GIMPS Credit for Exponents Between 79.3 Million and 596 Million

[jinydu]

Given that the LL tests for M100,000,007 are scheduled to finish next month, I am curious as to how GIMPS credit will be assigned for exponents from 79.3 million to 596 million.

It would nice if Mr. Woltman could reply here for everyone to see.

Thanks

[Uncwilly]

Like other exponents being worked on by LMH'ers there is likely to be no credit.

[jinydu]

That would be unfortunate... Could Mr. Woltman confirm this please?

[gribozavr]

In 100MDPP I keep my private version of stats. But, as I'm the only active contributor, those stats are rather boring.

[cheesehead]

IMO credit can and will be given eventually if the records of contributions have sufficient information, such as identities of contributors, along with the amounts of work accomplished in a form that can be converted to CPU-years.

[jinydu]

Originally Posted by cheesehead

IMO credit can and will be given eventually if the records of contributions have sufficient information, such as identities of contributors, along with the amounts of work accomplished in a form that can be converted to CPU-years.

Care to speculate on a formula for how credit will be given?

[gribozavr]

jinydu, some time ago I was thinking about the problem. My solution was
http://mersenneforum.org/showpost.ph...&postcount=284

[cheesehead]

Originally Posted by jinydu

Care to speculate on a formula for how credit will be given?

For L-L, just use the current formula with the higher exponents:

(Exponent-2) * (iteration time for the FFT length), where "iteration time" is in units of P90-CPU-years,

or

(Exponent-2) * (iteration time for the FFT length) / (number of seconds in a year),
where "iteration time" is in seconds on a P90,

For trial factoring:

Originally Posted by gribozavr

jinydu, some time ago I was thinking about the problem. My solution was
http://mersenneforum.org/showpost.ph...&postcount=284

Of course there has to be a multiplier/divider to convert the TF "points" (exponent 332,192,831 = 1 point) in that thread to CPU-years.

[gribozavr]

But are CPU years comparable? For example, TF > 64bits with mfactor on 32-bit 3.0 Ghz Pentium4 would be much slower than on Athlon64 1.8Ghz. But LL would be faster on P4 AFAIK.

[cheesehead]

We already have that problem now; it doesn't start just at 79.3M.

[jinydu]

Originally Posted by cheesehead

For L-L, just use the current formula with the higher exponents:

(Exponent-2) * (iteration time for the FFT length), where "iteration time" is in units of P90-CPU-years,

or

(Exponent-2) * (iteration time for the FFT length) / (number of seconds in a year),
where "iteration time" is in seconds on a P90,

So in other words, I would need to have a P90 in order to find the iteration times, not to mention the fact that I don't know what the FFT length is for exponents larger than 79.3 million...