On Sun, Jun 13, 1999 at 04:35:36PM -0700, Mersenne Digest wrote:
>Should do, however sometimes (for a reason I don't know) it's
>anything up to a day old.
I didn't find myself there at all (neither full name, nor username). And
I started this before PrimeNet became standard.
>Actually it's irrelevant what speed your processor is working at so
>far as credit is concerned. George Woltman (the author of the
>software) benchmarks the program using a "standard" system, which
>happens to be a P90 he has lying around somewhere (was probably state-
>of-the-art when he got it).
I've talked to him lately. The P90 is upgraded now, and Prime95 v19 will
provide a benchmark based on something newer (like his PII-400). It is
already reporting 0.3secs/iter on my PC, where it should be 0.2...
(Thankfully, the `moving average' function saves the estimates.)
>Conversely AMD
>K6 chips aren't as good as you'd expect (no pipelining in the FPU at
>all), and Cyrix chips are plain miserable.
K6s and Cyrixes are relatively good at integer maths -- let them factor
with the integer algorithm, would by my suggestion.
---snip---
>on the C400 making the speed comparison every more impossible:)
For benchmarks, turn to the benchmark pages on the GIMPS homepages :-)
---snip---
>> This is a fun project
>It is also to be useful.
I think George's original idea of it was fun.
---snip---
>btw, I disagree with the current policy of not letting slowish machines like
>this do LL tests, this machine cranks out a exponent every 60 days, has been
>an old faithful, I see no reason to retire it from the test, hence have not
>updated it to the current software, its still running 16.x.
You can choose exactly what assignments you want, even with 18.x. There are
checkboxes to override the automatic selection.
---snip---
>I thought that if no check-in was done in 60 days, the number was put back in
>the pool.
No, if no check-in has been done in 60 days _after the exponent was expected
to complete_, it is put back into the pool. Of course, once in a while,
the software will report `new expected completion dates' to the server, and
this date will be moved.
---snip---
>The owner is "dsh21". Mail him and ask why.
[...]
>The owner is "koma". Mail him and ask why.
[...]
>Why don't you mail "andres" and ask if it is abandoned first?
I have to agree on this policy. One mail, give them a week or so to reply,
and if they don't, take the exponent. (Didn't IPS have an automatic system
at one point? If you look in the readme file, it said it didn't work at
expected, so they took it out.)
---snip---
>This is an excellent point, I think we ought to take notice. It's
>certainly p*ss*d me off a bit with the Proth project to find that,
>out of 4 primes I've discovered, 2 are "rediscoveries" of numbers
>that other people have been working on without reserving ranges via
>the perfectly satisfactory facilities provided.
We should certainly prevent things like this from happening to GIMPS...
For now, we can rely on the users, I hope. Suggested policy: One mail,
one week.
---snip---
>Some people that are out of contact may be using the buggy version 17, and
>their work is wasted.
Probably not -- if they _have_ an Internet connection and have downloaded
v17 with it, they will probably also have received the v17 bug warning, and
they will probably at least make their computers report in now and then...
---snip---
>Also, I'm going to quit first time LL testing. Call me impatient, but I don't
>want to wait until early July for my exponent to finish, thus I'm going to
>switch to double-checking.
Now, isn't it great that v17/v18 actually _has_ something to do for impatient
users, namely double-checking? Remember, everybody: Nobody forces you to do
double-checking. It's _your_ option.
/* Steinar */
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
