Hi everyone,

I've just brought a new server on-line which has limited awareness of the
architectures of the various machines being used. The reservation/status
page has been tweaked to include this information.
If people want to volunteer more information about their machines, then just
shout, and I can edit the config file and get the server to re-read that config
without pulling the machine down. Which is nice.

As a minor change I've made the query for the stats
   http://62.236.152.56:8080/res
instead of ?res, but detect ?res queries, and send back a 301 pointing
you to the new address. All known browsers will automatically follow a
301. The publicly visable table is still at
   http://fatphil.org/maths/PIES/clientserver.html

Phil

=====
When inserting a CD, hold down shift to stop the AutoRun feature
In the Device Manager, disable the SbcpHid device.
http://www.cs.princeton.edu/~jhalderm/cd3/

__________________________________
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
http://photos.yahoo.com/

--- Bo Xiao <ebx@...> wrote:
> > > I am interested in testing bigger numbers. Phil
> > > mentioned current fft lost accuracy at around
> > > 100,000 digits. What does it take to fix that?
> > > Would a 64 CPU help?
> >
> > I tried to do the same thing with ForEis that Yves did with GeneFer, making
> it use 80-bit floats. However, I only got a tiny increase in range (about 10%
> max) for my 3x slowdown, and decided that wasn't worth it.
>
> [excause my ignorcence, a couple of weeks from now,
> I will talk this fft thing differently. :)]
>
> Genefer can test higher exponents. Just the ranges drop.

GeneFer, with a 3x slowdown, can test 50 times the range of Proth
at the same exponent.

I can't explain why I only get <1.1x the range.

> > For much bigger numbers, exponent 98304, I need a completely new FFT.
>
> Have you tested other available fft even if they are
> slower? Just to see if we can do it.

All the general FFTs are too slow, and all the optimised FFTs have
coupled the DWT tightly to the FFT and assume that they're doing
IBDWT or right-angle DWT. I use my own new DWT. I'd need to remove
their DWT from the FFT first, and then add my own DWT. It's not
easy, as the ones I've looked at have been very tightly coupled
indeed.

Phil

=====
When inserting a CD, hold down shift to stop the AutoRun feature
In the Device Manager, disable the SbcpHid device.
http://www.cs.princeton.edu/~jhalderm/cd3/

__________________________________
Do you Yahoo!?
New Yahoo! Photos - easier uploading and sharing.
http://photos.yahoo.com/

On Sat, Dec 06, 2003 at 03:49:03PM -0000, thefatphil wrote:
> --- In pies_project@yahoogroups.com, Bo Xiao <ebx@e...> wrote:
> > Unless we want to shuffle the order, we wont get
> > to the 100000 digit area in another 6 months.
>
> 10^(100000/16384) = 1269157.8

Ok, I calculated the number at 1270000 but in my mind,
I read it as 1700000. :(

[...]

> > I am interested in testing bigger numbers. Phil
> > mentioned current fft lost accuracy at around
> > 100,000 digits. What does it take to fix that?
> > Would a 64 CPU help?
>
> I tried to do the same thing with ForEis that Yves did with GeneFer, making it
use 80-bit floats. However, I only got a tiny increase in range (about 10% max)
for my 3x slowdown, and decided that wasn't worth it.

[excause my ignorcence, a couple of weeks from now,
I will talk this fft thing differently. :)]

Genefer can test higher exponents. Just the ranges drop.

>
> For much bigger numbers, exponent 98304, I need a completely new FFT.

Have you tested other available fft even if they are
slower? Just to see if we can do it.

Bo

>
> Phil

--- In pies_project@yahoogroups.com, Bo Xiao <ebx@e...> wrote:
> Unless we want to shuffle the order, we wont get
> to the 100000 digit area in another 6 months.

10^(100000/16384) = 1269157.8

We've handed out numbers up to 1229522 already, and we've only been going a
week. I think we'll have a 100000 digit prime before Christmas.

