Use integer power "**" operator for integer types

[MaxGraey]

Once day Vitalik Buterin published this twit about one counter example in Javascript using Fermat's Last Theorem:

57055**3 + 339590**3 == 340126**3
> true

That's happened because JS/TS using float point numbers all the time except logical ops.
In AS we have more rich set of types but power (**) operator still use Math.pow under the hood. This unfortunately cause to the same problem as in JS/TS and increase size of code btw.

[dcodeIO]

There is no specific reason to use floating point Math.pow for everything, except, maybe, that this is what JS does. I agree though that integer-specific exponentiation would be a nice addition.

[MaxGraey]

In python for integers operands, the result has the same type as the operands (after coercion) unless the second argument is negative in this case pow '**' return float result. Will be great have the same behaviour

[dcodeIO]

Do you know a good ipow (or how it's named) implementation that we could implement? :)

[MaxGraey]

Hmm, I don't know about performance of my implementation but I tried speedup ipow calculation as much as possible. We could get this as proof of concept but for smaller types implementation could be more simplicity without some extra checks. Also compiler should provide precomputing cases when all arguments is constant like 57055**3. I could start implement this for std, WDYT?

[dcodeIO]

@MaxGraey Has this been solved by your PR?

[MaxGraey]

Not yet. It require some stuff from compiler side as well.

[dcodeIO]

Closing this issue as part of 2020 vacuum as it is now tracked by the respective PR, which is nearing completion.

[MaxGraey]

It still WIP here: #1146 and will finished soon

[dcodeIO]

Closed this because I thought that having both an issue and a PR, that is nearly finished, isn't necessary. But if you prefer to keep it open, I'm ok with that :)

[MaxGraey]

Implemented in #1146