Mersenne Digest Saturday, June 12 1999 Volume 01 : Number 574
----------------------------------------------------------------------
Date: Fri, 11 Jun 1999 23:36:47 +0100
From: Gordon Spence <[EMAIL PROTECTED]>
Subject: Mersenne: Re: Optimisation
George:
If you goto this site there is a code profiler/optimiser available for free
download for 14 day evaluation. It is about 40mb.
http://developer.intel.com
Sorry, I can't remember the exact area or page :-(
regards
G
Gordon Spence, Nokia IP Telephony
Applications Engineer Grove House, Waltham Way,
[EMAIL PROTECTED] White Waltham, Maidenhead,
http://www.nokiaiptel.com/ Berkshire, SL6 3TN, UK.
Office: +44 1628 827204 GSM: +44 385 576623
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 16:31:42 -0700
From: "John R Pierce" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: Optimisation
> If you goto this site there is a code profiler/optimiser available for
free
> download for 14 day evaluation. It is about 40mb.
>
> http://developer.intel.com
>
> Sorry, I can't remember the exact area or page :-(
try http://developer.intel.com/vtune/analyzer/demo2.htm
- -jrp
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 01:39:58 +0200
From: "Otto Bruggeman" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: Optimisation
I believe this is the exact url:
http://developer.intel.com/vtune/icl/demo.htm
Otto
> George:
>
> If you goto this site there is a code profiler/optimiser available for
free
> download for 14 day evaluation. It is about 40mb.
>
> http://developer.intel.com
>
> Sorry, I can't remember the exact area or page :-(
>
> regards
>
> G
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 19:52:38 EDT
From: [EMAIL PROTECTED]
Subject: Mersenne: Re: Mersenne Digest V1 #573
Please - if you can, avoid quoting anything that doesn't directly pertain to
what you're responding to. Some people's posts have dozens of lines of
unnecessary quoting attached. This is an annoyance to (probably the very few)
those who use AOL and are forced to download an attachment when the digest
exceeds a certain size. Anyways:
<<As I said, unless there is an intervention or someone just takes it upon
themselves to double-check those exponents with software other than George's
(the very basis of doublechecking), we won't get confirmation of M37 until
2003>>>
We can doublecheck with Prime95 now, that's a new feature. But different
programs and systems are always a good idea.
I propose (for maybe not the first time) two new systems:
A) A doublechecking cleanup team of computers. A team of (say) five
PIII-500s, 64MB SRAM and suitable motherboards, with cheap everything else
(cases, etc, and probably only one old monitor to share among them all)
should be sent to work on the smallest exponents not doublechecked. I know
that we don't want to "poach" (good word) anyone's work, but exponents at the
VERY bottom of the list really *do* need to be finished. If the "owners" of
those exponents are annoyed, we can credit them for the CPU time they've
invested, or give them another exponent, etc. But the last dregs need to be
taken care of. The team of doublechecking cleanup computers should be pretty
small (I picked five) so that we don't poach too many exponents but still get
the work done.
B) A first-LL-testing cleanup team of computers. This should be even smaller.
One or two dedicated PIII-500s (and with the other characteristics I
mentioned) should be sent to work on the smallest exponents not
singlechecked. Again, we don't want to poach anyone's work, but oftentimes
anomalies crop up (a person who's forgotten about GIMPS and hardly ever turns
his computer on yet it occasionally reports from time to time, say) and
again, we need to finish exponents at the bottom of the list.
Of these two (probably not original) proposals, I consider the doublechecking
one the most important because the slowest computers are usually there. Of
course, these cleanup team(s) should be "offical" and given permission to do
this. Ah, if only I could build a team of LASTLYs. (LASTLY: A PII-400 I built
for the sole purpose of LL testing).
S.T.L.
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 20:34:31 -0400 (EDT)
From: lrwiman <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Poaching (was Mersenne Digest V1 #573)
> A) A doublechecking cleanup team of computers. A team of (say) five
> PIII-500s, 64MB SRAM and suitable motherboards, with cheap everything else
> (cases, etc, and probably only one old monitor to share among them all)
As a computer repairman, of two years, I agree with my father of (5 years
computer repair experience, ~20 years electronics experience), and his
co-worker (30 years of computer repair experience) that your *never ever*
cheap out on power supplies. And crappy cases almost always come with
crappy power supplies. This is expecially true for a team like that one
that would have to have constant operation.
Crappy power supplies can cause flaky performance in innumerable ways
relating to CPU and motherboard operation, as well as drive operation.
Just in case these things get built, what OS should they have, or does
it matter significantly for speed?
- -Lucas Wiman
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 17:58:04 PDT
From: david campeau <[EMAIL PROTECTED]>
Subject: Mersenne: Re: Mersenne Digest V1 #573
>I propose (for maybe not the first time) two new systems:
>A) A double-checking cleanup team of computers. A team of (say) five
>PIII-500s, 64MB SRAM and suitable motherboards, with cheap everything else
>(cases, etc, and probably only one old monitor to share among them all)
>should be sent to work on the smallest exponents not doublechecked. I know
>that we don't want to "poach" (good word) anyone's work, but exponents at
>the
>VERY bottom of the list really *do* need to be finished. If the "owners" of
>those exponents are annoyed, we can credit them for the CPU time they've
>invested, or give them another exponent, etc. But the last dregs need to be
>taken care of. The team of doublechecking cleanup computers should be
>pretty
>small (I picked five) so that we don't poach too many exponents but still
>get
>the work done.
2 (since madpoo stopped) mersenne participant ( that I know of at the
moment) are doing exactly what you are suggesting. Every night one of my
computer ask for 7 days of double-checking work. Every morning I
redistribute small exponnent to my other computers and release any exponent
that are over a certain limit (2,025,000 at the moment). When my computer as
not enough work queued I just raise my limit.
That way I know that exponent are done in a timely manner and they won't be
re-re-assign to someone who is going to "sit on it".
of course it's alot of monitoring:)
David Campeau
P.S: Hope no one (akruppa, madpoo?) got mad because they could never get
those realy small exponent :)
______________________________________________________
Get Your Private, Free Email at http://www.hotmail.com
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 18:03:40 -0700
From: "John R Pierce" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Poaching (was Mersenne Digest V1 #573)
> > A) A doublechecking cleanup team of computers. A team of (say) five
> > PIII-500s, 64MB SRAM and suitable motherboards, with cheap everything
else
> > (cases, etc, and probably only one old monitor to share among them all)
>
> As a computer repairman, of two years, I agree with my father of (5 years
> computer repair experience, ~20 years electronics experience), and his
> co-worker (30 years of computer repair experience) that your *never ever*
> cheap out on power supplies. And crappy cases almost always come with
> crappy power supplies. This is expecially true for a team like that one
> that would have to have constant operation.
>
> Crappy power supplies can cause flaky performance in innumerable ways
> relating to CPU and motherboard operation, as well as drive operation.
sure, but a nice ATX case like the InWin A500 with a reasonable 235W power
supply sells for a cheap $59 or so.... I've got a lab full of servers
running in these back at work, some have been on continuously about 2 years
now.
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 21:43:42 -0400
From: Chris Nash <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Poaching (was Mersenne Digest V1 #573)
> co-worker (30 years of computer repair experience) that your *never ever*
> cheap out on power supplies. And crappy cases almost always come with
> crappy power supplies. This is expecially true for a team like that one
> that would have to have constant operation.
> Crappy power supplies can cause flaky performance in innumerable ways
> relating to CPU and motherboard operation, as well as drive operation.
I have to back Lucas up on this one, and can't stress it enough. In one form
or another over the past few years I've been involved in mathematical
computing 24 hours a day, seven days a week. Apart from one of the first
Gateway P-120 motherboards (which apparently had a known tendency to
overheat and Gateway were aware of it), an early US Robotics 28.8 modem
(which was lousy design) and a modem which was struck by lightning (forgot
to unplug the phone cord), the only hardware component to have failed has
been power supplies. I'm currently on my fourth power supply in 2 years on
my current machine.
