Re: [eclipse-users] will IC work with 64-bit integers ?

From: Joachim Schimpf (Independent Contractor) <jschimpf_at_cisco.com>
Date: Fri, 29 Feb 2008 13:50:47 +0000
Evgeni Kornykhin wrote:
> Hi!
> Now "ic" deals with integers of bitlength up to 51 correctly. Is it 
> planned to expand its functionality for dealing with integers of 
> bitlength up to 64 ?

By design, lib(ic) uses double floats (53 bits mantissa) to represent
the bounds of both reals and integers.  Integers are just reals with
an additional integrality flag.  There are no plans to change that.

Outside the range -2^53+1..2^53-1, integers can still be handled,
but ic represents their bounds in an accurate, but imprecise way
(analogous to the accurate but imprecise representation of reals
with float intervals).

Maybe you can describe your problem, and someone can suggest a better
model for it.  It is very unlikely that finite domain technology will
be able to solve a model with that kind of astronomical domains.

-- Joachim
Received on Fri Feb 29 2008 - 13:51:09 CET

This archive was generated by hypermail 2.2.0 : Thu Feb 02 2012 - 02:31:58 CET