen or Bust, or ECM attack on C300+
 

As of December 2014, there were 14 composites having 300 or more decimal digits. Half of them are [url=http://www.rechenkraft.net/yoyo/download/download/stats/ecm/xy/wu_status]sheduled[/url] to run t65 by [url=http://www.rechenkraft.net/yoyo/]yoyo@home[/url]:
C305_142_139 (the only composite without cofactor)
C310_147_142
C311_148_133
C313_148_139
C316_149_132
C321_149_146 (the largest composite in the project)
C304_150_149 (the largest Leyland number in the project)

Two more are on the way to t60:
C300_141_133 (5k+ curves @260M)
C301_148_131 (20k curves@260M)

Others had been tested at least up to t55 (18k@110M):
C301_143_129
C300_145_136
C302_149_110
C307_149_120
C303_150_119

Their decimal expansion is available there: [url]http://www.primefan.ru/xyyxf/cmplist.txt[/url]

It would be fine to rain down more ECM power on them :-)

[QUOTE=XYYXF;373397]As of April 2014, there were 14 composites having 300 or more decimal digits. Half of them are [url=http://www.rechenkraft.net/yoyo/download/download/stats/ecm/xy/wu_status]sheduled[/url] to run t65 by [url=http://www.rechenkraft.net/yoyo/]yoyo@home[/url]:
C305_142_139 (the only composite without cofactor)
C310_147_142
C311_148_133
C313_148_139
C316_149_132
C321_149_146 (the largest composite in the project)
C304_150_149 (the largest Leyland number in the project)

Three more had been tested up to t55 (18k@110M):
C302_149_110
C307_149_120
C303_150_119

And the least four were brought only to t50 (7600@43M):
C300_141_133
C301_143_129
C300_145_136
C301_148_131

Their decimal expansion is available there: [url]http://xyyxf.at.tut.by/cmplist.txt[/url]

It would be fine to rain down more ECM power on them :-)[/QUOTE]

I guess most of them (thiose under 1018 bits) could be treated with GPU-ECM...

Luigi

OK, I will try raining 18432@110M on C300_145_136 using a GPU. It will take about nine days, plus maybe five days on 36 threads for stage 2.

ryanp is reserving C300_141_133.

18432@110M on C300_145_136 completed, no factor found.

Thank you Tom :)

As all of my ECM reservations are about to finish, I'd like to take C301_143_129 to t55. It will take a few weeks.

Anybody want to tackle C301_148_131? It's the last one.

I'll do it. I'd imagine a couple of weeks for me as well.

I ran 6000 B1=260M curves on C301_148_131 and will run 6000 more. That will take care of the t55, but you can run some 110M curves too.

Sounds good.

Smaller curves have a better chance _if_ the factor is in that range (50-55 digits).
And, the t55 is not complete yet, so you have a fair chance.

For the full t55, it would have taken you a core-year (e.g. 4 cores * 3 months).

That makes sense. I'm doing 8 curves at B1=110M (simultaneous with ecm.py) approximately every 20 minutes, so a quick calculation says it would take me about a month to run 18,000 curves.

Upon a second thought, you may want to switch to B1=260M curves, they will take ~50 minutes each.
This is because while your chances of getting a <t55 factor are not zero, but significantly decreased by the bayesian prior of knowing that 6000 (and some in progress) are run at B1=260M level without a factor. With 3000 more of my curves finishing, your chances will go down by the factor of 1/e (because the 3000 will complete the t55 level, and the chances of a factor will become ~ e[SUP]-1[/SUP] of the initial chances.

One thing that you would want to do is stagger the curves so the 2nd stage memory is used by 1-2 threads of the 8. Phase the curves to start at, say, two at the top of the hour, two more at :12, two more at :24 and two more at :36 and then leave them be.

[QUOTE=swellman;373913]ryanp is reserving C300_141_133.[/QUOTE]

Just an update - Ryan has run this through a gauntlet of ECM. Probably some bursts of high levels but no factors found.

