[Cfp-interest] change for canonicalize

[Jim Thomas]

Here's a change I came up based on our teleconference discussion today. Please review ASAP.

-Jim

14.9  The canonicalize functions

IEC 60559 requires an arithmetic convertFormat operation from each format to itself. This operation produces a canonical encoding and, for a signaling NaN input, raises the “invalid” floating-point and delivers a quiet NaN. C assignment (and conversion as if by assignment) to the same format may be implemented as a convertFormat operation or as a copy operation. The suggested change to C11 below provides the IEC 60559 convertFormat operation.

Suggested change to C11:

As the last subclause of 7.12.11, add:

7.12.11.n The canonicalize functions

Synopsis

            #include <math.h>

            int canonicalize(double * cx, const double * x);

            int canonicalizef(float * cx, const float * x);

            int canonicalizel(long double * cx, const long double * x);

Description

The canonicalize functions attempt to produce a canonical version of the floating-point representation in the object pointed to by the argument x and store the canonical result in the object pointed to by the argument cx. If the input *x is a signalling NaN, the canonicalize functions are intended to store a canonical quiet NaN. If a canonical result is not produced the object pointed to by cx in unchanged.

Returns

The functions return zero if a canonical result is stored in the object pointed to by cx. Otherwise they return a nonzero value.

After new F.10.8.4, add:

F.10.8.5 The canonicalize functions

The canonicalize functions produce[1] the canonical version of the representation in the object pointed to by the argument x. If the imput *x is a signaling NaN, the "invalid" floating-point exception is raised and a (canonical) quiet NaN (which should be the canonical version of that signaling NaN made quiet) is produced.  For quiet NaN, infinity, and finite inputs, the functions raise no floating-point exceptions.


[1] As if x*1e0 were computed.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mailman.oakapple.net/pipermail/cfp-interest/attachments/20120209/02a59534/attachment-0001.html