When the power supply fails, I have been fortunate and not had any permanent
damage to other hardware components, mainly because voltage regulators tend
to be quite robust components and, even in a failure, don't let much more
than 3.3V or 5.0V hit the board. (Though exploding capacitors in the power
supply *will* blank the CMOS). However a power supply is notoriously full of
very poor, very cheap components. It is the 10c resistor, or 50c smoothing
capacitor that fails - not very comforting when you may have thousands of
dollars of hardware hanging off it. People who drive sports cars don't use
the cheapest gasoline...
It seems lately though that component quality is decreasing. AT power
supplies seemed pretty indestructable, but ATX power supplies are much
weaker. John Pierce is right, a $59 case isn't bad - I could have got the
latest power supply without a case for $48. Resist the temptation though to
go to a high street store and pick up a cheap power supply for $30... and
believe me, a spare power supply you have hanging around, or a reconditioned
one is *not* an option.
It's not worth spoiling the ship for a ha-penny of tar... if you're building
a decent system, don't try to save a few bucks on a power supply. A good
brain is useless if the heart stops. I'm reminded of the Russian guy in
"Armageddon" - "American components, Russian components... all made in
Taiwan!".
Chris Nash
Lexington KY
UNITED STATES
=======================================================
Co-discoverer of probably the 8th and 11th largest known primes.
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 21:56:49 -0400
From: Chris Nash <[EMAIL PROTECTED]>
Subject: Re: Inane Stuff (Was: Mersenne: M38, SETI, and other random stuff )
Once again my apologies for lowering the tone, and many thanks for some
sensible and thought-provoking responses!
> Following conservative estimates of cpu power and number of participants
> doubling every two years, I'd guess that we will have a our first billion
> digit prime in 2021, when we have 40 million participants and Pentium XV
> 1000GHz processors.
10^12Hz... wow! Can you imagine the technical innovation needed to get a
machine where light only travels 0.3mm in a clock cycle? That's some densely
packed, erm, stuff... probably not silicon, the sort of thing we probably
can't conceive right now (electron obedience school?), but there's a good 22
years to go yet. Back in 1977, I seem to remember "VLSI" meant a digital
watch was $100, now they're free with Happy Meals. As for 40 million
participants, maybe they'll be giving away LL engines or Dubner crunchers
with Happy Meals by then, maybe every electronic device in my house will be
squaring and subtracting 2 in its idle time.
> > I'm not counting on seeing either in my lifetime.
> Well, I still plan on seeing it in my lifetime. ;-)
I'm with Spike on this one. If E.T. does make the call, there are going to
be a lot of people dropping EVERYTHING. And even if we don't have "Microsoft
SpaceBender V1.1a SR3" (courtesy of Bob Burrowes, thanks Bob) to make
interstellar communication a possibility, there'll be a lot of people
running for the cryogenic suspension chambers...
Meanwhile, the potential M38 gets ever nearer... not long to wait now I
shouldn't think. I hope Ernst or whoever is verifying will at least give us
a yes or no answer, even if the EFF rules mean that the exponent won't be
released to the world until it is in print...
Chris Nash
Lexington KY
UNITED STATES
=======================================================
Co-discoverer of probably the 8th and 11th largest known primes.
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: 11 Jun 99 20:24:18 MDT
From: Paul Derbyshire <[EMAIL PROTECTED]>
Subject: Re: Mersenne: status of exponents
JON STRAYER <[EMAIL PROTECTED]> wrote:
>> As I said, unless there is an intervention or someone just
>> takes it upon themselves to double-check those exponents
>> with software other than George's (the very basis
>> of doublechecking), we won't get confirmation of
>> M37 until 2003.... <sigh>
> So? It's not like we are running out of work.
No, nor will we ever, barring an ingenious mathematical proof that either
a) There are finitely many Mersenne primes (with upper bound or maximum
number of Mersenne primes) or
b) All (or all but finitely many with an upper bound or maximum quantity
for the sporadics) Mersenne primes follow some pattern.
