Mersenne Digest         Friday, March 10 2000         Volume 01 : Number 704




----------------------------------------------------------------------

Date: Tue, 07 Mar 2000 21:22:42 EST
From: "Nathan Russell" <[EMAIL PROTECTED]>
Subject: Mersenne: Dumb Newbie Question

Hello everyone,

I saw somewhere on the PrimeNet page that factoring time is weighted lower 
in the top producers list in order to encourage an ideal ratio of "about 10 
factoring assignments for 1 LL test".  This implies that 9 out of 10 
factoring assignments do find a factor.  I have completed 3 assignments to 
date without finding a factor.  Am I just unlucky?

Thanks for your time.

Nathan
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Tue, 7 Mar 2000 19:38:25 -0700
From: "Aaron Blosser" <[EMAIL PROTECTED]>
Subject: RE: Mersenne: why stay off the top producers list?

> > I'm glad you're doing it the "right" way, compared to the way I
> did it. :-)
> > http://www.sciencenews.org/20000304/bob1.asp

> Aaron, its *because* of your experience that I am going the
> slow legal way.  {8-]  Otherwise I mighta just sent out GIMPS
> as an enclosure and thought little of it.  I never did really get much
> management attention until I pointed out that a clever programmer
> could *already be using* spare CPU cycles on our machines and
> we wouldnt even know it, unless we had our own, well controlled
> background process to keep tabs on that.
>
> Those of you who work in big companies, feel free to give your
> IT manager a few sleepless nights with that line of reasoning, then
> hand her your clever solution...  {8^D  spike

Glad I could help.  I really don't want to see *anyone* go through what I've
gone through.  That's why when Ivars Peterson asked if I wanted to make any
comments for his article, I basically just advised anyone who would listen
that they had darn better go through the proper channels.

Through my contract work, I can't count how many times I saw people who
installed their own software onto their work PC.  For the most part, whether
the employees had local admin access or not (or on Win9x boxes), there was a
written policy that stated that no "outside" software could be installed.
Period.

I just happened to be in the unfortunate situation of breaking this rule,
not just on one computer but on many.

So...even if you have installed something you consider trivial, if you did
it without authorization and your company has a policy about that, that
qualifies as "exceeding authorized access" and of course, as I've found out,
damages do not have to be physical damages, or "making the machines run
slow", but can include the cost to remove the software.  So if some tech has
to come uninstall that software, and that tech gets paid $15 bucks an hour,
they can charge you for it.  Just try and keep the "damage" amount under
$1000 dollars to avoid prosecution under federal law. :-)

On that note...best wishes to all! :-)  I don't mean to be such a downer
about all that, but really, if anything, US WEST's reaction just goes to
show how uptight some companies are about things.

Aaron

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Tue, 7 Mar 2000 19:49:58 -0700
From: "Aaron Blosser" <[EMAIL PROTECTED]>
Subject: RE: Mersenne: Dumb Newbie Question

> I saw somewhere on the PrimeNet page that factoring time is
> weighted lower
> in the top producers list in order to encourage an ideal ratio of
> "about 10
> factoring assignments for 1 LL test".  This implies that 9 out of 10
> factoring assignments do find a factor.  I have completed 3
> assignments to
> date without finding a factor.  Am I just unlucky?

Sometimes you find a bunch all in a row, sometimes you trial factor a couple
dozen and don't find any.

I have about 5 machines doing factoring, and for a while, I wasn't finding
any, but just in the past week I've found something like 4.  It all just
depends...

Aaron

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Tue, 07 Mar 2000 23:56:11 -0600
From: Ken Kriesel <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: Large-Exponent QA

At 07:54 PM 3/7/2000 -0000, Brian J. Beesley wrote:

(many detail comments deleted, wherein Brian sets the record straight)

>I've just put the file back up. Sorry, it got removed accidentally 
>when I upgraded the server hardware about three months ago.

Thanks, that's useful.

>[ Ernst Mayer comments ]
>> >So, it doesn't look like testing exponents above ~40M is going to
>> >be practicable any time soon, where I mean doable in a year or less. But
>> >since the vast majority of GIMPS first-time LL tests won't even be
>> >approaching 20M for some years yet, having one or two double-checked
>> >exponents in each subrange below 39M would seem sufficient for the next
>> >couple of years.
>
>I agree. Of course, anyone who thinks it's _fun_ to tie up a system 
>for several years running a LL test on a larger exponent is quite 
>welcome to do so, so far as I'm concerned!

Well, I'm one of those folks who regards it as much more fun to have
cpus fully occupied than idle, and believes the scouts should be well
ahead of the army.

Ken said, in regard to partially completed runs:
>> I'd like to see them get cpu credit, but I am not in a position to
>> guarantee it.
>
>George seems to add the CPU credit from QA tests to his records. So 
>far as PrimeNet is concerned, we (deliberately) don't use PrimeNet to 
>communicate results; in any case PrimeNet doesn't recognize that we 
>own the QA assignments (some of them are actually triple-checks & the 
>rest are outside currently active ranges), which is why they "don't 
>count". This doesn't bother me, but it should be easy enough to fix.

So far as I know, George treats full LL tests from the QA group like any 
other manually reported results; credit is given.


>> Most of the effort would fall on someone else, and I'd rather see George
>> and Scott doing other things than the bookkeeping of apportioning credit
>> by iteration count and exponent size and checks of usable save files.  To
>> keep the minimum contribution sizable, I ask the volunteers to commit to
>> at least a half-PII-400-year; large contributions are more likely to
>> justify crediting the work or a partial large exponent.
>> 
>I'd strongly reccomend anyone running PrimeNet 10 million digit 
>assignments to place the following line in their prime.ini file:
>
>InterimFiles=1000000
>
>This will cause the interim residual to be written to results.txt 
>every 1000000 iterations. At the same time, an extra save file will 
>be written. The idea is that, when double-checking is scheduled on 
>this exponent, any error can be found without neccessarily having to 
>complete the whole run to find it, thus saving time. Also, in the 
>event that a prime is found, with a set of interim save files we will 
>be able to verify the result much more quickly by running 1000000 
>iterations on thirty odd systems in parallel.

Personally, I'd prefer 2000000 for the bigger exponents, but that's
personal taste.
Ideally, in a future version of prime95 and primenet, the interim residues
will be
handled over IPS and include some tiebreaker strategy.  The observed trend is
for LL tests to be less reliable with increasing exponent, due to
increasing runtime,
so an automated way of detecting and handling mismatches in interim residue 
becomes increasingly important as the average available exponent increases.

>If you're worried about disk space, delete the extra save files; the 
>interim residual has considerable value in itself.
>
>Also could anyone dropping out of a large exponent run (including 
>PrimeNet 10 million digit range assignments) please send in a copy of 
>their last savefile, so that work completed isn't lost. My server has 
>lots of space available for this task.

I've made the standing offer before, of anyone with large exponents & wanting
to quit, that the QA group will take them on for completion.
Brian's ftp server is an ideal drop point.


Ken

>
>Regards
>Brian Beesley
>
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 11:46:36 -0000
From: "Brian J. Beesley" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Dumb Newbie Question

On 7 Mar 00, at 21:22, Nathan Russell wrote:

> I saw somewhere on the PrimeNet page that factoring time is weighted lower
> in the top producers list in order to encourage an ideal ratio of "about
> 10 factoring assignments for 1 LL test".  This implies that 9 out of 10
> factoring assignments do find a factor.

No. We want to balance time to run a LL test against time to run 
factoring divided by the probability of finding a factor. So the 
implication is that about 9 out of 10 factoring assignments _fail_ to 
find a factor.

> I have completed 3 assignments to
> date without finding a factor.  Am I just unlucky?

I would say that your chance of _not_ finding a factor in N 
assignments should be approximately 0.9^N - for N=3 this evaluates at 
0.729. However, random events like finding factors, arrival of buses, 
airplane crashes etc. always seem to arrive in bunches. I've had a 
P100 solidly crunching its way through factoring assignments for a 
good while now - it completes about two a week - once it went about 
three months without finding any factors, once it found three in a 
fortnight.


Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 15:29:55 +0100 (CET)
From: Henrik Olsen <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Dumb Newbie Question

On Wed, 8 Mar 2000, Brian J. Beesley wrote:
<snip>
> 0.729. However, random events like finding factors, arrival of buses, 
> airplane crashes etc. always seem to arrive in bunches. I've had a 
finding factors only seems to come in cluster, but the other two does
haver mechanisms that does make them actually cluster up.

Bus arrival clustering is cause by a well known positive feedback
mechanism since any difference in the interval between busses will cause
more passenres to arrive at the stops before the late bus, making it even 
later.  Only way to prevent this is to include pauses in each bus's
schedule, but traffic plannners seem to be too stupid to understand this
and tries to do it with tighter schedules instead.

I would expect a feedback mechanism on plane crashes as well.
If there haven't been crashes for a while, inspection becomes lax, crashes
happen and after a delay inspection tightens up again.  It's the deay that
causes clustering here.

- -- 
Henrik Olsen,  Dawn Solutions I/S       URL=http://www.iaeste.dk/~henrik/
Many are my names in many countries. [EMAIL PROTECTED] among
the Elves, [EMAIL PROTECTED] to the Dwarves; [EMAIL PROTECTED] I was
in my youth in the West that is forgotten, in the South [EMAIL PROTECTED], 
in the north [EMAIL PROTECTED]; to the East I go not.
                                                   With thanks to Tolkien


_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 15:17:31 -0500 (EST)
From: Lucas Wiman  <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Dumb Newbie Question

> On Wed, 8 Mar 2000, Brian J. Beesley wrote:
> <snip>
> > 0.729. However, random events like finding factors, arrival of buses, 
> > airplane crashes etc. always seem to arrive in bunches. I've had a 
> finding factors only seems to come in cluster, but the other two does
> haver mechanisms that does make them actually cluster up.
> 
> Bus arrival clustering is cause by a well known positive feedback
> mechanism since any difference in the interval between busses will cause
> more passenres to arrive at the stops before the late bus, making it even 
> later.  Only way to prevent this is to include pauses in each bus's
> schedule, but traffic plannners seem to be too stupid to understand this
> and tries to do it with tighter schedules instead.

With regards to factoring, if prime finds a factor then it can start
on another one more quickly increasing the factors found per time.

- -Lucas Wiman

> 

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 20:30:46 +0000
From: "Steinar H. Gunderson" <[EMAIL PROTECTED]>
Subject: Mersenne: Re: Dumb Newbie Question

On Wed, Mar 08, 2000 at 03:17:31PM -0500, Lucas Wiman wrote:
>With regards to factoring, if prime finds a factor then it can start
>on another one more quickly increasing the factors found per time.

I thought it continued factoring (at least to a certain level), to
get more statistical data?

/* Steinar */
- -- 
Homepage: http://members.xoom.com/sneeze/
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 21:26:18 -0000
From: "Brian J. Beesley" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Dumb Newbie Question

On 8 Mar 00, at 15:29, Henrik Olsen wrote:

> finding factors only seems to come in cluster, but the other two does
> haver mechanisms that does make them actually cluster up.

I was using "random" in a loose sense, in the sense that I'd loosely 
refer to the digits of the decimal expansion of pi as being random. 
They may indeed pass all known tests for random sequences, but 
nevertheless form a very definitely non-random sequence.

(a) I think I've mentioned the theory of bus clumping on this list 
before.

(b) It's more than likely that aircraft accidents seem to be more 
densely clustered than they really are. In the wake of a major 
disaster, any minor incident involving the same operator or aircraft 
type is likely to be reported, whereas normally the media might not 
be interested in an incident causing very little damage and no 
casualties. 
In fact you can bet that, once a fault comes to light (whether caused 
by design, maintainance or operator error), extra checks are put in 
place very quickly. My guess is that the DC9/DC80/Boeing 717 family 
probably have the _safest_ horizontal stabilizer systems in the air 
at the moment, following the recent loss of the Alaskan aircraft off 
California.

(c) the fact that we can't predict beforehand which Mersenne numbers 
yield easily to trial factoring doesn't mean that there isn't a non-
random mechanism in effect. It might be that we're too dumb to have 
found it yet!


Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 18:08:37 EST
From: [EMAIL PROTECTED]
Subject: Mersenne: Mlucas: Concatenating worktodo files

Siegmar Szlavik writes:

>I have some bad news: I think I found a bug in Mlucas.
>I also have some good news: it is not related to the test itself. :)
>
>I had 2 unfinished exponents with savefiles and everything.
>I copied all the stuff to a directory and started Mlucas, which
>continued the work on the first exponent in worktodo.ini. So far so
>good, no problem. But: after finishing the first exponent it restarts
>the computation of the second one, instead of continuing the previous
>work :( all the work done so far was lost :( Is it possible, that you
>only check the first exponent for old work?

Hi, Siegmar:

Thanks for the bug, er, I mean, "new feature!!", report.

Indeed, Mlucas only checks for old savefiles upon program initiation-
I hadn't considered the case of people concatenating worktodo files
for several partially finished jobs. It just goes to show: for any
program of even modest complexity, people will always find ways to
use it that one didn't anticipate.

For those of you using Mlucas who in future encounter this scenario,
the short-term workaround is to move the savefiles for exponent #2
(or beyond) into the working directory, but don't actually add
exponent #2 to the worktodo.ini file until #1 finishes. (I.e. let
the current run complete and the program stop.) Alternatively, put
the exponent #2 savefiles into a separate directory, create a new
worktodo.ini file there, and start a second parallel run. (But that
is preferable only if the two jobs don't conflict with each other
too much, i.e. if the aggregate runtime needed for both LL tests
run in parallel isn't significantly more than to do them sequentially
would require.)

I've added this to my TODO list for v2.7a - it will check for old
savefiles whenever a new exponent is gotten from worktodo.ini.

Sorry about the lost iterations, Siegmar.

- -Ernst

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Wed, 8 Mar 2000 21:58:46 -0800
From: "Scott Kurowski" <[EMAIL PROTECTED]>
Subject: Mersenne: RE: Mersenne Digest V1 #703

> George seems to add the CPU credit from QA tests to his records. So
> far as PrimeNet is concerned, we (deliberately) don't use PrimeNet to
> communicate results; in any case PrimeNet doesn't recognize that we
> own the QA assignments (some of them are actually triple-checks & the
> rest are outside currently active ranges), which is why they "don't
> count". This doesn't bother me, but it should be easy enough to fix.

If the exponent is not assigned to someone else, the server automatically
registers your assignment and starts tracking it normally, including credit
for work done.  You are welcome to hook up your QA work to PrimeNet.

This is regardless of the exponents being within an open range or not.
(Safety note: This feature is currently switched on [and has been on since
April 1997], but can be toggled off if things get out of hand with up-range
exponent squatting.)

I originally added this capability to allow manual testing folks with
exponents in progress to start using PrimeNet and automagically transfer
tracking to the server for George.  At first it got a lot of mileage, but is
now seldom used as most new GIMPS members immediately sign up to use
Entropia's network.

Regards,
scott

[EMAIL PROTECTED]
Entropia.com, Inc.



_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Thu, 9 Mar 2000 07:25:02 -0000
From: "Brian J. Beesley" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: Dumb Newbie Question

On 8 Mar 00, at 20:30, Steinar H. Gunderson wrote:

> On Wed, Mar 08, 2000 at 03:17:31PM -0500, Lucas Wiman wrote:
> >With regards to factoring, if prime finds a factor then it can start on
> >another one more quickly increasing the factors found per time.
> 
> I thought it continued factoring (at least to a certain level), to
> get more statistical data?

Up to v18 the factoring ran in 16 streams; if it found a factor, it 
did the remaining streams up to the factor found so as to be sure the 
factor found was the smallest. (The size of the _smallest_ factor 
being of interest to at least some mathematicians).

v19 works slightly differently in that it works in single-bit layers 
instead of from bottom to top of the whole range. This saves a 
considerable amount of time when, as commonly happens, a factor is 
found at the low end of a later stream.

Note that, if trial factoring fails but a factor is found using P-1, 
ECM or some other advanced method, we _don't_ know that the factor we 
found is the smallest (though this is probable) - we do however know 
that there are no factors up to the trial factoring limit. Using 
advanced factoring techniques, the only sure way we can find the 
smallest factor is to completely factorize the number. Unfortunately 
this is not always feasible, even for some Mersenne numbers with 
fairly small exponents.


Regards
Brian Beesley
_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Thu, 09 Mar 2000 14:21:41 -0800
From: Stefan Struiker <[EMAIL PROTECTED]>
Subject: Mersenne: AMD, FPU, MOUSE...

TWIMC:
Anyone know where I can find WinBench 99/FPU  WinMark data for the
daisy-fresh 1GHz AMD, perhaps as interpreted by Compaq Productions?

                                                           TNX And Best Regards,
                                                                           Stefanovic

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

Date: Fri, 10 Mar 2000 12:13:15 -0800
From: Stefan Struiker <[EMAIL PROTECTED]>
Subject: Mersenne: There Is Speed, More Speed And Then The 21364

  TWIMC:

         Spec Int and Spec Fp comparison (CPU)
         Posted By johan
         Friday, March 10, 2000 - 1:28:01 PM

         Idiot found a huge spec benchmark fest here ! The German "Mac Info" compares 
the G4, the Athlon 550-1000, the PIII, the PII, the
         Alpha 21264 based on assembled Spec 95 Int and Spec 95 Fp benchmarks from 
everywhere. A few of the most interesting results:

          CPU
                                  Spec Int 95
                                             Spec fp 95
          Alpha 21264/466
                                     24.6
                                               47.9
          Alpha 21264/667
                                     32.1
                                               53.7
          Alpha 21264A/667
                                     37.5
                                               65.5
          Alpha 21264A/700
                                     39.1
                                               68.1
          Alpha 21364/1000
                                     70.0
                                              120.0
          AMD K7/550
                                     25.1
                                               20.6
          AMD K7/800
                                     35.0
                                               25.4
          AMD K7/1000
                                      ?
                                               29.4
          Intel Pentium III/E 733/133
                                     35.6
                                               28.1
          Intel Pentium III/E 800/100
                                     38.3
                                               24.5
          Intel Pentium III/E 800/133
                                     38.4
                                               28.9
          PowerPC 750/500/250/100
                                     23.9
                                               14.6
          PowerPC 7400/450
                                     21.4
                                               20.4
          Sun UltraSPARC-II/450
                                     19.7
                                               27.9
          Sun UltraSPARC-III/600
                                     35.0
                                               60.0

For more 21364 info, see:

http://www.aceshardware.com/

Regards,
 Stefanovic

_________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ      -- http://www.tasam.com/~lrwiman/FAQ-mers

------------------------------

End of Mersenne Digest V1 #704
******************************

Reply via email to