Re: Motion P1788/M0013.04 - Comparisons - Overflow / Infinity
Ian McIntosh wrote:
> I'm still unenthused about 0 times Infinity producing 0 because it needs
> either special instructions I don't expect to see or special checks after
> every operation.
In machines with floating-point control registers it might be possible
to have an "Interval-assist" mode in which *directed* rounding uses the
alternate inf*0 rule. The Vienna proposal mentioned a few cases where
the IEEE 754 Infinity rules would have to be bent in order to permit
efficient computation with interval endpoints.
Michel.
P.S. We (same group of correspondents) have been here before,
in April 2010 and August 1009. I also referred then to
a discussion in the stds-754 group in June 2007.
---Sent: 2010-09-20 23:27:22 UTC