(Preying for) World Record P-1!

[lisanderke]

I realize now that it would have made more sense to name this thread '(Re-)doing P-1 on small exponents' (with 're-' in brackets since there are also small exponents that have not had P-1 done before). This was my original intention, doing P-1 on exponents that had not had much (if anything) done before. With this in mind, I do believe it makes sense to expect that any factors found have a good shot at being the biggest of their kind. Though, portraying it as if finding a "world record" was the goal is indeed not right. Or, as you put it eloquently,

No talk of "world records" - but likes and makes others like factors. This is where world records really will come - as a byproduct, with a lot of diligent work.

I agree that it was wrong of me to give the thread its current title. You could have simply said that and suggested another name, if that is what you're bothered with.

[Prime95]

Originally Posted by Batalov

Pretentious?

or perhaps a simple marketing ploy to attract interest.

Think about it GIMPS does the same thing. "Join up - find world record prime, become famous!"
Sells a lot better than "Greatly increase your electricity bill for a miniscule chance of success finding a prime that few will care about"

Now, back to the topic at hand...
Next version of 30.8 has optimizations to make increasing a stage 1 bound faster (maybe 20 or 25%).
I'm testing a small improvement to stage 2 for big polys too - a modest 6.5% on the M79147 run I'm testing now (B1=1B, B2=20T)

[Batalov]

Originally Posted by lisanderke

I realize now that it would have made more sense to name this thread '(Re-)doing P-1 on small exponents' (with 're-' in brackets since there are also small exponents that have not had P-1 done before). ...
I agree that it was wrong of me to give the thread its current title. You could have simply said that and suggested another name, if that is what you're bothered with.

Good start. But why would anyone suggest a change of title if that title would contradict the contents of the thread. You don't suggest to rewrite the whole thread, I hope?
The title is the least of the problems. The title is a start.

Originally Posted by lisanderke

This was my original intention, doing P-1 on exponents that had not had much (if anything) done before.

There is a problem hiding right here and you don't see it. P-1 (and very large ones) were in fact done on them, - you simply don't know about that. There is a lot of folks (that not-to-be-mentioned too often) Paul Z had done tons, Ryan P had done much, SSW did monstrous tons of it. They keep their results elsewhere. Why? Because inserting data into GIMPS database had not been easy for decades when they did that. (I know that from George, first hand.) So they didn't. So, you assume that work has not been done (but it had), and the rest of the arguments falls apart (looking at it from a different angle i.e. bayesian-statistically, your prior is way off, so your posterior* will be not what you expect). If you will collect your future results (I hope that you will continue what you started) - then the statistical analysis of then will show that very fewer than expected results will be found.

That is why I keep repeating like a broken record: you are excited about your new idea. I get it! Next step is do research, what has been already done? what could I possibly miss about what has been already done? "An ounce of research will save a ton of resources" / "“An ounce of prevention is worth a pound of cure."

There is a very well known trend in general (don't think just of GIMPS) - negative results are much harder to find in prior art. In science, people have trouble publishing negative findings - so they don't. In applied science, people might have tried something and have found nothing so that also stays only in their logs.
________
*I swear that I am not swearing there. These are proper words.

Originally Posted by Batalov

Good start. But why would anyone suggest a change of title if that title would contradict the contents of the thread. You don't suggest to rewrite the whole thread, I hope?
The title is the least of the problems. The title is a start.


There is a problem hiding right here and you don't see it. P-1 (and very large ones) were in fact done on them, - you simply don't know about that. There is a lot of folks (that not-to-be-mentioned too often) Paul Z had done tons, Ryan P had done much, SSW did monstrous tons of it. They keep there results elsewhere. Why? Because inserting data into GIMPS database had not been easy for decades when they did that. (I know that from George, first hand.) So they didn't. So, you assume that work has not been done, and the rest of the arguments falls apart. If you will collect your future results (I hope that you will continue what you started) - then the statistical analysis of then will show that very fewer than expected results will be found.

That is why I keep repeating like a broken record: you are excited about your new idea. I get it! Next step is do research, what has been already done? what could I possibly miss about what has been already done? "An ounce of research will save a ton of resources" / "“An ounce of prevention is worth a pound of cure."

There is a very well known trend in general (don't think just of GIMPS) - negative results are much harder to find in prior art. In science, people have trouble publishing negative findings - so they don't. In applied science, people might have tried something and have found nothing so that also stays only in their logs.

Caustic wit with good content. A Scoville rating of 850k.
Re-reading certain threads (historical records a la Galaxy Quest) and posts and comparing them with the content of the newer ones gives a good recipe..sketch..outline..of a methodology to implement and then some. Hardware isn't everything though and solid theory* is prevalent/intrinsic within those posts.
As an afterthought, solidity based on mathematical proof rather than something like phlogiston. Dephlogistication is an ongoing exercise of mine.

[VBCurtis]

Originally Posted by jwaltos

Caustic wit with good content. A Scoville rating of 850k.
Re-reading certain threads (historical records a la Galaxy Quest) and posts and comparing them with the content of the newer ones gives a good recipe..sketch..outline..of a methodology to implement and then some. Hardware isn't everything though and solid theory* is prevalent/intrinsic within those posts.
As an afterthought, solidity based on mathematical proof rather than something like phlogiston. Dephlogistication is an ongoing exercise of mine.

What are you trying to say? This post seems devoid of content. What methodology do you have in mind that you refer to, to implement?

Originally Posted by VBCurtis

What are you trying to say? This post seems devoid of content. What methodology do you have in mind that you refer to, to implement?

"Caustic wit with good content. A Scoville rating of 850k.

Re-reading certain threads (historical records a la Galaxy Quest) and posts and comparing them with the content of the newer ones gives a good recipe..sketch..outline..of a methodology to implement and then some. Hardware isn't everything though and solid theory* is prevalent/intrinsic within those posts.
As an afterthought, solidity based on mathematical proof rather than something like phlogiston. Dephlogistication is an ongoing exercise of mine."

Perhaps it may seem devoid of content but that depends upon perspective. I'll explain myself to you:
Certain people within this forum are very good at expressing dry wit, implicit and explicit and sometimes at a level of abstraction greater than the originating post..ie data abstraction. I appreciate a good sense of humour as well the burn, hence my reference to Scoville. The Galaxy Quest reference was stated because it's a great parody rooted in interpretation.

On a serious note, you've been around this forum for awhile and you've seen posts by frmky, Gerbicz, xilman, Batalov, Greathouse, ewmayer, Sardonicus, LaurV. These are some names off the top of my head where their posts regarding hardware, firmware and software implementation, development and execution have helped me through the years. Ancillary subjects such as chess and astronomy have also interested me. Rogue's listing of science news, Kriesel's reporting on GPU's performance; Penne, Buhrow, JasonP have also made this forum a "must check" because of their excellent software. Many aspects of "number crunching" have been discussed from backdating hardware (virus proofing), back-dating operating systems (some awesome software just won't work on a newer linux realease) to CUDA and OpenCl code discussions which have helped me out here and there as well as advanced cooling mechanisms.
Nick's thread is always worthwhile and Sardonicus has helped me find a paper or two I couldn't locate online.

All of the above have helped me develop a methodology in approaching a question, understanding what that question is and appreciate what it is that I'm looking at. This forum helps people like myself do that, to develop an understanding and an appreciation of the "state" of a question and my (or yours for that matter) ability to resolve it. Phlogiston, N-rays, cold-fusion (there are books on these farces/delusions) as well as others on mathematical cranks and others on those with psychological issues.

The methodology that I have in mind is that you must be capable of proving whatever it is you state as fact, at the very least a concise empirical statement. This all begins with an awareness. This means learning everything you can about what the question requires you to know. This is a stepping stone to asking questions that may not have asked. This is difficult and requires a good "architect-like" imagination. I've worked on certain questions for decades..perhaps the mathematical concepts exist or perhaps not to solve such conundrums but I'll be damned if I'll be waiting patiently for something to be published while I'm still capable of thinking along those same lines.

Like a game of poker when the stakes become seriously "un-fun." Unless you're able to play the game as it should be played then be able take of yourself afterwards you shouldn't be in the game to begin with, by this I mean that a reputation is built on professionalism which is based on a track record. My track record isn't here. I hope I was able to convey my methodology as always being actively prepared and pro-active and to address any shortcomings immediately and thoroughly. Try skydiving where you need to cut away to your reserve chute or experience a regulator failure at depth while scuba diving. Things happen..scouts motto..be prepared.

I echo the sentiment below.

[Zhangrc]

Enough, enough.
Remember this one.
What we should do now is to really find a factor to prove the project is feasible!

[lycorn]

Originally Posted by lisanderke

You could have simply said that and suggested another name, if that is what you're bothered with.

Seconded.

[Prime95]

My first find for this project!

{"status":"F", "exponent":79259, "worktype":"P-1", "factors":["5397816232739964621072948142158502627401583"], "b1":1000000000, "b2":1000000000, "fft-length":4096}

[lisanderke]

Originally Posted by Prime95

My first find for this project!

{"status":"F", "exponent":79259, "worktype":"P-1", "factors":["5397816232739964621072948142158502627401583"], "b1":1000000000, "b2":1000000000, "fft-length":4096}

Awesome! That's quite a big find as well :)


I'll be taking 41k and 42k up to B1=1B too. That leaves us with the following reservations currently:
George Woltman - 79k 80k 81k
Lisander Viaene - 40k 41k 42k

[ATH]

I did a big P-1 run with GMP-ECM on M4933 with the 10 known factors removed:

(2^4933-1)/29599/3637821119/25117652081/35368771391/169745595529/90043391515951/740782867588969631/359805386252506578683471/47332662224988461128710060020435516120344835557993/1552544580255765059978618542093161661106227966313804593681333703

P-1: B1=200*10⁹ B2=5*10¹⁶

I attached the stage1 file if anyone will ever want to take it further. Stage2 can be done in stages so can be continued from 5*10¹⁶.
You just have to write for example (M4933.txt has the line above with all the factors):
ecm.exe -v -pm1 -maxmem 16000 -resume M4933pm200.txt 200e9 50000e12-50100e12 < M4933.txt


I did not add the limits I did to the primenet server, it will just generate a ridiculous number of Ghz-Days. Unless the information can be added manually.