Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

Re: Motion P1788/M007.01_NaI



Vincent Lefevre schrieb:
On 2009-08-24 19:50:56 +0200, Arnold Neumaier wrote:
Michel Hack schrieb:
Nevertheless I agree that when a product of endpoints is of the
form 0*oo the indefiniteness of the limit may indeed play a role.
Arnold Neumaier however seems to have good reasons for expecting
0*oo = 0 in this case too.
which are spelled out in Part 7 of the Vienna Proposal.

If I understand correctly, this would be a useful property of
FP arithmetic which would be nice to have in order to implement
interval arithmetic efficiently (but this is not an operation
in IA). However more generally in FP, a NaN result would be
more useful for the user (because of indeterminate forms), and
it is the only conforming result for IEEE 754.

This would just mean that if a processor wants IA to be implemented
more efficiently, it could have some (preferably static) attribute
to return either NaN or 0.

It simply means that interval bounds don't behave in this case like
the generic case, so that a special treatment is benefician.

Note that Part 7 of the Vienna Proposal should only be a recommendation for the next revision of IEEE 754, and a guide for hardware implementors who can make decisions independent of IEEE 754.


Arnold Neumaier