0
Fixed

Inconsistency in handling division by zero

Ash 5 years ago updated by Pavel Holoborodko 5 years ago 2

Using double precision:

>> 1/0

ans =

Inf

And :

>> 0^(-1)

ans =
Inf

The results should be consistent using either method as shown above. However, using quad precision, these computations do not give consistent results:

>> 1/mp(0)

ans =
Inf
>> mp(0)^(-1)
ans =
Inf - NaNi

I'm not sure what's under the hood, but it seems to me that either computation should give the same answer.

Under review

In power function we handle inputs like complex numbers by default. Semantic for various singular cases in complex arithmetic is not defined uniformly and implementation dependent (e.g. see here: http://www.advanpix.com/2015/10/19/devnotes-1-multiplicative-operations-with-complex-infinity/)


That is the reason for such misbehavior. We will fix this today.


Thank you for reporting the issue

Fixed

This bug has been resolved in Windows version. You can download it from our website.

GNU Linux & Mac OSX will follow today.