> Or if we can attract more folks.

Given that the project is very finite, I'm tempted to drag it out as long as
possible.

> I am interested in testing bigger numbers. Phil
> mentioned current fft lost accuracy at around
> 100,000 digits. What does it take to fix that?
> Would a 64 CPU help?

I tried to do the same thing with ForEis that Yves did with GeneFer, making it
use 80-bit floats. However, I only got a tiny increase in range (about 10% max)
for my 3x slowdown, and decided that wasn't worth it.

For much bigger numbers, exponent 98304, I need a completely new FFT.

Phil

--- In pies_project@yahoogroups.com, "David Broadhurst" <d.broadhurst@o...>
wrote:
> PS: The Wagstaff cumulant line is shown in
> http://opteron.mersenneforum.org/png/log2_log2_Mn.png
> There was an unlucky glitch between M31 and M32, but
> thereafter the best fit gradient looks Wagstaffian.

One thing that annoys me about that chart (and the one on the prime pages) is
the fact that it's actually meaningless.

The only more sensible way of plotting "how on target are we" would be some kind
of statistical confidence level that the green line hypothesis were true.

The green line is otherwise utterly meaningless as it stands.
Once you've found primes up to M127, it means nothing to talk about the
"expected number of primes to M127" when projecting a line from 127 onwards, as
the number of primes is known and fixed. And so for every other point. The green
line implies that the primes affect each other's probabilities, which has no
heuristic backing at all.

Mersenne primes, like a roulette wheel, have no memory.

Phil

Note - I'm a few days behind reading these as yahoo has decided that it doesn't
want to deliver mails from this forum to my preferred address :-( (I get
primeform ones to that address though?!?)

--- In pies_project@yahoogroups.com, "David Broadhurst" <d.broadhurst@o...>
wrote:
> > For reference (and David can probably do
> > the sums), I think that over the last 6
> > GIMPS primes, George has been quite lucky
> > with distribution
>
> The last 6 points on
>
> http://primes.utm.edu/gifs/lg_lg_Mn.gif
>
> are in fair accord with the Wagstaff slope:
>
> 2^exp(-Euler)

I've worked out why I thought George had been lucky -
his CPU power growth is greater than Moore's law.
Assuming expenential growth of CPU power along the lines
of Moore's law, I think he should have expected a prime
in 4 years, not 2. (I did the calculation after the
previous prime, I was sure that it would be 4 years to
the next one, and was thus surprised when this one came
along.)

Phil

corection
we have to multiply all the primes from 2 till "end of the range"
do we have them all?

On Friday 05 December 2003 22:42, Bo Xiao wrote:
> Relaying the message ...
>
> ----- Forwarded message from Juraj Holtak <e0125383@...>
> -----
>
> > From: Juraj Holtak <e0125383@...>
> > Reply-To: e0125383@...
> > To: Bo Xiao <ebx@...>
> > Subject: Re: [pies_project] (unknown)
> > Date: Fri, 5 Dec 2003 22:34:23 +0100
> > User-Agent: KMail/1.5
> >
> > Yes :-)
> >
> > page with statistics about the hardware, compiler, compiler options and
> > dedicated/not dedicated pies-maschine and the amount of number done per
> > day or something like that
> >
> > Phil when we find all the primes in the range and multiply them and add 1
> > do we get another not yeat known *huge* prime?
> > or r there stil primes missing?
> >
> >
> > hmm my english is getting worse and worse....
> > sorry guys
> >
> > Juraj
> >
> > On Friday 05 December 2003 20:03, Bo Xiao wrote:
> > > Somewhere we should see the longer term plan of the
> > > project. The available ranges should be done in less
> > > than a year even with no new computers added.
> > >
> > > A benchmark page would be nice too.
> > >
> > > Bo
> > >
> > > On Thu, Dec 04, 2003 at 10:48:08PM +0200, Phil Carmody wrote:
> > > > I've just noticed from
> > > >   http://primes.utm.edu/primes/status.php?hours=18
> > > > that the windows (boo hiss!) guys are churning out primes at a rate
> > > > of knots, with Steven now joining the ranks of the 49152 finders with
> > > > his f2 b=110234. So there's 4 of us now! (Don't worry Chris, Raffi,
> > > > and Bo, there are plenty in there, and with 1 prime expected every
> > > > ~3500 tests, it won't be too long, I'm sure.)
> > > >
> > > > Presently there's no way of getting the "file" testers' results
> > > > integrated with the client-server data, except when they both finally
> > > > reach my hand-edited tables at fatphil.org.  Is anyone perturbed by
> > > > this apparent schizophrenia?
> > > >
> > > > Does anyone have any suggestions regarding teh client-server setup?
> > > >
> > > > Do people want personal info, or icons or machine information, or
> > > > stuff like that on the client-server status page? Or on the
> > > > hand-edited pages?
> > > >
> > > > (Funnily enough, that bizarre looking proth on the prime pages status
> > > > page is also one of my babies, one of the proth-racers I was working
> > > > with Robert on! So I've t0ta11y 0wnz0r3d the submissions table today
> > > > :-) )
> > > >
> > > > Phil
> > > >
> > > > --
> > > > People generally seem to want software to be free as in speech and/or
> > > > free as in beer. Unfortunately rather too much of it is free as in
> > > > jazz. -- Janet McKnight
> > > >
> > > >
> > > >
> > > > To unsubscribe from this group, send an email to:
> > > > pies_project-unsubscribe@yahoogroups.com
> > > >
> > > >
> > > >
> > > > Your use of Yahoo! Groups is subject to
> > > > http://docs.yahoo.com/info/terms/
> > >
> > > To unsubscribe from this group, send an email to:
> > > pies_project-unsubscribe@yahoogroups.com
> > >
> > >
> > >
> > > Your use of Yahoo! Groups is subject to
> > > http://docs.yahoo.com/info/terms/
>
> ----- End forwarded message -----
>
>
> To unsubscribe from this group, send an email to:
> pies_project-unsubscribe@yahoogroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/

Relaying the message ...

----- Forwarded message from Juraj Holtak <e0125383@...> -----

> From: Juraj Holtak <e0125383@...>
> Reply-To: e0125383@...
> To: Bo Xiao <ebx@...>
> Subject: Re: [pies_project] (unknown)
> Date: Fri, 5 Dec 2003 22:34:23 +0100
> User-Agent: KMail/1.5
>
> Yes :-)
>
> page with statistics about the hardware, compiler, compiler options and
> dedicated/not dedicated pies-maschine and the amount of number done per day
> or something like that
>
> Phil when we find all the primes in the range and multiply them and add 1 do
> we get another not yeat known *huge* prime?
> or r there stil primes missing?
>
>
> hmm my english is getting worse and worse....
> sorry guys
>
> Juraj
>
> On Friday 05 December 2003 20:03, Bo Xiao wrote:
> > Somewhere we should see the longer term plan of the
> > project. The available ranges should be done in less
> > than a year even with no new computers added.
> >
> > A benchmark page would be nice too.
> >
> > Bo
> >
> > On Thu, Dec 04, 2003 at 10:48:08PM +0200, Phil Carmody wrote:
> > > I've just noticed from
> > >   http://primes.utm.edu/primes/status.php?hours=18
> > > that the windows (boo hiss!) guys are churning out primes at a rate of
> > > knots, with Steven now joining the ranks of the 49152 finders with his f2
> > > b=110234. So there's 4 of us now! (Don't worry Chris, Raffi, and Bo,
> > > there are plenty in there, and with 1 prime expected every ~3500 tests,
> > > it won't be too long, I'm sure.)
> > >
> > > Presently there's no way of getting the "file" testers' results
> > > integrated with the client-server data, except when they both finally
> > > reach my hand-edited tables at fatphil.org.  Is anyone perturbed by this
> > > apparent schizophrenia?
> > >
> > > Does anyone have any suggestions regarding teh client-server setup?
> > >
> > > Do people want personal info, or icons or machine information, or stuff
> > > like that on the client-server status page? Or on the hand-edited pages?
> > >
> > > (Funnily enough, that bizarre looking proth on the prime pages status
> > > page is also one of my babies, one of the proth-racers I was working with
> > > Robert on! So I've t0ta11y 0wnz0r3d the submissions table today :-) )
> > >
> > > Phil
> > >
> > > --
> > > People generally seem to want software to be free as in speech and/or
> > > free as in beer. Unfortunately rather too much of it is free as in jazz.
> > > -- Janet McKnight
> > >
> > >
> > >
> > > To unsubscribe from this group, send an email to:
> > > pies_project-unsubscribe@yahoogroups.com
> > >
> > >
> > >
> > > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
> >
> >
> > To unsubscribe from this group, send an email to:
> > pies_project-unsubscribe@yahoogroups.com
> >
> >
> >
> > Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>

----- End forwarded message -----

Somewhere we should see the longer term plan of the
project. The available ranges should be done in less
than a year even with no new computers added.

A benchmark page would be nice too.

Bo

On Thu, Dec 04, 2003 at 10:48:08PM +0200, Phil Carmody wrote:
> I've just noticed from
>   http://primes.utm.edu/primes/status.php?hours=18
> that the windows (boo hiss!) guys are churning out primes at a rate of
> knots, with Steven now joining the ranks of the 49152 finders with his f2
> b=110234. So there's 4 of us now! (Don't worry Chris, Raffi, and Bo, there
> are plenty in there, and with 1 prime expected every ~3500 tests, it won't
> be too long, I'm sure.)
>
> Presently there's no way of getting the "file" testers' results integrated
> with the client-server data, except when they both finally reach my
> hand-edited tables at fatphil.org.  Is anyone perturbed by this apparent
> schizophrenia?
>
> Does anyone have any suggestions regarding teh client-server setup?
>
> Do people want personal info, or icons or machine information, or stuff
> like that on the client-server status page? Or on the hand-edited pages?
>
> (Funnily enough, that bizarre looking proth on the prime pages status page
> is also one of my babies, one of the proth-racers I was working with
> Robert on! So I've t0ta11y 0wnz0r3d the submissions table today :-) )
>
> Phil
>
> --
> People generally seem to want software to be free as in speech and/or free
> as in beer. Unfortunately rather too much of it is free as in jazz.
> -- Janet McKnight
>
>
>
> To unsubscribe from this group, send an email to:
> pies_project-unsubscribe@yahoogroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>

> On Thu, 4 Dec 2003, Phil Carmody wrote:
> > I've just noticed from
> >   http://primes.utm.edu/primes/status.php?hours=18
> > that the windows (boo hiss!) guys are churning out
> primes at a rate of
> > knots, with Steven now joining the ranks of the
> 49152 finders with his f2
> > b=110234. So there's 4 of us now! (Don't worry
> Chris, Raffi, and Bo, there
> > are plenty in there, and with 1 prime expected
> every ~3500 tests, it won't
> > be too long, I'm sure.)

For the record, I'm running on linux, just don't have
a networked internet connection.

Steven Harvey

=====
harvey563@...

__________________________________
Do you Yahoo!?
Protect your identity with Yahoo! Mail AddressGuard
http://antispam.yahoo.com/whatsnewfree

On Thu, 4 Dec 2003, Phil Carmody wrote:
> I've just noticed from
>   http://primes.utm.edu/primes/status.php?hours=18
> that the windows (boo hiss!) guys are churning out primes at a rate of
> knots, with Steven now joining the ranks of the 49152 finders with his f2
> b=110234. So there's 4 of us now! (Don't worry Chris, Raffi, and Bo, there
> are plenty in there, and with 1 prime expected every ~3500 tests, it won't
> be too long, I'm sure.)

Perhaps, but I have not sually been lucky.  I am happy
plodding along.

> Presently there's no way of getting the "file" testers' results integrated
> with the client-server data, except when they both finally reach my
> hand-edited tables at fatphil.org.  Is anyone perturbed by this apparent
> schizophrenia?

"Not I" says Chris, returning to the video game that has wasted many hours
of his and his kids time for no good reason.  <grin>

I also of course have no opinions on what data shows.  Wait, maybe I do.
If you want the project to grow, I'd guess that what you need to to offer
what folks want, and that is...?  I actually don't know.  Folks seem to
like rankings, especially if it is clear how they can move up.

If you find out what folks want, let me know, I have been curious for
sometime what the average person submitting primes at my site wants as
well.  Whatever the ECPP folks want--I don't provide it, they often do not
even submit.

Chris

I've just noticed from
   http://primes.utm.edu/primes/status.php?hours=18
that the windows (boo hiss!) guys are churning out primes at a rate of
knots, with Steven now joining the ranks of the 49152 finders with his f2
b=110234. So there's 4 of us now! (Don't worry Chris, Raffi, and Bo, there
are plenty in there, and with 1 prime expected every ~3500 tests, it won't
be too long, I'm sure.)

Presently there's no way of getting the "file" testers' results integrated
with the client-server data, except when they both finally reach my
hand-edited tables at fatphil.org.  Is anyone perturbed by this apparent
schizophrenia?

Does anyone have any suggestions regarding teh client-server setup?

Do people want personal info, or icons or machine information, or stuff
like that on the client-server status page? Or on the hand-edited pages?

(Funnily enough, that bizarre looking proth on the prime pages status page
is also one of my babies, one of the proth-racers I was working with
Robert on! So I've t0ta11y 0wnz0r3d the submissions table today :-) )

Phil

--
People generally seem to want software to be free as in speech and/or free
as in beer. Unfortunately rather too much of it is free as in jazz.
-- Janet McKnight

Unless we want to shuffle the order, we wont get
to the 100000 digit area in another 6 months.

Or if we can attract more folks.

I am interested in testing bigger numbers. Phil
mentioned current fft lost accuracy at around
100,000 digits. What does it take to fix that?
Would a 64 CPU help?

Time to pick up the math books again.

Bo

My congratulations to Phil too!

It's time to get our first 100000-digits prime for n=49152 and
b>1270000.  Who will be that lucky guy?

-- Thomas.

PS: The Wagstaff cumulant line is shown in
http://opteron.mersenneforum.org/png/log2_log2_Mn.png
There was an unlucky glitch between M31 and M32, but
thereafter the best fit gradient looks Wagstaffian.

Cool our chef has found the second prime in
the c/s arena. I wouldnt call it huge but
it certainly signals the right direction.
Go go!

Bo
PS: just dont want it be shaded by m40. ;)

> For reference (and David can probably do
> the sums), I think that over the last 6
> GIMPS primes, George has been quite lucky
> with distribution

The last 6 points on

http://primes.utm.edu/gifs/lg_lg_Mn.gif

are in fair accord with the Wagstaff slope:

2^exp(-Euler)

Like Phil, I wrote a happy message to George.
If others feel happy, why do not so also?

David

George Woltman, via the massive array of machines called GIMPS has just found
(and verified, he's always very strict about not announcing until an independent
verification) the world's new largest prime:

2^20996011-1 6320430 digits.

Those who know George and GIMPS may wish to drop him a congratulatory mail, as
I'm about to do.

For reference (and David can probably do the sums), I think that over the last 6
GIMPS primes, George has been quite lucky with distribution, and therefore this
might be the last one for quite a while (if things are planing on levelling
themselves out). So make the most of this announcement while it's fresh!

http://primes.utm.edu/primes/status.php?hours=12
http://primes.utm.edu/primes/page.php?id=67522
http://primes.utm.edu/bios/code.php?code=G6


Phil

This 6,320,430 digit prime was found on Nov 17th.

I like the article Chris wrote.

Among others, he mentioned both gfn and phi. I
hope we can pull more participants to the project.
We need that. Mersenne numbers goes up too steeply
and there are too few of them out there(within
reach).

Bo

Phil:

> discovered one that was actually prime, so they
> obviously had to be kept in

Indeed. Here is a small titanic example

Calling Brillhart-Lehmer-Selfridge with factored part 33.94%
Phi(2^7*3^2,2^3*3^4) is prime! (1.7337s+0.0434s)

David (not actually a pies-artist)

It appears Kimmo has just encountered something that Raffi encountered a while
back - the dreaded pseudoprime!

Phi(2^s*3^t, 2^a*3^b) is very often a 2-SPRP and 3-SPRP (always?)

However, the pseudoprimes will _fail_ a Pocklington test,
which is what the '-t' switch performs.

I nearly (before I looked at them more closely) removed such
numbers from the list of candidates. However, just as I ws about to do that, I
discovered one that was actually prime, so they obviously had to be kept in.
Fortunately these numbers are mostly sieved out, and I think there will only be
one other such number in the ranges being handed out.

I don't think anyone will ever encounter it, but there's also the occasional
number that is described as "unknown" - and basically that's what it is. The FFT
overflowed catastrophically, and no conclusion could be made. However, just
before the fatal FFT, I did a quick check of the number, to see if it had a form
which PRPs often had, a "fingerprint", so to speak. These numbers are more often
than not primes, and so are worth throwing at PFGW. I'll be amazed if anyone
actually meets one of these things though.


Phil

--- In pies_project@yahoogroups.com, "raffi_raccoon" <raffi@s...> wrote:
> Congratulations Pie Chef and Thomas for the largest PIES prime yet
> at 99617 digits!
>
> http://primes.utm.edu/primes/page.php?id=67494
>
> It's a great looking prime and it sure is huge!
>
> CHEERS!

Seconded Raffi! Congrats to Thomas for finding the bug in my server that was
failing to include peaches such as these! Now fixed.

   http://62.236.152.56:8080/?res  (realtime)
   http://fatphil.org/maths/PIES/clientserver.html  (updated once per hour).

We shouldn't forget that Kimmo has already found something like _7_ primes in
the ~70-80 kdigit range, which is a lot more work (and a lot more credit for the
PIES project on the top-20 table.)

Phil

Congratulations Pie Chef and Thomas for the largest PIES prime yet
at 99617 digits!

http://primes.utm.edu/primes/page.php?id=67494

It's a great looking prime and it sure is huge!

CHEERS!

--- In pies_project@yahoogroups.com, Bo Xiao <ebx@e...> wrote:
> I better find something in the next 24 hours or my new
> prover code will be taken away. BTW, how do I credit
> sieve and PIES for any findings? I only have foreis
> now(f10).

I would suggest letting it be taken away.  Then, when you get a
prime, request a new code like before

   (1) Start at your bio page and click on "create
       new proof code" at the bottom of the page.
   (2) Click on "Phil Carmody's ForEis"
   (3) In the spot for other info, where it now says "none",
       type instead:  "PhiSieve, PIES"

The confirmation page should list all off the appropriate
entries.

Chris.

Ok I am biting. Shift to 49152 on next chunk request.
(changed exp and chunk size in my cfg) Also updated
my client.

I better find something in the next 24 hours or my new
prover code will be taken away. BTW, how do I credit
sieve and PIES for any findings? I only have foreis
now(f10).

Bo

On Thu, Nov 27, 2003 at 12:25:05AM -0000, thefatphil wrote:
> Teething trouble galore...
>
> Well the last 24 hours has been _hectic_.
> Several bugs have raised their ugly head, several 'issues'
> have been raised too.
>
> 1) Bo's "locking up" last night - inexlicable.
> 2) Bo's inability to restart - changes to both client and server.
> 3) Raffi's name-changing issue - don't change machine names!
> 4) Raffi's name choice issue - changes to the server.
> 5) Raffi's PRP causing knickers to get twisted - client changes.
> 6) Richard's desire for progress info - server changes.
> 7) Phil wanting to get 49152 online as well as 24576 - server changes.
> 8) No way to stop requesting ranges - server changes.
>
>
> Basically, about 2 minutes ago, I brought a completely re-vamped
> server program online. I'll be staring at the screen for a couple
> of hours to check that at least a few people exchange blocks
> happily.
>
> The server now manages _both_ the 24576 and 49152 ranges.
>
> I will be advising that those with fast machines move over to
> 49152, but I don't want everyone to go over at the same time, I'd rather
migrate people one at a time. "Fast" here means anything 1GHz or above. The
migration's very easy, do I have any volunteers?
> (Raffi, Bo, and Chris all have at least one fast machine, IIRC form my server
logs.)
>
> Those who are already "client/server enabled" can take a peek at the
client/server progress by pointing their web-browser at the range server
>   http://62.236.152.56:8080/?res
> Note - it's not a webserver, and it won't serve you favicon.ico, no matter how
often you ask. If browsers keep asking for that file, I shall be forced to serve
malformed .ico files in order to crash them :-P .
>
>
> I've also put a new perl client on the clients page on the web-site. I
recommend everyone updates as soon as possible, as
> some of the bugs were pretty critical. (e.g. if you find a PRP, the current
client will fall over horribly on the next chunk, which is a bit stupid.)
>
> The new client permits you to indicate that you don't want any more chunks
simply by setting chunk=0 in the config file. When you restart the client, it
will continue the currently reserved chunk, andcotinue processing. On completion
it will hand the results to the server, the server will say thank you, and the
two can then part company, sob.
>
>
> Alpha users - watch this space - there will be a new Alpha testing program
which Juraj kindly compiled using CCC which is >5% faster than the current
client.
>
> Phil
>
>
>
> To unsubscribe from this group, send an email to:
> pies_project-unsubscribe@yahoogroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>

Teething trouble galore...

Well the last 24 hours has been _hectic_.
Several bugs have raised their ugly head, several 'issues'
have been raised too.

1) Bo's "locking up" last night - inexlicable.
2) Bo's inability to restart - changes to both client and server.
3) Raffi's name-changing issue - don't change machine names!
4) Raffi's name choice issue - changes to the server.
5) Raffi's PRP causing knickers to get twisted - client changes.
6) Richard's desire for progress info - server changes.
7) Phil wanting to get 49152 online as well as 24576 - server changes.
8) No way to stop requesting ranges - server changes.


Basically, about 2 minutes ago, I brought a completely re-vamped
server program online. I'll be staring at the screen for a couple
of hours to check that at least a few people exchange blocks
happily.

The server now manages _both_ the 24576 and 49152 ranges.

I will be advising that those with fast machines move over to
49152, but I don't want everyone to go over at the same time, I'd rather migrate
people one at a time. "Fast" here means anything 1GHz or above. The migration's
very easy, do I have any volunteers?
(Raffi, Bo, and Chris all have at least one fast machine, IIRC form my server
logs.)

Those who are already "client/server enabled" can take a peek at the
client/server progress by pointing their web-browser at the range server
   http://62.236.152.56:8080/?res
Note - it's not a webserver, and it won't serve you favicon.ico, no matter how
often you ask. If browsers keep asking for that file, I shall be forced to serve
malformed .ico files in order to crash them :-P .


I've also put a new perl client on the clients page on the web-site. I recommend
everyone updates as soon as possible, as
some of the bugs were pretty critical. (e.g. if you find a PRP, the current
client will fall over horribly on the next chunk, which is a bit stupid.)

The new client permits you to indicate that you don't want any more chunks
simply by setting chunk=0 in the config file. When you restart the client, it
will continue the currently reserved chunk, andcotinue processing. On completion
it will hand the results to the server, the server will say thank you, and the
two can then part company, sob.


Alpha users - watch this space - there will be a new Alpha testing program which
Juraj kindly compiled using CCC which is >5% faster than the current client.

Phil

--- In pies_project@yahoogroups.com, Bo Xiao <ebx@e...> wrote:
> Hi,
>
> Just to stir the water.
>
> What is the density after sieving?

24576  -  .33039
49152  -  .31889

> For the chunks
> I am getting, it is right at 50%. For the current
> 24576, there is about 1 prime in every 5000 ragne
> (obvervation from phil's status page). In other
> words, one every 2500 candidates.

I thought it was more like 7000, which  puts the post-sieving density at 1/2300.
That figure could be out by quite a margin
as I've not calculated the density.

> The size of the numbers is comparable to gfn 8192.

The original 8192s, yes, not the extended ones.

> There is about 1 prime in every 2000 candidates
> over there, where the candidatedensity is at about
> 12.5%.

Yeah - he can't test the even candidates, I can :-)
Instant 2x density boost.

> Now for the same XP1900+, I can test 1 phi 24576
> in 1.5 minutes and 1 gfn 8192 in 4.5 minutes. So
> the time it takes to hit a phi prime is 4 times
> better, or 9 to 2.25. Is that about right?

The larger 8192 is "worth" about 1.6 times as much on the Prime Pages top-5000.
So the relative worth/cost ratio is closer to
1.8:1

They are quite good, aren't they :-)

> (All timings are obervations/estimates. Not hard
> banchmarks.)
>
> The side question is, how is our sieveing doing?

I've stopped 49152 sieving now, at the limit of my DEC alpha's FPU accuracy.
49152 is now being tested, in fact.

More news on this later...

Phil

On Wed, Nov 26, 2003 at 12:53:17PM -0800, Bo Xiao wrote:
> Hi,
>
> Just to stir the water.
>
> What is the density after sieving? For the chunks
> I am getting, it is right at 50%. For the current
> 24576, there is about 1 prime in every 5000 ragne
> (obvervation from phil's status page). In other
> words, one every 2500 candidates.
>
> The size of the numbers is comparable to gfn 8192.
> There is about 1 prime in every 2000 candidates
> over there, where the candidatedensity is at about
> 12.5%.
>
> Now for the same XP1900+, I can test 1 phi 24576
> in 1.5 minutes and 1 gfn 8192 in 4.5 minutes. So
> the time it takes to hit a phi prime is 4 times
> better, or 9 to 2.25. Is that about right?

OK it is 9 to 3.375.

>
> (All timings are obervations/estimates. Not hard
> banchmarks.)
>
> The side question is, how is our sieveing doing?
>
> Happy Hunting!
>
> Bo
>
>
> To unsubscribe from this group, send an email to:
> pies_project-unsubscribe@yahoogroups.com
>
>
>
> Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
>
>

Hi,

Just to stir the water.

What is the density after sieving? For the chunks
I am getting, it is right at 50%. For the current
24576, there is about 1 prime in every 5000 ragne
(obvervation from phil's status page). In other
words, one every 2500 candidates.

The size of the numbers is comparable to gfn 8192.
There is about 1 prime in every 2000 candidates
over there, where the candidatedensity is at about
12.5%.

Now for the same XP1900+, I can test 1 phi 24576
in 1.5 minutes and 1 gfn 8192 in 4.5 minutes. So
the time it takes to hit a phi prime is 4 times
better, or 9 to 2.25. Is that about right?

(All timings are obervations/estimates. Not hard
banchmarks.)

The side question is, how is our sieveing doing?

Happy Hunting!

Bo