Adding in Parameters a flag for parrallel execution?

[lballerio]

Might it be useful to actually expose a flag in the experiments Parameters for executing in parallel or not a specific experiment?
Of course this is in the same fashion of the eventual refactor of qibocal protocols.

[alecandido]

Not sure, I feel like we should either do or not do, and do not add a flag. Otherwise, we would end up in the same situation of hardware_average and similar.

@RoyStegeman?

[RoyStegeman]

I would say that if a list of multiple targets is passed to the executor, it should try to execute the protocol for all targets simultaneously and raise an error if this is not possible (e.g. for 2q protocols where one qubit may be shared among to pairs). We can already add explicitly a for loop inside or outside the Executor so giving a third option through a flag (support for which then has to be propagated inside the protocol) seems overkill.

[lballerio]

Checking if parallel execution is possible is for sure something we have to implement and I was already thinking about that, but I also think this might be step1.
Honestly I think this is quite a sensible topic, especially when flux experiments comes into the game since there flux crosstalk is indeed an important factor;
of course for those cases we can do those checks, but I also think this parallel execution option should be a functionality that the user is able to control.

[RoyStegeman]

Maybe I misunderstood your original suggestion. I thought you meant a flag that if True runs runs the protocol in parallel on all targets (which should be the default, but probably is not consistently handled for all protocols) and if False run them in series (a for-loop).

This is independent of the handling of flux, for which I would say we should turn on the flux only for those qubits being used simultaneously. I.e if you do a single-qubit experiment, only one qubit should be fluxed, and if at a later stage of the calibration process you want to do a simultaneous calibration then all of them are fluxed and the flux-crosstalk matrix is used to set the fluxes. For the time being all fluxes are always activated (at least in qblox), which is something that has to be changed.