At this point, he has documented 5000 curves @B1=260M with more to come.

Sean

[QUOTE=Batalov;385889]Upon a second thought, you may want to switch to B1=260M curves, they will take ~50 minutes each.
This is because while your chances of getting a <t55 factor are not zero, but significantly decreased by the bayesian prior of knowing that 6000 (and some in progress) are run at B1=260M level without a factor. With 3000 more of my curves finishing, your chances will go down by the factor of 1/e (because the 3000 will complete the t55 level, and the chances of a factor will become ~ e[SUP]-1[/SUP] of the initial chances.

One thing that you would want to do is stagger the curves so the 2nd stage memory is used by 1-2 threads of the 8. Phase the curves to start at, say, two at the top of the hour, two more at :12, two more at :24 and two more at :36 and then leave them be.[/QUOTE]

I'll give that a shot. Thanks!

I ran 20,000 B1=260M curves on C301_148_131 altogether and stopped.

Checked the Fourteen in the factordb.com ... so far it looks like a Bust. :-(

[QUOTE=Batalov;385981]I ran 20,000 B1=260M curves on C301_148_131 altogether and stopped.[/QUOTE]

So, did you run 20,000 B1=260M curves in a day or two? If so, that's truly impressive! Do you have access to lots of computers (or just processors) at a University, or something similar? I'm currently working on a C251 and have done about 26,400 B1=260M curves in 20 days. This is with 32 cores on 3 different machines (16+12+4, not all equal). I certainly wouldn't mind seeing my curve count speed up by about 20x. Which makes me wonder, did you use about 500-600 cores to run that many curves that fast?

[QUOTE=WraithX;386987]So, did you run 20,000 B1=260M curves in a day or two? If so, that's truly impressive! Do you have access to lots of computers (or just processors) at a University, or something similar? I'm currently working on a C251 and have done about 26,400 B1=260M curves in 20 days. This is with 32 cores on 3 different machines (16+12+4, not all equal). I certainly wouldn't mind seeing my curve count speed up by about 20x. Which makes me wonder, did you use about 500-600 cores to run that many curves that fast?[/QUOTE]
Batalov has some serious firepower, I don't know how much exactly, but he can do a [URL="http://mersenneforum.org/showpost.php?p=382864&postcount=9"]SNFS(diff 176) in half a day[/URL] .

[QUOTE=swellman;385810]As all of my ECM reservations are about to finish, I'd like to take C301_143_129 to t55. It will take a few weeks.[/QUOTE]

C301_143_129 survived 18000+ curves @B1=110M with no factors found. Releasing number.

yoyo@home performed 9k@850M on all of them. Still no factors.

[url]http://www.primefan.ru/xyyxf/status.html#work[/url]

Could Ryan help with B1=850M on the remaining numbers?
C301_143_129
C300_145_136
C301_148_131
C302_149_110
C303_150_119

Ryan is running ECM on C301_143_129, with some 10k curves @B1=76e8 completed so far. He is planning on running another 16k curves soon, at which point it may be safe to declare victory and flee the field.

BTW, 142[SUP]139[/SUP]+139[SUP]142[/SUP] is the only number without known factors :)

[QUOTE=XYYXF;446742]BTW, 142[SUP]139[/SUP]+139[SUP]142[/SUP] is the only number without known factors :)[/QUOTE]

Ryan is running ECM @B1=76e8 on this number, as well as C301_143_129. 16k curves at a time (alternating between the two composites). No luck so far.

Not sure how many curves are required at this level before declaring both numbers "NFS ready", but I'm thinking we are rapidly approaching that point.

[QUOTE=XYYXF;446742]BTW, 142[SUP]139[/SUP]+139[SUP]142[/SUP] is the only number without known factors :)[/QUOTE]

Ryan has started sieving this number.