The Entries:

ocia@tekig5.pen.tek.com (Stephen Blazo)

From Steve Blazo who passed along the program source code, although
he did not write it.  The program does not print to disk, and is quite
slow.

   "I don't know if you can use the following or not. I will work on
    getting an executable.  If all else fails I will put something in
    the mail to  you. This was put on the net (in a trivially
    different form) maybe a year ago.  I took it, modified it,
    compiled it, and it seems to work at least up to several thousand
    digits. I need to find my photocopy of the 100,000 digits of pi
    from Mathematics of Computation, 16, No. 77, Jan 1962, pp 76-99,
    to check the answer.

    'Spigot' algorithm originally due to Stanley Rabinowitz."

Look for BLAZO.TTP, and enter the number of digits of PI in the
command line.


                * * * * * * * * * * * * * * 

Gio Ciampa <GIO@wolfson-computer-laboratory.birmingham.ac.uk>

GIO_PI.PRG - Check the biography for Gio in the GIO folder.  Gio uses
both the 2 step Machin algorithm (GIO_PI.PRG) and the 3 step Stormer
variation (GIO_PI3B.PRG).  Just double-click on either program, and
watch it go: the second one is 11% faster than the first, demonstrating
that the right *algorithm* is essential.

I got to know Gio the best in all this phoo-fer-rah, as we passed
dozens of Internet messages back and forth.


                * * * * * * * * * * * * * * 

Stephan Olbrich    17.10.1993
Wullwinkel 38
D-30823 Garbsen
Germany

Tel.:  xx49137-13627
Email: olbrich@rrzn.uni-hannover.d400.de

PI4.TOS - from Stephan Olbrich of Garbsen, Germany.  This one uses
the two step Machin Algorithm - but incorporates multiprecision
arithmetic techniques - thus rendering it 5 times faster than
GIO_PI.PRG.  It proves the power of the multiprecision arithmetic.  It
also writes to either screen or disk.

Warning - this will accept requests of over 17,650 digits, but will do
strange things if you do - one of which is *not* accurately
calculating PI.  If you plug in 25000, you get a 3.8.... value.  If you
plug in 50000, you get PI accurately to about 1200 digits, then the
next 48800 are all wrong.  I have asked the author to iron out the
bugs.



Times:

               PI4.TOS              GIO_PI3B.PRG         BLAZO.TTP

  100 digits            .24 sec                .79 sec          1.5 sec
 1000 digits          16.29 sec              63.38 sec        123   sec
10000 digits   25 min 54 sec         123 min 56 sec      200 min flat.
 
Calc's up to    17,650 digits          32,254 digits      48,370 digits


I am hoping for someone to write a program which makes use of *both*
of multiprecision arithmetic *and* one of the Borwein algorithms. 
Also, a program that will not be limited, other than by size of RAM.

Any takers?


Stuart A. Lyster.
s.lyster@genie.geis.com
slyster@outb.wimsey.bc.ca


My thanks to those who expressed interest and commented as the contest
went on:

mith@RMIT.EDU.AU (James Alan Hall)
ocia@tekig5.pen.tek.com (Stephen Blazo)
mark anthony kolesar <kolesar@cis.ohio-state.edu>
sanders@sci.kun.nl
pwp@cs.indiana.edu
wells_s@kosmos.wcc.govt.nz
tony@microware.co.uk (Tony Mountifield)
ocia@tekig5.pen.tek.com (Stephen Blazo)
warwick@cs.uq.oz.au
guerard@etca.fr
Steve.Yelvington@lambada.oit.unc.edu
George Kirkaldie <gkirkaldie@sanity.tdkcs.waterloo
" (Felix Rauch)" <rauch@avalon.unizh.ch>
Gio Ciampa <GIO@wolfson-computer-laboratory.birmingham.ac.uk>
rened@htsa.aha.nl (Rene Drost)
dvasilef@copper.ucs.indiana.edu
jmcdonal@garnet.acns.fsu.edu

