comp.lang.ada
 help / color / mirror / Atom feed
From: georg@ii.uib.no (Hans Georg Schaathun)
Subject: Re: Large numbers (or is Ada the choice for me?)
Date: 10 Mar 2001 16:49:41 GMT
Date: 2001-03-10T16:49:41+00:00	[thread overview]
Message-ID: <slrn9akmp5.n5v.georg@admiral.ii.uib.no> (raw)
In-Reply-To: 98bp0c$nsq$1@nh.pace.co.uk

On Fri, 9 Mar 2001 18:28:43 -0500, Marin David Condic
  <marin.condic.auntie.spam@pacemicro.com> wrote:
: Well, let me check a few assumptions. 1) We can never run out of numbers.
: (Go ahead. Use all you want. We'll make more! :-) 2) We *can* and *will*
: eventually run out of memory. Hence, even if you did all the math with some
: sort of fractional representation rather than a decimal representation, it
: would be possible to construct numbers that exceed the capacity of the
: machine.

That is just as simple and likely as creating a problem it takes 50 years
to solve (with infinite memory resources).  The solution is rather simple,
the program returns error, `sorry, sam, your problem is to large'.

I am certainly aware that I won't be able to solve all the problems
I might want to create, but test runs in maple indicates that _this_
problem is solvable if memory is handled with care.

:           Hence, I think it stands to reason that you would be off on a
: fool's errand to insist on no approximations or limitations whatsoever.

I don't insist on no limitations.  I accept the limitations of memory 
and CPU resources, but I want to do as much as possible within these
limits.  My problem is primarily to determine whether the system has
integer solutions or not, which means that I can't accept (not even risk)
any rounding errors of � or more.  I doubt that floating point can save 
significant amounts of memory with this requirement.

: There has to be some sort of practical upper limit imposed by the available
: memory if nothing else.

Yes, so I won't cry over problems which can't be solved.

: As I said elsewhere, I rather hastily picked a bad example - but I think the
: point still stands that one will have to live with some sort of
: approximation on the representation - even if in practice, it may be so
: small as to not matter.

Wrong, my outset is to solve a couple of practical problems, not to 
solve any problem you might think of.

:-- Hans Georg
-- 
Signature en panne.



  reply	other threads:[~2001-03-10 16:49 UTC|newest]

Thread overview: 22+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2001-03-09 18:58 Large numbers (or is Ada the choice for me?) Hans Georg Schaathun
2001-03-09 19:35 ` Marin David Condic
2001-03-09 20:44   ` David Starner
2001-03-09 23:12     ` Marin David Condic
2001-03-10  2:56       ` David Starner
2001-03-10 11:37         ` Florian Weimer
2001-03-10  6:08       ` tmoran
2001-03-09 21:01   ` Randy Brukardt
2001-03-09 23:02   ` Robert A Duff
2001-03-09 23:28     ` Marin David Condic
2001-03-10 16:49       ` Hans Georg Schaathun [this message]
2001-03-10 11:59   ` Jeffrey Carter
2001-03-09 20:37 ` Brian Catlin
2001-03-09 21:26 ` JP Thornley
2001-03-09 21:59 ` Tucker Taft
2001-03-15  8:33   ` Modular type (Re: Large numbers) Hans Georg Schaathun
2001-03-15 10:58     ` Florian Weimer
2001-03-15 11:12       ` Hans Georg Schaathun
2001-03-15 16:24         ` Tucker Taft
2001-03-10  1:42 ` Large numbers (or is Ada the choice for me?) Keith Thompson
2001-03-19 20:48 ` Robert I. Eachus
2001-03-20  3:33   ` Brian Rogoff
replies disabled

This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox