home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!mcsun!sun4nl!dutrun!dutiws!dutiag!hdev
- From: hdev@dutiag.tudelft.nl (Hans de Vreught)
- Newsgroups: sci.math
- Subject: Re: Looking for fast methods of computing PI
- Message-ID: <hdev.719107527@dutiag>
- Date: 15 Oct 92 00:05:27 GMT
- References: <1992Oct13.025820.4593@eecs.nwu.edu> <-g1zrgc@rpi.edu>
- Sender: news@dutiws.tudelft.nl
- Lines: 29
-
- fokp@cary112.its.rpi.edu (Pingtao (Ben) Fok) writes:
-
- >I don't know what method you are talking about but I have been
- >thinking this problem for a long time and I found a way(the hard way)
- >to find PI. I use the idea of polygon in a circle. I first use
- >a square to start with. Then I turn the square into a octagon and find
- >the length of the side in terms of square root of 2. And then I
- >double the number of side again and again and again.....
- >It works fine but the algebra is heavy. I figure square root of 2
- >is easier to find than PI. A friend of mine used taylor's series
- >to find Pi up to 5000 digit.
-
- The method with polygons dates back to the Greeks. Taylor series (Machin's
- formula) works fine with this small number of digits. But the method used for
- the record is based on the AGM (arithmetic and geometric mean). The method of
- Brent and Salamin is well known (which is one such a method):
- a_{n+1}=(a_n+b_n)/2 (arithmetic mean)
- b_{n+1}=sqrt(a_n*b_n) (geometric mean)
- c_{n+1}=(a_n-b_n)/2
- 2(a_{n+1})^2
- pi_n=----------------------
- 1-sum(2^m*c_m^2,m=0,n)
- Roughly pi_{n+1} has twice as many correct digits as pi_n. Look in the faq
- list for the references.
- --
- Hans de Vreught | John von Neumann:
- hdev@dutiaa.tudelft.nl | Young man, in mathematics
- Delft University of Technology (TWI-ThI) | you don't understand things,
- The Netherlands | you just get used to them.
-