Subject: Re: CMUCL18d on Alpha?
From: Erik Naggum <>
Date: Fri, 19 Apr 2002 19:16:44 GMT
Newsgroups: comp.lang.lisp
Message-ID: <>

* Bulent Murtezaoglu
| Yes, but if you assume that the numbers read have been printed by some 
| other program that uses the same internal float representation, then this
| complication disappears, does it not?  

  No.  The problem lies in the rounding errors that can easily be
  introduced by the division and multiplication by (powers of) 10.  You
  have to be _very_ careful with these operations to avoid introducing
  errors.  If done incorrectly, they may even be relatively large and lead
  to propagating problems.

| That is, given A.BdEE, printed from an IEEE double in decimal radix with
| B l digits log, it seems we should be able to assume that (* (+ Ad0 Bd-l)
| 1dEE) will retrieve the original IEEE double.  At least I don't see how
| this would fail.

  Then try it out.  This is easy as long as you have an ample supply of
  precision bits, but that is not true for the least significant bits.

  In a fight against something, the fight has value, victory has none.
  In a fight for something, the fight is a loss, victory merely relief.

  Post with compassion: