🚀 go-pugleaf

In article <1993Apr27.184408.10491@ast.saic.com>, hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> Hello Netters,
|> 
|> We'd like to ask for your help (actually your CPU's help) in doing
|> a large scale parallel computation. This computation will give
|> upper and lower bounds on the area of the Mandelbrot set.
 ...
|> The mail server mset@kepler.ucsd.edu can send you 
...
|> 
|> To get a piece of the action (a segment to compute) send email to
|> mset@kepler.ucsd.edu.  The message should have the word "join",

Please no more 'join'  since the cells are all assigned. For some reason
the deamon will now send out cell #48 again, which has already been
assigned. If you want to experiment with #48, feel free. Just, please
do not send in the result to the  mset@kepler.ucsd.edu account.

Thank you for the overwhelming response. Many machines are expected to be
continuing the calculation for a week or two. Then we can all see the
result.  Don't feel left out if you didnot get in on GIPPP#3.  The
independant and fundamentally different approach to the Mandelbrot
Set area calculation known as GIPPP#2 (Robert P. Munafo - 
mrob@world.std.com) is still waiting to be run.

|> When the 'smoke clears' in a week or two or three, we'll have an
|> area estimate.  Your contributations will be logged and acknowledged.
|> 
|> Thanks,
|> 
|> Jay Hill
|> Yuval Fisher
|> 

Jay
-- 
{ hilljr@jupiter.saic.com } begin writeln(3*ln(640320)/sqrt(163):17:15) end.
void main(){double sqrt(), y=1/sqrt(2.), a=.5, m=1,z; int n=0;
for(;m*=2,z=sqrt(1-y*y),y=(1-z)/(1+z),a=a*(1+y)*(1+y)-m*y,n<3;n++);
printf("%17.15lf\n",1/a);}

In article <1993Apr30.204926.6970@ast.saic.com>
hilljr@jupiter.saic.com (Jay R. Hill) writes:
>
>Please no more 'join'  since the cells are all assigned.

Could this be done more efficiently?
Now some computational resources left over, and that means
that cells were too large and the accepted resources used more
time to calculate than needed.

Is it possible arrange a server, which process the calculation, to
each computer node used in computation, and a program, which shedules
smaller cells to nodes, to jupiter.saic.com or to some main internet
node.

This way
 -cells doesn't run out if late 'join' requests to project comes
 -slow nodes doesn't slow down the project by calculating large cell
  longer than fast nodes -- slow nodes may calculate couple smaller
  cells, instead of many smaller cells as calculated by faster nodes
 -it is automatic for nodes

Could this be done without security faults or misuse?

The scheduler has a huge list of computers (nodes) used to
the calculation. Communication is done either via e-mail or via socket
depending on is the site in Internet (or on is it possible to make a
special e-mail account for calculation at all).

Ideas?
Just imagine how many flops we could get out of the system.... ahhh..


Juhana Kouhia

In article <1993Apr30.204926.6970@ast.saic.com>, hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> In article <1993Apr27.184408.10491@ast.saic.com>, hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> |> Hello Netters,
|> |> 
|> |> We'd like to ask for your help (actually your CPU's help) in doing
|> |> a large scale parallel computation. This computation will give
|> |> upper and lower bounds on the area of the Mandelbrot set.
...
|> 
|> Thank you for the overwhelming response. Many machines are expected to be
|> continuing the calculation for a week or two. Then we can all see the
|> result.  Don't feel left out if you did not get in on GIPPP#3.  
...
There are a few cells which require very long computer run times still
available. We have seen runtimes from 1 minute to 6000 minutes on a 15MFlop
machine. That is the nature of the MSet. Some parts are easy, some hard.
If you have a fast machine available to run any of these special hard
cells, please email me directly. 

Of the 119 cells originally sent out, 77 are returned completed. The rest
are assigned except for these special ones. We had at least 41 different
people respond with about a dozen joining to late to get a cell. Some 
of you have a very fast machine and did not get a cell. If you do still
want to get in on this, let me know. 
|> 
|> |> When the 'smoke clears' in a week or two or three, we'll have an
|> |> area estimate.  Your contributations will be logged and acknowledged.
|> |> 
|> |> Jay Hill
|> |> Yuval Fisher
 
Thank you all for helping with GIPPP#3
 
Jay 'Not tonight honey, I have a fractal' Hill
 
{ hilljr@jupiter.saic.com } begin writeln(3*ln(640320)/sqrt(163):17:15) end.
void main(){double sqrt(), y=1/sqrt(2.), a=.5, m=1,z; int n=0;
for(;m*=2,z=sqrt(1-y*y),y=(1-z)/(1+z),a=a*(1+y)*(1+y)-m*y,n<3;n++);
printf("%17.15lf\n",1/a);}

In article <1993Apr30.204926.6970@ast.saic.com>, 
   hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> In article <1993Apr27.184408.10491@ast.saic.com>, 
   hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> |> Hello Netters,
|> |> We'd like to ask for your help (actually your CPU's help) in doing
|> |> a large scale parallel computation. This computation will give
|> |> upper and lower bounds on the area of the Mandelbrot set.
|>  ...
|> Thank you for the overwhelming response.  Many machines are expected
|> to be continuing the calculation for a week or two.

