On 12 May 2001, at 10:22, George Woltman wrote:
> I did this yesterday. Not all exponents are triple checks, some were exponents
> that were never added to the servers list of exponents to double-check.
It so happens I grabbed 7 DC assignments around 4.9 million this
morning. In George's hrf3 database file (the version dated May 6th),
two of them have one entry, two have two entries and three have three
entries. Obviously this is a small sample, but this appears to
suggest a high rate of submission of results with mismatching
residuals. In the case of each of the three exponents with three
completed LL tests, one of these was done using an old version which
did not support offset whilst the other two were done with versions
which do support variable offset. The chance of offsets matching
(thus making the double check invalid even if the residuals match) is
rather small, so the reason a third or fourth run is neccessary is
very probably because results with mismatching residuals are being
submitted for some reason. Most probably random glitches?
>
> To make matters worse, Primenet was not originally designed to
> hand out double-checking assignments. This has resulted in some minor
> glitches, especially when double-checking and first-time checking ranges
> overlap.
>
> Scott/Entropia gave me a simple tool to do some remote server database
> maintenance. I can schedule triple-checks by essentially telling the server
> to forget about the double-check result it already has.
Presumably what happens is as follows:
(a) if a _first_ test result is submitted in this range, PrimeNet
"chalks it up" as a "cleared exponent"; then George recycles the
exponent for double-checking, and it loses its "cleared" status;
(b) when a DC assignment completes, again PrimeNet marks the exponent
as "cleared"; if it turns out that the residuals don't match, or (on
the odd occasion) the double check is invalid because the offsets
match, the exponent has to be recycled and again the "cleared" status
count drops.
>
> This is not a setback for the project. To get accurate counts of the master
> database, visit http://www.mersenne.org/status.htm. There you will see
> that 157,484 exponents have been double-checked and 127,720 exponents
> have been tested once.
Yes, I fail to see how _any_ completed assignment can possibly set
the project back!
Regards
Brian Beesley
1775*2^332181+1 is prime! (100000 digits) Discovered 22-Apr-2001
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
