Add cirq.PhasedXPowZPowGate #2566
Conversation
- Add cirq.CircuitDiagramInfoArgs.format_real helper method This gives a canonical single qubit operation to optimize into when doing microwave rotations with virtual Zs
cirq/ops/phased_x_z_gate.py
Outdated
| x_exponent: The $x$ in $Z^z Z^a X^x Z^{-a}$. | ||
| z_exponent: The $z$ in $Z^z Z^a X^x Z^{-a}$. Note that the $Z^z$ | ||
| operation happens last. | ||
| axis_phase_exponent: The $a$ $Z^z Z^a X^x Z^{-a}$. |
There was a problem hiding this comment.
I like these argument names
…ecompose - This code is not very performant, but it does the job
|
|
||
| @value.value_equality(approximate=True) | ||
| class PhasedXZPowGate(gate_features.SingleQubitGate): | ||
| """A single qubit operation expressed as $Z^z Z^a X^x Z^{-a}$. |
There was a problem hiding this comment.
This expression is matrix multiplication with time going to the left? Would be nice to make this clear so readers don't misinterpret this as a "circuit" with time going to the right. (This is implicit in the constructor docstring for z_exponent, but would be good to be clear up front.)
cirq/ops/phased_x_z_gate.py
Outdated
| axis_phase_exponent: Union[numbers.Real, sympy.Basic]) -> None: | ||
| """ | ||
| Args: | ||
| x_exponent: The $x$ in $Z^z Z^a X^x Z^{-a}$. |
There was a problem hiding this comment.
nit: I'd suggest not repeating this expression. Can say something like The $x$ in the gate expression above and refer back to the class docstring. Same for the following params.
There was a problem hiding this comment.
In many contexts, "above" will not refer to anything. For example, autocomplete popups in IDEs listing the parameter will not also list the docstring of the class. I tried to make these clearer.
| z = self.z_exponent | ||
| a = self.axis_phase_exponent | ||
|
|
||
| # Canonicalize X exponent into (-1, +1]. |
There was a problem hiding this comment.
Could also canonicalize x to [0, 1] and adjust the axis phase accordingly if needed. This is generally how we do things internally so that the pulse amplitude is always a positive number.
There was a problem hiding this comment.
Acknowledged. The exact canonicalization is not too important; it's only really relevant for equality. The user doesn't see it.
| if z > 1: | ||
| z -= 2 | ||
|
|
||
| # Canonicalize axis phase exponent into (-0.5, +0.5]. |
There was a problem hiding this comment.
As mentioned above, it would match our internal convention more closely if we instead canonicalize x to [0, 1] and a to (-1, 1].
There was a problem hiding this comment.
Acknowledged. The exact canonicalization is not too important; it's only really relevant for equality. The user doesn't see it.
cirq/ops/phased_x_z_gate.py
Outdated
| q = qubits[0] | ||
| yield ops.Z(q)**-self._axis_phase_exponent | ||
| yield ops.X(q)**self._x_exponent | ||
| yield ops.Z(q)**(self._z_exponent + self._axis_phase_exponent) |
There was a problem hiding this comment.
Nit: write this as self._axis_phase_exponent + self._z_exponent.
cirq/ops/phased_x_z_gate.py
Outdated
|
|
||
|
|
||
| @value.value_equality(approximate=True) | ||
| class PhasedXZPowGate(gate_features.SingleQubitGate): |
There was a problem hiding this comment.
I wonder if we should change the name here; I think of a "Pow" gate as a gate that can be raised to an exponent, which is not the case here, at least as implemented. I guess we have a similar issue with PhaseXPowGate.
There was a problem hiding this comment.
I renamed to PhasedXPowZPowGate.
Alternatives could be...
cirq.google.MicrowaveGate
cirq.google.MicrowaveFrameGate
PhysicalXVirtualZsGate
There was a problem hiding this comment.
I personally like WZ, but I guess we've moved away from W for the phase X gate, so that's probably too cryptic...
There was a problem hiding this comment.
PhasedXPowZPowGate sounds like something someone would come up while satirizing cirq's gate naming conventions
There was a problem hiding this comment.
What about PhasedXZGate
There was a problem hiding this comment.
Renamed to PhasedXZGate.
But what about AbstractPhasedXZOperationFactory?
There was a problem hiding this comment.
Reminds me of https://hackernoon.com/a-factoryfactoryfactory-in-production-822478b5afbd by a former coworker :)
cirq/ops/phased_x_z_gate.py
Outdated
| args: 'cirq.CircuitDiagramInfoArgs') -> str: | ||
| """See `cirq.SupportsCircuitDiagramInfo`.""" | ||
| return (f'PhXZ(' | ||
| f'p={args.format_real(self._axis_phase_exponent)},' |
There was a problem hiding this comment.
Call this a instead to match constructor args?
# Conflicts: # cirq/protocols/circuit_diagram_info_protocol.py
Strilanc
added
the
automerge
CirqBot
added
the
front_of_queue_automerge
CirqBot
removed
automerge
This gives a canonical single qubit operation to optimize into when doing microwave rotations with virtual Zs