Maintain global phase

[aaronszasz]

Feature request: an option when compiling to maintain the global phase of the operator

[edyounis]

Awesome suggestion! In 1.2, we will update the instantiation pipeline a bit, and this should definitely go in there. In the meantime, if this is important or timely for you, we can accomplish this by building a cost function for it.

[WolfLink]

I think the difficulty here is it’s hard to consistently define a global phase.

For example, consider a circuit containing gate Rz(pi/4). There are two different conventions for defining the unitary of Rz(pi/4), and they have global phases e^(pi/8) off from each other. So BQSKit might consider a circuit to match the desired phase or not, depending on which convention is used.

Note that in the actual convention that BQSKit uses, Rz(pi) and Z differ by a global phase.

[edyounis]

We can always easily add a global phase or correct it at the end. Don't worry about doing it during compilation.

[aaronszasz]

I think there is a well-defined choice of global phase, namely that before/after compiling, the get_unitary() method should give the same matrix with the same phase. So everything should be internally consistent within bqskit, even if other programs that read qasm files later might be inconsistent