Imagine there turned out to be a link between primes patterns (or Mersenne
prime patterns) and the Mandelbreot set? It's not out of the question. That
thing has interesting additive combinatorics, also doubling patterns,
Fibonacci sequences, and the like hidden in it.
____________________________________________________________________
Get free e-mail and a permanent address at http://www.netaddress.com/?N=1
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: 11 Jun 99 20:35:58 MDT
From: Paul Derbyshire <[EMAIL PROTECTED]>
Subject: Re: [Re: Inane Stuff (Was: Mersenne: M38, SETI, and other random stuff )]
Chris Nash <[EMAIL PROTECTED]> wrote:
> 10^12Hz... wow! Can you imagine the technical innovation needed to get a
> machine where light only travels 0.3mm in a clock cycle? That's some
> densely packed, erm, stuff... probably not silicon, the sort of thing
> we probably can't conceive right now (electron obedience school?)...
The technical term is "superconductor" and I can conceive it quite fine :-)
(This will probably provoke more cryonics postings.)
Room temp superconductors... or we can pull a star trek and use the warp drive
to speed up the speed of light and thus the optical components inside the CPU.
(This is how the computers on the NCC-1701-D supposedly work...if you don't
believe me, read the technical manual, available at fine bookstores everywhere
and on multimedia CD-ROM.)
>...maybe every electronic device in my house will be squaring and
>subtracting 2 in its idle time.
I'd rather compute a Mersenne LL test. Failing that, I'd devote the idle
cycles to exploring the Mandelbrot set, not some Julia set. And the Julia set
in question here is the world's least interesting...just a line segment.
Square and add i and you get something a tad more interesting...
like from a storm chaser's lucid dreams.
> I'm with Spike on this one. If E.T. does make the call, there are going
> to be a lot of people dropping EVERYTHING. And even if we don't
> have "Microsoft SpaceBender V1.1a SR3"...
Now that would give a whole new meaning to "Internet Exploder". Can you spell
"core breach"?
Please, powers that be. DON'T TRUST *THAT* APP TO MICROSOFT! ANYONE BUT
MICROSOFT!
____________________________________________________________________
Get free e-mail and a permanent address at http://www.netaddress.com/?N=1
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: 11 Jun 99 20:37:49 MDT
From: Paul Derbyshire <[EMAIL PROTECTED]>
Subject: Mersenne: CORRECTION about Microsoft software!
URGENT CORRECTION to the preceding Microsoft-related article.
"Core breach" wouldn't begin to describe it. It would probably leave one of
those subspace rifts that hangs around and swallows hapless ships long after
the original disaster becomes last century's news...
____________________________________________________________________
Get free e-mail and a permanent address at http://www.netaddress.com/?N=1
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 23:12:41 -0400
From: Chris Nash <[EMAIL PROTECTED]>
Subject: Mersenne: Connection with Mandelbrot? (was: status of exponents)
Me again, my wife's out of town so I'm surfing the net too much. Apologies
if you're sick of me.
> b) All (or all but finitely many with an upper bound or maximum quantity
> for the sporadics) Mersenne primes follow some pattern.
> Imagine there turned out to be a link between primes patterns (or Mersenne
> prime patterns) and the Mandelbreot set? It's not out of the question.
That
> thing has interesting additive combinatorics, also doubling patterns,
> Fibonacci sequences, and the like hidden in it.
It's absolutely doubtless there is a link with the Mandelbrot set - and
hopefully I won't fall in to the usual trap of trendy mumbo-jumbo and black
magic that the subject usually yields. It might be doubtful that we could
ever *use* this fact, but who knows, stranger things have happened,
apparently the pretty colors can be generated by charging a metal plate in
the shape of the set itself...
As Paul says, the thing is oozing with combinatorics. Each of its components
has an associated cycle length, in fact, each component has a root point
which, if the iteration is applied n times, the point is mapped back to
itself. Components (or "sprouts") are connected at single "attachment
points", and the cycle length of a child component is a multiple of its
parent. A lot of analysis has been done on the location of these points.
But think of it like this. Suppose you wanted to know how many components
were of each cycle length n. You'd have to find the root points, ie solve
"n'th iterate of x=x", which is an equation of order (you guessed it)
2^(n-1). Factor out the root x=0 (the root of cycle length 1) and the
equation has a Mersenne number degree. Of course, some of these roots are
also roots for factors of n, but you can enumerate them - you'll also end up
running for the Cunningham tables. The strange thing is, since the number of
components of given cycle length grows exponentially, there are not enough
"attachment points" for them all. Hence the familiar "mini-Mandelbrot" sets
have to appear, seemingly in the middle of nowhere (though they are
connected via some very convoluted infinite sequences - the set is, I think,
proven to be a single connected piece).
<Hand-waving>
Prime cycle lengths are interesting, because of the connection points
problem. A prime cycle length *should* produce more "island molecules" than
a composite would. In theory then we could do a primality test; zoom very
closely on a *very* tiny area of the Mandelbrot set (which we could compute
by some iterative root-finding procedure) and have a look. If there's an
"island molecule" there, N is prime, if not, N is composite. Obviously,
mathematical accuracy - and rendering time - is quite an issue if you're
staring so closely at the thing, but it's an intriguing thought experiment.
Sounds radical? Not at all, we've all seen this process 38 times before...
just not in the complex plane:
The Mandelbrot set iteration z -> z^2+c in C
The Lucas test iteration x -> x^2-2 in Z(N).
It's just too much to be a coincidence. A Lucas test is nothing more than a
very thin slice of the Mandelbrot set. Island molecules have a lot in common
with Lucas pseudoprimes.
</Hand-waving>
Chris Nash
Lexington KY
UNITED STATES
=======================================================
Co-discoverer of probably the 8th and 11th largest known primes.
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Fri, 11 Jun 1999 23:37:58 -0400
From: Chris Nash <[EMAIL PROTECTED]>
Subject: Re: Inane Stuff (Was: Mersenne: M38, SETI,and other random stuff )
I'm on a roll tonight. Here's another one from me. Sorry guys, but this one
was just too plain freaky to be a coincidence.
> The technical term is "superconductor" and I can conceive it quite fine
:-)
> (This will probably provoke more cryonics postings.)
I can see it now, George will be asking all you hardware guys how to
optimize version 118 for the 65536-bit Intel superconductor architecture.
The hardware guys better start working on this stuff, software guys like
these sort of gifts :)
> Room temp superconductors... or we can pull a star trek and use the warp
drive
> to speed up the speed of light and thus the optical components inside the
CPU.
> (This is how the computers on the NCC-1701-D supposedly work...if you
don't
> believe me, read the technical manual, available at fine bookstores
everywhere
> and on multimedia CD-ROM.)
And darn fine it is too. And all true ya know. There's only one leap of
faith required for it all to be possible (subspace), ok, maybe the
transporter is something else, but E=mc^2 makes that already seem like fact.
I had to stop myself going into Star Trek mode on my last post. Paul saved
me the embarrassment because I would have made a mess of it.
> >...maybe every electronic device in my house will be squaring and
> >subtracting 2 in its idle time.
> I'd rather compute a Mersenne LL test.
Erm, Paul.... it *is* an LL test. I forgot the mod N, but there's no charge
for that :)
> cycles to exploring the Mandelbrot set, not some Julia set. And the Julia
set
> in question here is the world's least interesting...just a line segment.
... and it's "*the*" LL test as well. Freaky, eh? But surely not the world's
least interesting? Let's go Star Trek with it, this uninteresting,
line-segment Julia set is folded in subspace (ok, modulo N space, it's
getting late and I really need a trek fix right now) into an LL test. I just
posted that... how strange. Great minds think alike, and all that.
> Square and add i and you get something a tad more interesting...
> like from a storm chaser's lucid dreams.
A very different LL test, but still an LL test. A lucid dream, most
definitely, but perhaps a catalog of all the Lucas pseudoprimes
discriminant -1 (so of the form 4n+3)?
>>> have "Microsoft SpaceBender V1.1a SR3"...
>> Now that would give a whole new meaning to "Internet Exploder". Can you
spell
>> "core breach"?
"Gimme an M, Gimme an I, Gimme a C.....".
>URGENT CORRECTION to the preceding Microsoft-related article.
I love the list server when the "preceding article" arrives afterwards. Some
relativistic effect, obviously.
>"Core breach" wouldn't begin to describe it. It would probably leave one of
>those subspace rifts that hangs around and swallows hapless ships long
after
>the original disaster becomes last century's news...
Or maybe worse. One might board the hapless ship and find Bill Gates rather
than Montgomery Scott suspended in the transporter pattern buffer...
Enough already, I'm probably already in too many people's killfiles than is
healthy.
Chris Nash etc etc....
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 07:47:38 +0200
From: Sturle Sunde <[EMAIL PROTECTED]>
Subject: Re: Mersenne: status of exponents
> If I see a test that will take well over a year, is it wrong of me to just
> do it myself with a manual assignment?
Yes. I would get very pissed if someone snatched an exponent which I
already spent a year of work on, and am still working on, without even
telling me in advance. Hey, you coud have the P###-files, but tell me!
As long as the people working on the exponent are actualy working of them,
I think it is very little nice of you to hijack their exponents without
even sending them an email in advance!
At least: Stay away from my exponents! If you touch them, I will find
you with the completed exponents report, track you down, tell the FBI that
they don't have all your computer equipment and that you are searching for
primes again, force you to to switch to SETI@home and factor the exponents
you've tested to pieces, so that you'll loose credit for all the CPU time!
Be afraid. B-)
Also, most of my assignments, both double cheking and first time, are
running on non-Intel Unix boxes. These are reserved directly form George,
not from Primenet, so you won't find them. The Primenet software have
flaws which makes it unuseable for other packages than Prime95/mprime.
Primenet once hijacked 58 of the double cheking assignments I was working
on due to a bug in the Primenet software. Unfortunately the next batch I
got mostly contained exponents which wanted 256K FFT size instead of 128K.
256K FFT size means more than tvice the CPU-time, which means to slow to
test on an Indy. Those 58 machines are therefore retired from GIMPS. 8-(
- --
Sturle URL: http://www.stud.ifi.uio.no/~sturles/ Er det m}ndag i dag?
~~~~~~ MMF: http://www.alladvantage.com/go.asp?refid=BUP399 - St. URLe
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 01:10:38 -0600
From: "Aaron Blosser" <[EMAIL PROTECTED]>
Subject: RE: Mersenne: status of exponents
> At least: Stay away from my exponents! If you touch them, I will find
> you with the completed exponents report, track you down, tell the
> FBI that
> they don't have all your computer equipment and that you are
> searching for
> primes again, force you to to switch to SETI@home and factor the
> exponents
> you've tested to pieces, so that you'll loose credit for all the
> CPU time!
> Be afraid. B-)
Yow! I'll be sure and stay away from your exponents anyway!!! :-) The last
thing I need is for the FBI to come and take the 3 new machines I have at
home (and 2 of them belong to my company)!! :-)
The ones I grabbed (I went back and looked at the assigned work page on
Primenet...so if you manually got some from George, they wouldn't show up
there in theory) were *first time* checks that had completion dates at least
a year from now.
That way, when the original assignee checks in their result, it's at least
good for a double-check, which is what a machine that slow should've been
doing anyway. These are exponents in the 4M-5M range...there were very few
that were going to take over a year...thank goodness.
> Also, most of my assignments, both double cheking and first time, are
> running on non-Intel Unix boxes. These are reserved directly
> form George,
> not from Primenet, so you won't find them. The Primenet software have
> flaws which makes it unuseable for other packages than Prime95/mprime.
I think that's what Jud McCranie was talking about with the un-tested
first-time exponents under 4M. I guess those are some that show up on the
mersenne.org status list. Now...correct me if I'm wrong, but not all
exponents that George keeps track of end up in Scott's Primenet
database...right? George still keeps a few that he assigns to people
manually? Or are there still some manual assignments that got passed out
*before* the advent of Primenet that STILL haven't been finished? Yow!
> Primenet once hijacked 58 of the double cheking assignments I was working
> on due to a bug in the Primenet software.
That's happened to me, strangely enough...every now and then when a machine
of mine update expected completion dates, it will release some exponents
because it says they were already tested (happened more often with factoring
assignments). I don't know why, and I haven't seen that happen lately...I
assume some bug was causing Primenet to assign the same exponent twice
maybe? No sweat off my brow though...I've turned in a few LL results that,
for whatever reason ended up being counted as a double-check instead of the
first time LL test it started out as, but again...each one get's
double-checked anyway, so it's all good.
Aaron
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 01:19:44 -0600
From: "Aaron Blosser" <[EMAIL PROTECTED]>
Subject: RE: Mersenne: Re: Mersenne Digest V1 #573
> 2 (since madpoo stopped) mersenne participant ( that I know of at the
> moment) are doing exactly what you are suggesting. Every night one of my
> computer ask for 7 days of double-checking work. Every morning I
> redistribute small exponnent to my other computers and release
> any exponent
> that are over a certain limit (2,025,000 at the moment). When my
> computer as
> not enough work queued I just raise my limit.
>
> That way I know that exponent are done in a timely manner and
> they won't be
> re-re-assign to someone who is going to "sit on it".
>
> of course it's alot of monitoring:)
>
> David Campeau
>
> P.S: Hope no one (akruppa, madpoo?) got mad because they could never get
> those realy small exponent :)
Actually, I'm team madpoo (I changed names after the problem last year...to
avoid some attention if you know what I mean).
I had some curious behaviour on some machines...I had built about 20 or so
P-166's for a test lab and they got a whole bunch of double-checks, but then
people started nabbing our test machines for real use and I went in and
unreserved them since they were going to rebuild the machines...well, a few
machines didn't get rebuilt and were still working on the original
exponents...whoops! I think those are worked out of their systems now
though (it's been a few months).
I was trying to be nice and unassign them rather than wait for the
completion_time+60_days to expire before they were returned to the pool...so
I guess some numbers were triple-checked? I hate when that happens.
Strangely enough, I had a similar experience with US WEST. The moment I was
walked out the door over a year ago, I knew they would just stop the service
on each machine so I unreserved every exponent I had checked out (there were
a lot ya know :-)
But even months later, those silly US WEST people had apparently missed some
machines since I caught at least 2 that kept on running, checking in/out
exponents. Sheesh. I guess they eventually found them since I knew they
must have been monitoring my prime stats page... But that's another big
reason I changed my account name...to keep those darn machines from
bothering me. I even considered writing an email to someone there to let
them know what was happening, but I could just imagine it: "How do you know
some machines are still running the program? You must be hacking in!" Doh!
Don't need that! :-(
Aaron
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 07:49:22 -0500
From: Gary Diehl <[EMAIL PROTECTED]>
Subject: Mersenne: Top Producers
Ok. So I've only just recently joined GIMPS, and I finished LL testing
one exponent a day or so ago. It was in the M787xxxx range, and it took
27 days. Maybe I'm just being silly but I wonder...How come I'm not on
the (longer) top producer's list (albeit near the bottom of the list).
Doesn't the list update automatically?
I had my system overclocked to P2-333 for 21 of the 27 days, and then I
put the clock speed back due to heat and finished the test at P2-266 for
the remaining 6 days.
123.49 days = 1 p90 year at 266mhz (365/(266/90))
98.64 days = 1 p90 year at 333mhz (365/(333/90))
(these assume that MHZ is the deciding factor of speed, i.e. my 266 is
2.955556 times faster than a p90)
The 333mhz run accounted for 21.28% of one p-90 year. (21/98.64)*100
The 266mhz run accounted for 4.85% of one p-90 year. (6/123.49)*100
That totals 26.13% (approx) of one p-90 year that I have used in testing
one exponent. Did I do the math right?
Did the change of MHZ on my config mess up the primenet server?
Gary Diehl
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 09:28:14 EDT
From: [EMAIL PROTECTED]
Subject: Re: Mersenne: Top Producers
> Ok. So I've only just recently joined GIMPS, and I finished LL testing
> one exponent a day or so ago. It was in the M787xxxx range, and it took
> 27 days. Maybe I'm just being silly but I wonder...How come I'm not on
> the (longer) top producer's list (albeit near the bottom of the list).
> Doesn't the list update automatically?
I imagine you're referring to the Top Producer list at:
http://www.mersenne.org/top.htm
With the longer list at:
http://www.mersenne.org/top2.htm
These are update MANUALLY. Lately it has been about
every other week. At the bottom of the list is says when
it was "last updated" -- June 06 is the latest I've seen.
On the other hand, the list at:
http://entropia.com/ips/tops.shtml
is updated every hour. However, it only counts time through
the IPS. This won't matter since you joined recently, but
"old-timers" will have more time on the former lists.
Randy Given
[EMAIL PROTECTED]
http://members.aol.com/GivenRandy
public key at http://members.aol.com/GivenRandy/pgpkey.asc
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 09:37:32 -0400
From: Jud McCranie <[EMAIL PROTECTED]>
Subject: Re: Mersenne: status of exponents
At 07:47 AM 6/12/99 +0200, Sturle Sunde wrote:
>Yes. I would get very pissed if someone snatched an exponent which I
>already spent a year of work on, and am still working on, without even
>telling me in advance.
We are all in some danger of this happening because of non-GIMPS people working
on Mersenne primes (with Crays, etc).
+----------------------------------------------+
| Jud "program first and think later" McCranie |
+----------------------------------------------+
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 09:48:37 -0400
From: Jud McCranie <[EMAIL PROTECTED]>
Subject: Mersenne: status pages
There seems to be a big discrepancy between what the status page at
merseme.org shows (updated 6-6-99) and what the PrimeNet status page
(updated hourly) shows as far as the exponents under 4,000,000. So maybe
these small exponents that the former page shows that I was concerned about
have actually been done.
+----------------------------------------------+
| Jud "program first and think later" McCranie |
+----------------------------------------------+
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
Date: Sat, 12 Jun 1999 15:56:39 +0100
From: "Brian J. Beesley" <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Poaching
> A) A doublechecking cleanup team of computers. A team of (say) five
> PIII-500s, 64MB SRAM and suitable motherboards, with cheap everything else
> (cases, etc, and probably only one old monitor to share among them all)
I'd rather drop to PII-400. The memory bus speed is the same, so the
PIII-500 will not be as much as 25% faster. You might see 15% if
you're lucky. A "retail box" PII-400 is less than half the price of a
PIII-500, so you will probably find that you could build (at least) 5
complete PII-400 systems for the price of 4 PIII-500s, and end up
with more total power for less expense. Plus, you'd have more upgrade
potential, when faster processors come down to a sane price.
You might get even more "bangs per buck" by using Celeron 400
processors, though these will be less quick than PII-400 (Celeron
memory bus is only 66 MHz) and will have less upgrade potential
(seems impossible to get Slot 1 Celerons these days, and the Socket
370 to Slot 1 converters may have a reliability penalty, from what
some people say) Again, the price premium on the faster Celerons is
not justified by the small increase in performance.
Since you're only going for 64 MB RAM (more than sufficient) I'd
reccomend spending a little extra on ECC SDRAM (obviously only
applicable to PII/PIII systems, no support on Celeron). This should
make the system a deal less likely to produce wrong results as a
consequence of random hardware glitches.
And fit a UPS, or at least a decent mains surge suppressor.
> If the "owners" of
> those exponents are annoyed, we can credit them for the CPU time they've
> invested, or give them another exponent, etc.
If you really are going ahead with this, I think you should agree
formally to do something about sharing the credit/any prize money
etc., should it be found that one of the exponents you have "poached"
from its "rightful owner" actually generates a Mersenne prime.
Finally, I think you should make determined efforts to contact any
user who has an assignment allocated to them through official
channels before you "poach" it. Out of sheer courtesy, if nothing
else.
Regards
Brian Beesley
________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
------------------------------
End of Mersenne Digest V1 #574
******************************
