Hi all,
> >There are 3 SunOS workstations in this office that sit doing little
> >more than reading and displaying files - so they can be kept busy
> >testing for primes.
>
> You can always try to compile mprime on these. The network code might need
> platform-specific tweaking but the guts of the primality testing stuff
> ought to be portable C.
MPrime's Intel assembly code will be tough to build on a Sun. Michiel Van
Loon's PrimeOS2 implementation of PrimeNet's network layer might be more
portable.
> Today, Primenet status page contains:
(cut)
> 50.314 years/day !!!
>
> It seems that a massive number of new members have sent results.
Interpretation of the CPU rate figures is dependent upon the date stamps below
the subheadings and at the top of the summary report. In the example 50+ years
accumulated over 32 hours instead of the usual 24 hours. 37.8 years/day was the
interesting figure, not far from the week average of 38.
[Gary:]
> This appears to be a glitch in the program. The "CPU years" cannot be
> greater than the "CPU yr/day".
Actually, operator intervention. As Gary later pointed out, the time stamps and
daily cumulative count can on occasion skip a day but the CPU rates remain
normalized. In this particular case, the server was offline at 06:00 UTC/GMT
(22:00 Pacific) for a snapshot test copy of its database for the new 4.0
server's final tests. It caused the server to miss its daily self-maintenance
pass that updates the stats.
(The daily CPU rate also varies cyclically throughout the day and week. To see
a graph of the daily cycle, see http://entropia.com/ips/stats.html for an
analysis of PrimeNet's first 7 months. The 7-day average CPU yr/day and GFLOP/s
rates smooth out the daily spikes and dips, and is what Entropia.com pays most
attention to.)
The current stats show an additional CPU yr/day is coming in now relative to a
few days ago:
*** Aggregate CPU Statistics, P90 Units ***
Last 7 Days Average Cumulative Today
from 99-Jan-28 06h from 99-Feb-04 06h
Test Type CPU yr/day GFLOP/s CPU years CPU yr/day
------------ ---------- ---------- ---------- ----------
Lucas-Lehmer 37.489 451.287 8.586 41.510
Factoring 1.457 17.536 0.255 1.235
---------- ---------- ---------- ----------
TOTALS 38.946 468.823 8.842 42.745
> This page also shows that the number of inactive accounts is raising
> harder then the active accounts. Since the last newsletter was from may
> 22 (or did i miss one?) it might be a good idea for a new one to wake
> all these people because not all people who are testing are members of
> this mailing list
The inactive accounts figure is not as meaningful as originally conceived. We
removed it from the 4.0 server's status page, which if you don't mind ignoring
bogus test data is at:
http://207.104.25.139/primenet/status.shtml
A few people quit GIMPS but the vast majority are simply changing account IDs
and leaving the old ones behind. In contrast to this, PrimeNet has seen a net
average of 21 people join GIMPS every day for over the last year.
-----
In other PrimeNet news, the new 4.0 server is in final configuration and testing
this week and part of next, and the PrimeNet proxy server beta is running
successfully at almost all of the test sites that signed up. We expect to cut
over to the new system right after Internet verification completes with a final
database snapshot.
Regards,
scott
(for the Entropia.com team)