Make that yet another few weeks.  Of the 119 cells originally sent out,
97 are returned completed.  Thanks. Some folk are really struggling with
more difficult cells (taking up to 3 weeks on a 4MF SUN).  There are 
several 18MFlop and a few 50-70MFlop machines working hard on this. One
fellow offered a Cray, however we encourage 'nice' background running.
But at least 2 486/33 MSDOS machines got in a cell or two.  One guy has 
Spark10s which someone hits the red button on them every week or so :-(
Two have been killed by lightning.  There was a posting on sci.fractals
about the 9 billion pixels of God.  Hmmm. We could pass that any second
now!  If we finish, it will be equivalent to 274877906944 pixels.  
...
|> |> When the 'smoke clears' in a week or two or three, we'll have an
|> |> area estimate.  
And that was posted before the lightning. ;-)

|> |> Your contributations will be logged and acknowledged.
|> |> Thanks,  Jay Hill    Yuval Fisher

Jay 'Not tonight honey, I have a bifurcating fractal' Hill
-- 
{ hilljr@jupiter.saic.com } begin writeln(3*ln(640320)/sqrt(163):17:15) end.
void main(){double sqrt(), y=1/sqrt(2.), a=.5, m=1,z; int n=0;
for(;m*=2,z=sqrt(1-y*y),y=(1-z)/(1+z),a=a*(1+y)*(1+y)-m*y,n<3;n++);
printf("%17.15lf\n",1/a);} /* 9 billion digits of pi, anyone? */

In 1990 Christopher Kane wrote (see 'FRACTAL' discussion list <FRAC-L@GITVM1>
Fri, 12 Jan 90 11:59:00 CST Subject: RE: Area of the M-Set; archive filename
is NOTEBOOK, 60% into file)

  "About two months ago, I send a query to the list regarding the area of the
Mandelbrot Set. Well, the answer seems to be 1.508 to four significant digits.
This is according to Dave Rabenhorst in The Journal of Chaos and Graphics,
vol. 2 August 1987.  He uses a brute-force approach to finding this result.
Basically, he took a typical Mandelbrot calculating/displaying program and
modified it to count the number of points that fall in the set and dividing
by the number of points in the square (-2,2)-(2,-2) that surrounds the set."

The University of California San Diego library doesn't seem to have 
The Journal of Chaos and Graphics.  Does this journal really exits? :-)
The value 1.508 is amazingly close to what we are now getting, especially
considering Rabenhorst apparently used 129x129 pixels!  Does anyone have
any earlier references to Mandelbrot Set area calculations?  The first
estimate could be from Dr. Mandelbrot's first ASCII printout of the set.

Jay 'Not tonight honey, I have a bifurcating fractal' Hill
-- 
{ hilljr@jupiter.saic.com } begin writeln(3*ln(640320)/sqrt(163):17:15) end.
void main(){double sqrt(), y=1/sqrt(2.), a=.5, m=1,z; int n=0;
for(;m*=2,z=sqrt(1-y*y),y=(1-z)/(1+z),a=a*(1+y)*(1+y)-m*y,n<3;n++);
printf("%17.15lf\n",1/a);}

Re:GIPPP#3 paper now available.
In article <1993May14.214702.29442@ast.saic.com>, hilljr@jupiter.saic.com (Jay R. Hill) writes:
|> Hello Netters,
|> We'd like to ask for your help (actually your CPU's help) in doing
|> a large scale parallel computation. This computation will give
|> upper and lower bounds on the area of the Mandelbrot set.
|>  ...
Only a week or two to go.  Of the 119 cells originally sent out, 105 are 
returned completed.  Thanks.   A paper (post script) describing the method
and preliminary results is now available at mset@kepler.ucsd.edu. Just
email the word 

paper

and "Bounding the Area of the Mandelbrot Set" by Yuval Fisher and Jay Hill
will be sent to you by email.  We give results of a test run to depth 17
(thank you Bill Broadley <Broadley@neurocog.lrdc.pitt.edu>).  The 
Grand Internet Parallel Processing Project (GIPPP#3) is collecting
calculations by at least 38 subscribers to the net. These calculations
make an area estimation to depth 19 (equivalent to 4^19 pixels). We now
know the upper bound is less than pi/2 and the lower bound is greater 
than 1.503.  I estimate there are about 1000 million million floating
point operations in this calculation on more than 50 machines.  With
lightning stopping two sites from completing their cells,  several
sites having fingers applied to the big red button, and SysAdmin types
trying to maintain machine idleness :-), it's been an interesting project.

Thanks again for all who have or tried to participate.  Only 14 cells to
go.  It has been just one month, not too bad.

|> Thanks,  Jay Hill    Yuval Fisher
 
Jay 'Not tonight honey, I have a bifurcating fractal' Hill
-- 
{ hilljr@jupiter.saic.com } begin writeln(3*ln(640320)/sqrt(163):17:15) end.
void main(){double sqrt(), y=1/sqrt(2.), a=.5, m=1,z; int n=0;
for(;m*=2,z=sqrt(1-y*y),y=(1-z)/(1+z),a=a*(1+y)*(1+y)-m*y,n<3;n++);
printf("%17.15lf\n",1/a);} /* 9 billion digits of pi, anyone? */