[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: reading NaNs



 | From: "Marcin 'Qrczak' Kowalczyk" <qrczak@xxxxxxxxxx>
 | Date: Tue, 25 Oct 2005 02:39:04 +0200
 | 
 | Aubrey Jaffer <agj@xxxxxxxxxxxx> writes:
 | 
 | >  | >   An implementation may report a violation of an
 | >  | >   implementation restriction in any calculation for which
 | >  | >   the result would be 0/0.
 | >  | 
 | >  | It's not compatible with IEEE-754 because it doesn't guarantee
 | >  | that by default operations like (/ 0.0 0.0) produce NaN.
 | >
 | > Aren't there IEEE-754 modes which throw an exception rather than
 | > returning NaN?
 | 
 | There are, there are useful too, but they must be explicitly turned
 | on.  So the above description is good as long as "may" means that
 | the programmer chooses the behavior.
 | 
 | > If the hardware returns a NaN, what the Scheme implementation
 | > does with the NaN is not within the scope of IEEE-754.
 | 
 | I disagree. IEEE-754 specifies the language-independent interface,
 | as seem from the point of view of the programmer.

According to IEEE-754:

  user.  Any person, hardware, or program not itself specified by this
  standard, having access to and controlling those operations of the
  programming environment specified in this standard.

A Scheme implementation is a program; thus it qualifies as a user.
IEEE-754 does not constrain its behavior.