>From: "Brian J. Beesley" <[EMAIL PROTECTED]>
>To: "Nathan Russell" <[EMAIL PROTECTED]>
>CC: [EMAIL PROTECTED]
>Subject: RE: Mersenne: The recent popularity of Factoring
>Date: Sun, 26 Mar 2000 18:57:51 -0000
>
> > What about this: 300 assignments per month. Going on vacation? No
> > problem.
>
>If you're running say a dozen systems then 12 * 300 = (IMHO) too
>many.
Then make it the lesser of 300 assignments per month per machine, or
100/machine-month divided by the base-10 log of the number of machines on
the account. Even someone with 1000 machines running could still do an
assignment a day on each (well, not quite, due to the length of a month, but
I'll round). This, of course, raises the question of how to account for
multi-processor machines. The machine with 2 300 mhz P2 processors can be
safely ignored, but the large university that wants to run mprime in its
free time on its Beowulf cluster can't.
>
>If you're going on vacation, why not feed it some assignments which
>will take a reasonable time each to complete?
There is a point there, but I'm sure there are a few folks who just enjoy
the thrill of finding a factor. There is, we must admit, a pleasure of
discovery to seeing
"Mxxxxxxx has a factor: yyyyyyyyyyyyyyy" several times a month that is
utterly absent from seeing
"Mxxxxxxx is composite" only monthly.
>
> > Just so we know, how long would it take a 1.5 Ghz Athlon to factor 100
> > assignments through 64 bits?
>
>How big are the exponents? The time required will be inversely
>proportional to the exponent, though there comes a point where larger
>exponents should be factored deeper.
>
>I'm not sure there is much to choose between Athlon and PIII for
>factoring speed. Take the figure for a 500 MHz system and divide it
>by 3, that will be fairly close.
>
>Personally I think GIMPS/PrimeNet is about spending most of the CPU
>time crunching numbers rather than communicating with the server.
>IMHO if your system is taking less than about a day to complete a
>particular type of assignment, you should switch to something better
>suited to your system. There is a great deal of trial factoring to 69
>bits or more required for 10 million digit range exponents, I reckon
>these would take something of the order of a week each on the (so far
>hypothetical) 1.5 GHz system.
To my knowledge, the current versions don't allow us to select
factoring-only assignments on the higher exponents. It certainly is an idea
that I feel deserves some discussion, though.
>Regards
>Brian Beesley
Regards,
Nathan Russell
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
