A Desperate appeal! (by Richard K. Guy)... deadline is September 30, 2016

[rajula]

Originally Posted by Prime95

-- When the 260M column is marked "Done", the 800M column starts counting from 36,000.

This I did not know (or had forgotten). It explains the unexpectedly high count.

[GP2]

Reviving an old thread from 2013, since the deadline is now almost exactly a month away.

From the first post in this thread:

A Desperate appeal!! by Richard K. Guy
At the time of my 80th birthday, when John Conway and I were
writing The Book of Numbers, John bet that there would be no
further complete factorizations of Fermat numbers in the next 20
years. I bet him $20.00 that there would be. Now that my 100th
birthday is only 3 years away, I'm beginning to get a bit worried!
I'll even go so far as to split the $20.00 with anyone who
completely factors F12 or F13 or any larger Fermat number!!

(Note: The actual deadline is September 30, 2016;
please compare UPINT, 3rd edition, page 16.)

Nearly three years have passed, and it looks like Professor Guy's $20 is doomed.

I think there are two methods to try to find factors of Fermat numbers. One is sort of the equivalent of trial factoring (searching ranges of N and k). The FermatSearch sub-forum has details for reserving ranges. I'm not sure what software is used for that.

The other method is ECM testing, using mprime/Prime95.

F12 is now at the stage where ECM testing is looking for factors in the range of about 65 digits (about 215 bits). Each curve takes about 3.1 hours on a 2.3 GHz Haswell architecture.

F13 is now at the stage where ECM testing is looking for factors in the range of about 60 digits (about 200 bits). Each curve takes about 1.7 hours on a 2.3 GHz Haswell architecture.

Finding a new factor would be interesting in its own right, albeit rather improbable since you might have to run tens of thousands of curves or even hundreds of thousands before you hit a factor at this level, if one even exists. And even then, saving Professor Guy from financial ruin would require the cofactor to be prime as well.

Nevertheless, in case anyone wants to make a token effort, here are some lines you could add to your worktodo.txt (requires stopping the program first, or to avoid having to do that, just create a worktodo.add file that the program will automatically read and append to worktodo.txt the next time it does a disk write). There are 10 curves for F12 (2⁴⁰⁹⁶+1) and 18 curves for F13 (2⁸¹⁹²+1), so maybe 30 hours of CPU time for each on a modern Haswell machine. F12 needs 535MB of memory in stage two, and F13 needs 579MB.

ECM2=1,2,4096,1,800000000,80000000000,10,"114689,26017793,63766529,190274191361,1256132134125569,568630647535356955169033410940867804839360742060818433"

ECM2=1,2,8192,1,260000000,26000000000,18,"2710954639361,2663848877152141313,3603109844542291969,319546020820551643220672513"

Here is the ECM status report from PrimeNet.

[ET_]

Originally Posted by GP2

Reviving an old thread from 2013, since the deadline is now almost exactly a month away.

Here is the ECM status report from PrimeNet.

I'm in

[Batalov]

Luigi, can you keep the desperate appeal alive by reposting it on fermatsearch.org ?

Now, both W.Keller's site and prothsearch joined the choir invisible. Shuffled off their mortal coil, ran down the curtain. Their metabolic processes are now 'istory!

[ET_]

Originally Posted by Batalov

Luigi, can you keep the desperate appeal alive by reposting it on fermatsearch.org ?

Now, both W.Keller's site and prothsearch joined the choir invisible. Shuffled off their mortal coil, ran down the curtain. Their metabolic processes are now 'istory!

Done!

[pinhodecarlos]

Maybe contact yoyo. What would be the ecm standalone client flags?

EDIT: I want to help or using Prime95 or using ecm. For Prime95 I only want to use 2 of the 4 cores available from my laptop but I need someone to support me on getting the worktodo.txt setup or the prime.ini (not sure about the file names).
For ecm what are the command lines I need to setup to run some curve on those two Fermat numbers.

[ET_]

Originally Posted by pinhodecarlos

Maybe contact yoyo. What would be the ecm standalone client flags?

I already contacted them. Unfortunately, the parameters for the ECM search on Fermat numbers are out of their reach.

[pinhodecarlos]

Originally Posted by ET_

I already contacted them. Unfortunately, the parameters for the ECM search on Fermat numbers are out of their reach.

Share with me those parameters and I will try to bring some support from SETI.USA team and also from myself.

[pinhodecarlos]

Here's my doubt.

worktodo.txt content

[Worker #1]
ECM2=1,2,4096,1,800000000,80000000000,1000,"114689,26017793,63766529,190274191361,1256132134125569,568630647535356955169033410940867804839360742060818433"

[Worker #2]
ECM2=1,2,4096,1,800000000,80000000000,1000,"114689,26017793,63766529,190274191361,1256132134125569,568630647535356955169033410940867804839360742060818433"

local.txt content

OldCpuSpeed=2864
NewCpuSpeedCount=0
NewCpuSpeed=0
RollingAverage=1000
RollingAverageIsFromV27=1
ComputerGUID=ec875508223b69f7b24450ba96978009
WorkerThreads=2
Affinity=100
ThreadsPerTest=2

On the local.txt content flag in bold, I want to run one worker per core. How do I set it up? Appreciated some support.

Carlos

[GP2]

Originally Posted by pinhodecarlos

Here's my doubt.

On the local.txt content flag in bold, I want to run one worker per core. How do I set it up? Appreciated some support.

WorkerThreads=<<number of cores you have>>
ThreadsPerTest=1

[axn]

I think it might be ever so slightly more efficient to use a B2/B1 multiplier of 200, since that gives roughly equal times for stage 1 & 2 (which should be optimal according to RDS).
Anyone care to compute the expected curves for both configuration and see whether it makes any difference?