Closed speller26 closed 8 months ago
As far as tests go, the existing tests actually cover the implementation changes, since the transpiler target gatesets change in the tests too; an incorrect translation will lead to mismatch between the Aer and Braket simulations.
Alright, approving but feel free to give some thoughts about https://github.com/qiskit-community/qiskit-braket-provider/pull/141#discussion_r1481786603. As for the tests, you are right. I just feel a bit uncomfortable that the transpiler mixes everything and our tests become less specific.
Summary
Right now, the adapter transpiles circuits to a target gateset of all gates supported by Braket, which can cause problems if the device itself doesn't support a gate. This PR changes the adapter to transpile to the device's target gateset.
Details and comments
For devices that support control modifiers, named controlled gates that are not already in Braket will be translated to the corresponding uncontrolled gate with control modifiers applied.