On 22 Aug 00, at 19:15, xqrpa wrote:
> 1. Forming a group with a few slow machines, then adding a slew of faster
> ones bit by bit, causes an underestimate to be reported, but one
> which gradually increases as the new machines report in.
Similarly if you run a whole lot of systems for a while then stop
most of them, your hours/day will be well in excess of the systems
apparently allocated - but falling asymptotically to the true value
as time goes on.
>
> 2. How Ms are picked for processing *seems* to influence the numbers,
> too:
> PrimeNet can assign exponents automatucally, according to a (to me)
> rather
> mysterious protocol, or numbers can be gotten using Manual Testing
> Forms.
Not quite right.
The "mysterious" protocol is simply to allocate the smallest exponent
available in each assignment type. If an assignment expires it is
made available for reallocation in a daily job starting at 0600 UTC.
If a user voluntarily returns an assignment it becomes available for
reassignment immediately.
PrimeNet credits you with CPU time if you use the automatic method to
report the result of an assignment but not if you use the manual
assignment form. It doesn't matter whether you got the assignment
using the automatic method or using the manual assignment form.
Regards
Brian Beesley
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.exu.ilstu.edu/mersenne/faq-mers.txt
