Naive crosstalk #175
Naive crosstalk #175
Conversation
| // outcomes to the OutputEngine, which should be ignored. How do we do this? | ||
| // Perhaps there should be another add_collapse() method that applies a | ||
| // measurement and discards the outcome. | ||
| builder.add_measurements(&affected_qubits); |
There was a problem hiding this comment.
Comment from Ciaran on the commit:
Hmmmm, let's see. Measurements end up kinda of being two rounds... First the noise model sends a measurement request to the simulator and then noise model gets a result from the simulator. Then then noise model can intercept/forward/etc. results to the "classical engine". This is done here:
PECOS/crates/pecos-engines/src/noise/general.rs
Line 362 in 55994b8
fn continue_processing(I kinda of had to deal with a similar situation of distinguishing measurement results for leakage... and I ended up just storing the measurement type here:
PECOS/crates/pecos-engines/src/noise/general.rs
Line 332 in 55994b8
measured_qubits: Vec<(usize, bool)>,But I feel like what needs to be done is to add some tagging mechanism in ByteMessage
I might... try adding result tagging to ByteMessage...
There was a problem hiding this comment.
Ok, I think I understand. Without tagging in ByteMessage, what I could do is add an extra bool to measured_qubits, so that's Vec<(usize, bool, bool)> where the second bool corresponds to is_crosstalk to indicate its origin.
For the naive crosstalk, I'd just throw away the measurement outcome at apply_noise_on_continue_processing here whenever is_crosstalk is True. In future versions, it could be used to make a decision on how to apply crosstalk.
That makes sense, I can do it tomorrow morning. I guess there's no need for tags for the first version, but if you get them in, I'll use them.
There was a problem hiding this comment.
Yeah, sounds good. Later I can add a tag system and replace that.
| p2_idle: 0.0, | ||
| leaked_qubits: HashSet::new(), | ||
| rng: NoiseRng::default(), | ||
| initialized_qubits: BTreeSet::new(), |
There was a problem hiding this comment.
Comment from Ciaran in the commit:
dcc5fb7#commitcomment-164014201
Would prefer initialized -> prep or prepared or something
Note: crosstalk for measurement and prep/reset/init can look differently. But can change that later.
There was a problem hiding this comment.
Sure, happy to change that.
Creating a PR just so that it's easier to track conversations.