Infinite loop on exponent somehow

[jrmoserbaltimore]

I don't know what particularly causes this, but it hangs forever.

from fxpmath import Fxp
x = Fxp(0,dtype='UQ0.64')
x**4

[francof2a]

That is because when you use a constant as is, in this case the number 4, as operator of a Fxp object, this constant is converted to an Fxp object with same dtype. Then, the power operation are using a huge raw number as exponent.

You can solve this being explicit about dtype of the exponent:

x**Fxp(4)
# or
x**Fxp(4, dtype='some fixed-point dtype')

[jrmoserbaltimore]

Interesting. Should this produce a warning or raise an exception (probably the best route) or be a different behavior, or is that just impossible? It's not obvious what's happening and an infinite loop is probably not the programmer's intent.

[francof2a]

Yes, it could be! I think a good change can be modify the op_input_size default value from 'same' to 'best'. In the first case the number 4 is converted to UQ0.64 before to perform the power operation. In the second case, using 'best' the number for is converted to Q4.0. This behavior will prevent long calculation time and intensive memory usage.

But the idea of the warning is really important in order be aware of this automatic sizing. Definitely I'll include this in next release.

[francof2a]

solved in v0.4.9