how would you distribute the $100k prize?

[T.Rex]

What would happen if the 10M+ digit prime Mersenne was found by using Glucas or Mlucas, instead of prime95 ?

I'm now doing some experiments with Glucas and Mlucas, measuring their performance and trying to understand and improve their scalability when run multi-threaded (POSIX threads or OpenMP) on a large (16x) NUMA machine.
Rather than using a well-known prime Mersenne, I use "manual testing" exponents I got from George.
So the probability I find 10M+ with Glucas or with Mlucas is VERY small ! But not null.

What would happen then ?

Tony

[Jushi]

Originally Posted by Xyzzy

I wish the were no prize at all. It sure would simplify things.

On the other hand, I'm sure it helps to get more people in the project. Part of the popularity of GIMPS no doubt is because of the price money.

Originally Posted by drew

Those earlier tests don't change the fact that the 10 million digit test is the 'bread winner', so to speak. It seems fair to give a disproportionatlely high payout to that particular test.

I disagree, because the person whose computer finds the prime will have been assigned at random. So, that person really didn't do anything to find the prime, he was just lucky.

[Prime95]

Originally Posted by T.Rex

What would happen if the 10M+ digit prime Mersenne was found by using Glucas or Mlucas, instead of prime95 ?

What would happen then ?

I don't think Glucas or Mlucas have any restrictions in their license agreements. GIMPS does not "own" the exponents. The license agreement is tied to the use of prime95.

Thus, you can legally claim the entire $100K. Or, your own moral code might say that you found this as part of the GIMPS community and will abide by its prize rules.

[T.Rex]

Originally Posted by Prime95

Thus, you can legally claim the entire $100K. Or, your own moral code might say that you found this as part of the GIMPS community and will abide by its prize rules.

I think I would divide the prize in at least 3 parts: my-self, the developer of XLucas, and the GIMPS community. Probably something between: 1/2 - 1/4 - 1/4 and 2/3 - 1/6 - 1/6 .
But keep in mind that the probability I find a 10M+ digit prime with XLucas is probably less than 1/200,000 ! (But it seemed to me important to discuss of this now. Just in case !)
Your opinion ?
Tony

[R.D. Silverman]

Originally Posted by Xyzzy

I wish the were no prize at all. It sure would simplify things.

It would also get rid of those who are just doing it for the money.

Call me a purist snob. I think people should participate out of
interest in the subject, and not for a reward. Or rather, the discovery
itself and accompanying recognition should be a sufficient reward.

[R.D. Silverman]

Originally Posted by Jushi

On the other hand, I'm sure it helps to get more people in the project. Part of the popularity of GIMPS no doubt is because of the price money.



I disagree, because the person whose computer finds the prime will have been assigned at random. So, that person really didn't do anything to find the prime, he was just lucky.

Everyone seems to assume that the winner will be a Mersenne prime.

I am working on a different, constructive approach. I am up to 820,000
digits. I am using a single 166MhZ PII. The prime P that I have has:

P-1 = k * P1 where P1 is a prime > P^1/3
P1 -1 = k1 * P2 where P2 is a prime > P1^1/3
p2 - 1 = k2 * P3 where P3 is a prime > p2^1/3
etc.

I am doing this with an ALL INTEGER convolution code.

[drew]

Originally Posted by Jushi

I disagree, because the person whose computer finds the prime will have been assigned at random. So, that person really didn't do anything to find the prime, he was just lucky.

As were the people who found the previous Mersenne primes. The fact is that nobody who wins 'deserves' the prize money. The only people who really deserve anything are those who maintain GIMPS and primenet. It's like a raffle. If your lucky ticket is drawn, you're the winner. Let's face it, the 10,000,000 digit prime is the one that will get all the attention. If you're going to give prizes to 'raffle winners', it might as well have the allure of a big payout.

Drew

[T.Rex]

What is the goal of the GIMPS ?
Find new Mersenne primes. Break records. "Make Math History". Not get money.

What is the goal of the prize ?
Drive more people to Mathematics and to Number Theory and "contribute to solving huge scientific problems involving massive computation".

If the rules would give all the money to the GIMPS Project and to people who have written the extraordinary programs that enable us to prove prime a Mersenne number, I'm not sure so many people would have joined the project last year.

We will have a real idea of the impact of the prize to GIMPS contributors once the 10M+ prime is found, when people will see that testing a Mersenne number around 100M+ digits requires about 7 years of computation, 7/7 24/24 with a fast Pentium4 3.8 GHz, and that thus we will have to wait more than 10 years before reaching the next $150,000 EFF prize seems possible.

So, since the 10M+ prime will be a unique event in a decade, we should prepare our-selves to promote this news really widely.
As an example, the NewYorkTimes talked about M42. Not about M43. Though I've tried to warn them 3 or 4 times.

Tony

[xilman]

Originally Posted by R.D. Silverman

It would also get rid of those who are just doing it for the money.

Call me a purist snob. I think people should participate out of
interest in the subject, and not for a reward. Or rather, the discovery
itself and accompanying recognition should be a sufficient reward.

Fair enough: you are a purist snob.

My views on prize money for achievements in computational number theory are well documented in easily obtainable sources, including the Mersenne forum.

It's a rather interesting irony that I've won prize money from competitions for which you did the work for your previous employers.


Paul

[Zeta-Flux]

While I didn't join GIMPS for any prize money, I can definitely say that the settings I put on the program were motivated by prize money (i.e. the setting is *not* "do what makes the most sense for your computer").

Personally, I like the way George is distributing the money if he wins. Maybe a little less could go to the finder of the 10 million digit prime, and the extra could go to future finders.

Once a 10 million digit prime is found (whether here or elsewhere) I will probably start running a different program (like 17 or Bust, or Odd Perfect stuff).

[jinydu]

Originally Posted by T.Rex

We will have a real idea of the impact of the prize to GIMPS contributors once the 10M+ prime is found, when people will see that testing a Mersenne number around 100M+ digits requires about 7 years of computation, 7/7 24/24 with a fast Pentium4 3.8 GHz, and that thus we will have to wait more than 10 years before reaching the next $150,000 EFF prize seems possible.

You think it will be 10 years before a 100 million digit prime is found?