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

RE: 754R vs pow



In response to Mark Erle's note, Dave James wrote:

We would have to also exclude

  * every rounded operation that could underflow (and round differently)
                                                    +++++++++++++++++++++

The latter is not an issue; the results (in default exception handling
mode) are identical in the two options.  The only difference is whether
the Underflow flag is asserted.

  * processing of sNaNs and qNans  (that encode differently)

Here, the issue goes beyond encoding (which could be covered up
by explicit access functions, as Dave is about to propose):  NaN
propagation would have to be standardised for multiple-operand
cases.

Michel.

Sent: 2007-06-01 20:23:22 UTC

754 | revision | FAQ | references | list archive