Mersenne Digest Sunday, February 11 2001 Volume 01 : Number 815
----------------------------------------------------------------------
Date: Thu, 08 Feb 2001 19:35:26 -0800
From: "Terry S. Arnold" <[EMAIL PROTECTED]>
Subject: Mersenne: P 4 Optimiztion
I have heard rumblings that George has put together a version of the LL
test optimized for the P 4. Does anyone know what the performance is like
compared to a P 3?
Terry
Terry S. Arnold 2975 B Street San Diego, CA 92102 USA
[EMAIL PROTECTED] (619) 235-8181 (voice) (619) 235-0016 (fax)
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Thu, 08 Feb 2001 23:09:51 -0500
From: George Woltman <[EMAIL PROTECTED]>
Subject: Re: Mersenne: P 4 Optimiztion
Hi Terry,
At 07:35 PM 2/8/2001 -0800, Terry S. Arnold wrote:
>I have heard rumblings that George has put together a version of the LL
>test optimized for the P4. Does anyone know what the performance is like
>compared to a P3?
George is working on it, but is a long way from completion. Progress is
slow, primarily due to my own laziness. My estimate for a 512K FFT is 0.4
seconds
on a 1.4GHz P4. You can compare that to other machines at
http://www.mersenne.org/bench.htm.
As an aside, I've started putting together a document on what I'm learning
about the P4 at http://www.mersenne.org/p4notes.doc
I'll try to be more diligent...
Happy hunting,
George
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Thu, 08 Feb 2001 23:52:22 -0500
From: Jud McCranie <[EMAIL PROTECTED]>
Subject: Re: Mersenne: P 4 Optimiztion
At 07:35 PM 2/8/2001 -0800, Terry S. Arnold wrote:
>I have heard rumblings that George has put together a version of the LL
>test optimized for the P 4. Does anyone know what the performance is like
>compared to a P 3?
The P4 doesn't do very well on the version that isn't optimized for the P4:
http://www.mersenne.org/bench.htm
+-----------------------------------------------------------+
| Jud McCranie |
| |
| Think recursively( Think recursively( Think recursively)) |
+-----------------------------------------------------------+
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Thu, 08 Feb 2001 23:55:34 -0500
From: Jud McCranie <[EMAIL PROTECTED]>
Subject: Re: Mersenne: P 4 Optimiztion
At 11:09 PM 2/8/2001 -0500, George Woltman wrote:
> George is working on it, but is a long way from completion. Progress is
slow, primarily due to my own laziness. My estimate for a 512K FFT is 0.4
seconds
>on a 1.4GHz P4.
It probably isn't that important right now because few people have a P4.
+-----------------------------------------------------------+
| Jud McCranie |
| |
| Think recursively( Think recursively( Think recursively)) |
+-----------------------------------------------------------+
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Fri, 09 Feb 2001 12:07:48 -0500
From: Jeff Woods <[EMAIL PROTECTED]>
Subject: Re: Mersenne: P 4 Optimiztion
At 11:09 PM 2/8/01 -0500, you wrote:
>George is working on it, but is a long way from completion. Progress is
>slow, primarily due to my own laziness. My estimate for a 512K FFT is 0.4
>seconds on a 1.4GHz P4. You can compare that to other machines at
>http://www.mersenne.org/bench.htm.
Can we assume you meant 0.04?
According to that site, A P-III 1.0 Ghz is already clipping such a number
at 0.145, and a P4 is already listed there doing such in the current
version at 0.126....
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Fri, 09 Feb 2001 14:36:27 -0500
From: George Woltman <[EMAIL PROTECTED]>
Subject: Re: Mersenne: P 4 Optimiztion
Hi,
At 12:07 PM 2/9/2001 -0500, Jeff Woods wrote:
>>My estimate for a 512K FFT is 0.4 seconds on a 1.4GHz P4. You can
>>compare that to other machines at http://www.mersenne.org/bench.htm.
>
>Can we assume you meant 0.04?
I wanted to make a safe prediction :)
Yes, I meant 0.04 seconds.
Egg on face,
George
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Fri, 09 Feb 2001 17:56:26 EST
From: [EMAIL PROTECTED]
Subject: Mersenne: Re: screen savers
Just my 2 cents' worth with respect to the screen saver
proposals: how about the following?
1) (This is along the lines of the popular "swarm of
bees" screensaver) Have some bee (or other - perhaps
allow the user to choose from a menu) icons move around
the user's computer screen according to simple dynamical
rules, e.g. allowing them to avoid collisions and
treating the window edges as either reflecting or
periodic boundaries. Every time an LL iteration
completes, use the bottom X ( X is a small integer )
bits of the interim LL residue, treated as a signed
quantity, to set the x-acceleration of widget #1, and
the next X bits for the y-acceleration. Widget #2 would
similarly use the next 2X bits, etc. We use the bits to
set the acceleration rather than the speed to keep the
movements from being too herky-jerky. We'd need to make
sure the computations involved are sufficiently simple
so as not to steal too much runtime from the actual LL
test.
We could also combine the above with some way of
tracking the progress of the current assignment. For
instance, if the number being tested is 2^p - 1 and the
user's screen has N pixels, then on average every p/n
iterations, block one more pixel of the imaginary bee
box, say in inward-spiraling fashion. As the test
proceeds, the walls of the box slowly close in.
(Admittedly, not a good screen saver for the
claustrophobic :)
2) For the folks who like something that looks more
scientific, we could have a vertical box that looks
like a control panel. Every iteration, the bottom Y
bits of the LL residue get printed to the bottom line
of the panel, and the others move up. We could colorize
the individual digits of the hex residues, or pick a
new color for each new Residue, so things look a bit
more dynamic. The faster the user's machine, the more
dynamic the output.
- -Ernst
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Fri, 09 Feb 2001 20:14:52 -0500
From: Nathan Russell <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: screen savers
On Fri, 09 Feb 2001 17:56:26 EST, you wrote:
>Just my 2 cents' worth with respect to the screen saver
>proposals: how about the following?
>
>1) (This is along the lines of the popular "swarm of
>bees" screensaver) Have some bee (or other - perhaps
>allow the user to choose from a menu) icons move around
>the user's computer screen according to simple dynamical
>rules, e.g. allowing them to avoid collisions and
>treating the window edges as either reflecting or
>periodic boundaries. Every time an LL iteration
>completes, use the bottom X ( X is a small integer )
>bits of the interim LL residue, treated as a signed
>quantity, to set the x-acceleration of widget #1, and
>the next X bits for the y-acceleration. Widget #2 would
>similarly use the next 2X bits, etc. We use the bits to
>set the acceleration rather than the speed to keep the
>movements from being too herky-jerky. We'd need to make
>sure the computations involved are sufficiently simple
>so as not to steal too much runtime from the actual LL
>test.
My hunch is that this wouldn't steal much time if the bees move by
jumping rather than sliding.
>We could also combine the above with some way of
>tracking the progress of the current assignment. For
>instance, if the number being tested is 2^p - 1 and the
>user's screen has N pixels, then on average every p/n
>iterations, block one more pixel of the imaginary bee
>box, say in inward-spiraling fashion. As the test
>proceeds, the walls of the box slowly close in.
>(Admittedly, not a good screen saver for the
>claustrophobic :)
The thing here is that this would make the rate of progress slow. I
know that when I first started GIMPS, I got a little depresseed
thinking things like "I've been here an hour and it's not even a third
of a percent done!"
There are possible ways to make it appear that work is being done
quickly- the Tower of Hanoi comes to mind, though we would need to
make double moves on some iterations, since runtimes can only be a
power of two steps.
Perhaps there could be an option to play a fanfare or some such every
time one of the six or seven bottom discs in the tower is moved.
>2) For the folks who like something that looks more
>scientific, we could have a vertical box that looks
>like a control panel. Every iteration, the bottom Y
>bits of the LL residue get printed to the bottom line
>of the panel, and the others move up. We could colorize
>the individual digits of the hex residues, or pick a
>new color for each new Residue, so things look a bit
>more dynamic. The faster the user's machine, the more
>dynamic the output.
That's also a fairly interesting idea.
Maybe the colorization could vary with the current iteration number in
some fashion - for example, we could have the lines of text each
broken into blocks of rainbow color, with each column of blocks
shifting according to the base 7 number of the current iteration - it
would start partway through, according to how far below the next
highest power of 7 the exponent was.
Then, as the run continued, the right-hand word in each line would
shift by one color for each iteration, wrapping around - when the
bottom line was solid violet, you'd be done.
Nathan
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Fri, 09 Feb 2001 20:37:36 -0500
From: Jud McCranie <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: screen savers
At 08:14 PM 2/9/2001 -0500, Nathan Russell wrote:
> The thing here is that this would make the rate of progress slow. I
>know that when I first started GIMPS, I got a little depresseed
>thinking things like "I've been here an hour and it's not even a third
>of a percent done!"
When I joined, I was getting a LL test in about 9 hours on my P-60. Now it
takes 3 months on my P-300.
+-----------------------------------------------------------+
| Jud McCranie |
| |
| Think recursively( Think recursively( Think recursively)) |
+-----------------------------------------------------------+
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Sat, 10 Feb 2001 08:23:04 -0500
From: "H David & Carol Friedberg" <[EMAIL PROTECTED]>
Subject: Mersenne: re 102357
Guys
I dont know where to send this so I am sending to the list and ask that it
be sent on to the right folk
It really is not for publication, but I would like an address of a contact
person.
I am 102357.
My computer 01 with a Pentium II at 233 mHz has died a horrible death from
excessive viremia
All calculations either stopped or proceeded so slowly as to be confused
with molasses
The work is lost
I have replaced it with this Celeron at 766 an I am calling it 001
I still have a Pentium II at 266 mHz and a Celeron at 366 running,
uninfected
Thats it just three machines all going 24 hours a day for GIMPS
Cheers
H David Friedberg MD
[EMAIL PROTECTED]
102357
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Sat, 10 Feb 2001 11:58:04 -0600
From: Ken Kriesel <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: screen savers
At 05:56 PM 2/9/2001 EST, [EMAIL PROTECTED] wrote:
>Just my 2 cents' worth with respect to the screen saver
>proposals: how about the following?
>
>2) For the folks who like something that looks more
>scientific, we could have a vertical box that looks
>like a control panel. Every iteration, the bottom Y
>bits of the LL residue get printed to the bottom line
>of the panel, and the others move up. We could colorize
>the individual digits of the hex residues, or pick a
>new color for each new Residue, so things look a bit
>more dynamic. The faster the user's machine, the more
>dynamic the output.
I had been thinking that it would be interesting to map the entire savefile,
one byte per pixel, into a rectangular area centered on the screen,
updating every time a new intermediate savefile is saved. I think
default is every 30 minutes, but that can be adjusted.
Below the graphic could be displayed the iteration number and exponent.
A historical bar graph of iterations throughput per minute, scaled relative
to the maximum ever achieved on that machine, for the current runlength,
is another. It has the advantageous property that some end users might
be encouraged to maximize their throughput via the easy feedback.
Make a change, wait, check the graph.
Ken
_________________________________________________________________________
Unsubscribe & list info -- http://www.scruz.net/~luke/signup.htm
Mersenne Prime FAQ -- http://www.tasam.com/~lrwiman/FAQ-mers
------------------------------
Date: Sun, 11 Feb 2001 18:12:56 +0000 (GMT)
From: Russel Brooks <[EMAIL PROTECTED]>
Subject: Re: Mersenne: Re: screen savers
I was just thinking that screen savers perform 2 functions;
1. The flashy display and
2. Locking the keyboard with a pw.
The second is more important and must be included otherwise it
will not be able to be used instead of the system lock.
Cheers... Russ
_________________________________________________________________________
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 #815
******************************
