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

Motion P1788/M0033.01 Number Format



Alan Eliasen wrote, in his NO justification:
>    Currently, the text of the motion is ambiguous or impossible to meet
> if the implementation of the numerical type uses arbitrary-precision
> numbers.  (And good numerical implementations will have arbitrary
> precision numbers.)

I thought this point had been addressed in Vincent's Rationale:

| On A1, we require the subset to be finite in order to be able to
| specify the rounding without ambiguity. A discrete subset would be
| sufficient for this purpose, but P1788 will probably also need the
| notion of smallest and largest finite element (see the discussions
| about the midpoint). Arbitrary precision formats need to be
| parameterized (for instance, by the precision and possibly the
| exponent size), so that each format instance is finite.


>    In order for me to vote YES on this motion, it would need to be
> amended to allow a constraint on the *precision* as well as the rounding
> direction, and remove text about "closest" value.

I think an implementation that passes a precision constraint as an
additional parameter in functions where this matters would be conforming.

> ...                         If the exact precision constraints are not
> meetable for any reason, then the implementation should be allowed to
> return a less "sharp" value that still preserves containment.

This gets into the tightest/accurate/valid qualification debate (see
Vienna proposal), which has not yet come up in a Motion.

Michel.
---Sent: 2012-06-06 09:54:31 UTC