Hi Steve,
lets do more chasing ;-)
Because its a very long time ago I had a look on program codes I'm not sure
to understand all of this stuff.
But I think you are using the P-90 h/day values without changing them, don't
you?
I think the actual performance values are to be computed again to be useful,
because there are many
people upgrading and therefore ranked high but showing lower performance
values at P-90
h/day as other participants ranked nearby.
i.e. today I'm ranked at 445 with 22.882 years/148 exponents(IPS)/182,22
P-90 h/day
Currently I'm running Prime95 on three machines (2 GHz total) equal to 22,0
P-90 CPUs.
This means 528 P-90 h/day but completing one LL-test at 12M range only adds
less than five P-90 h/day.
There's a difference between 182 P-90 h/day at top producers list and 528
P-90 h/day (actual performance).
How to calculate the actual performance of an upgraded account of user X?
To do this I'm calculating the average P-90 h/day of such participants near
the rank of user X having higher values than X. The other users near X with
lower values did some upgrading recently too and therefore are to be
neglected. Computing the average out of 10 such values in front and 10 such
values behind the rank of user X works well. I called it the local average.
Using 20 such values nearby my own ranked ID (dismit) gives the local
average of 310 P-90 h/day. Of course, this isn't a very good approximation.
But the 182 P-90 h/day on the list means 58.7% of the local average of 310
P-90 h/day. Thus divide the local average of 310 by 0,587 and use the result
which is very near to the already known performance of 528 P-90 h/day.
Does anybody know a pretty formula of this algorithm? Why does it work at
all?
Have fun
Dieter Schmitt
-Ursprngliche Nachricht-
Von: "Steve" [EMAIL PROTECTED]
An: [EMAIL PROTECTED]
Gesendet: Donnerstag, 15. Februar 2001 16:56
Betreff: Mersenne: PrimeStats, Perl script for the PrimeNet Top Producers
Table
Hi Primepickers
PrimeStats is a perl script that interogates the top producers table
at: http://mersenne.org/ips/topproducers.shtml, after you've saved it to
disk, it will give you a report on participants as specified by yourself
in a
seperate data file. The script will also give you a detailed report of the
future prospects of one user ID that you supply to the script via the
command
line, this detailed report tells you how many people are in front of the
user
but going slower and how many people behind but going faster and gives you
estimates of when the user will catch the pack in front and when the
chasing
pack will catch up etc.
The script is available here:
http://www.zeropps.uklinux.net/linstuff.html
And here is a sample report that finds current details for 10 users and
gives
a detailed report for one of them (me):
Run Date: Thu 15 Feb 2001
Participants counted: 20,018
Extra details for user ID: sjlen
Position User Name CPU YearsExponentsCPU P90
Tested Hrs Per Day
1264 S18743 8.921 16225.16
3056 felipel 3.667 8 49.83
3171 sjlen3.512 7 66.59
3496 mbandsmer3.050 19 24.85
4021 Pse 2.488 5 45.91
5394 Lalo11.549 18 12.62
5981 S16318 1.319 3 28.17
6773 mage21 1.010 3 17.49
7411 S17376 0.838 2 19.11
7434 Paradoks 0.833 7 6.84
3171 sjlen3.512 7 66.59
712 people are faster than you but behind you,
at an average speed of 115.27 CPU hours per day
they are approximately 1.75 years behind you.
You will be in the center of the chasing pack in
something like 312.84 days.
825 people are slower than you but in front of you,
at an average speed of51.04 CPU hours per day
they are approximately 1.37 years in front of you.
You should be in the middle of the pack that you are
chasing in something like 763.26 days.
Any comments or suggestetions for improvements or error
reports are welcome.
--
Cheers
Steve email mailto:[EMAIL PROTECTED]
%HAV-A-NICEDAY Error not enough coffee 0 pps.
web http://www.zeropps.uklinux.net/
or http://start.at/zero-pps
3:39pm up 13 days, 17:18, 3 users, load average: 1.22, 1.18, 1.18
_
Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
_
Unsubscribe list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