| Thread Links | Date Links | ||||
|---|---|---|---|---|---|
| Thread Prev | Thread Next | Thread Index | Date Prev | Date Next | Date Index |
Arnold On 11 Sep 2009, at 11:04, Arnold Neumaier wrote:
In any case, if the vector has length 1, the result should be the same as for converting a float datum to an interval. Whatever the result of converting a single inf or NaN will be decided upon, that must be the result of a sum that evaluates exactly to NaN or inf.
Yes, I agree on that. John