home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.lang.c
- Path: sparky!uunet!haven.umd.edu!darwin.sura.net!mips!sdd.hp.com!news.cs.indiana.edu!umn.edu!thompson
- From: thompson@atlas.socsci.umn.edu (T. Scott Thompson)
- Subject: Re: POWER OF LARGE NUMBERS
- Message-ID: <thompson.714074422@daphne.socsci.umn.edu>
- Sender: news@news2.cis.umn.edu (Usenet News Administration)
- Nntp-Posting-Host: daphne.socsci.umn.edu
- Reply-To: thompson@atlas.socsci.umn.edu
- Organization: University of Minnesota
- References: <1992Aug12.135106.12924@cs.uow.edu.au> <thompson.713734408@daphne.socsci.umn.edu> <1992Aug16.052157.1991@organpipe.uug.arizona.edu> <QUANSTRO.92Aug16132245@mari.StOlaf.edu> <1992Aug17.060243.11404@organpipe.uug.arizona.edu>
- Date: Mon, 17 Aug 1992 18:00:22 GMT
- Lines: 47
-
- dave@cs.arizona.edu (Dave Schaumann) writes:
-
- >In article <QUANSTRO.92Aug16132245@mari.StOlaf.edu>, quanstro@mari (goon) writes:
- >>In article <1992Aug16.052157.1991@organpipe.uug.arizona.edu> dave@cs.arizona.edu (Dave Schaumann) writes:
- >>
- >> >> Does any one know how to raise power of possibly 100 digits
- >> >>to possibly 100 digits. Any algorithm or code will be appriciated.
- >> >
- >> >Is there any hope of solving this?
- >>
- >> Only if you're willing to give up most of the significant digits.
- >[...]
- >>
- >>You just contradicted yourself. It is possible. It just takes more
- >>memory and more programming (you can't expect the hardware to handle
- >>this sort of case). Check out the man page for mp (a set of arbitrary
- >>precision arith. functions) if you are on a UNIX system. There is also
- >>a GNU version of these functins.
-
- >Take a look at the specification again. We're talking about calculations
- >that will yield results with 10 ** 10 ** 100 digits. Try to do that with
- >a so-called `arbitrary precision' software package, and you'll find out
- >that it wants memory for that precision. We're talking about numbers so
- >hugely vast we can't even talk about how many *digits* they have without
- >using exponents! A megabyte has (about) 10**6 bytes in it, so we're talking
- >about 10**94 megabytes just to hold the result. Suppose you can load 1
- >megabyte in a 10**-12 seconds (new warp-drive technology), it would take you
- >10**78 seconds to load your value. That's more than 10**71 years.
-
- >So I restate my point: yes, you can write software to deal with numbers
- >that large, but NOT without dropping 99.9% of your accuracy.
-
- Actually it is worse than that since the number of digits needed is
- (roughly!)
- 10 ** (10 ** 100)
-
- which is _much_ larger than (10 ** 10) ** 100. (Remember folks,
- exponentiation is _not_ associative!)
-
- I am no physicist but I'll go out on a limb with this conjecture:
-
- No "arbitrary precision" package can compute the solution because the
- number of bits needed to represent the answer exceeds the number of
- atoms in the universe.
-
- (Well, I suppose we could wait for the new "quantum" computers to
- come along. :-) )
-