Julien Gacon
Julien Gacon
Update: this should get fixed by #12384.
Nice! Did you already get to benchmark the basis translator with the oxidized equivalence library? 🙂
@mtreinish's update LGTM 👍🏻
https://github.com/Qiskit/qiskit/pull/12825#pullrequestreview-2404327735 is a very good catch -- the newly added gates that implement the DD sequences didn't have a `duration` attached, since they weren't present when the scheduling pass was...
After discussion with @ElePT I also updated the default behavior of `min_joinable_duration`: On realistic devices, the 2q gate lengths vary a little -- yet might not want to insert DD...
I removed any setting and usage of `Instruction.duration` in 21c4ea0, so hopefully you don't have to update the context-aware DD in #13506 🙂 That means the timeline drawer will fail...
Yep, it's good from my side now 🙂 The number of X gates is rather large though, per design of the algorithm, I think that was just surprising to see....
Re: `min_joinable_duration` -- this shouldn't set the amount of DD inserted, but rather control how adjacent idle blocks are joined. It's a knob that lets you trade-off between potentially many...
Fair point, let's rename it 👍🏻
Nice catch, that seems to be unique only to the Pauli-I operator. Would you like to open a PR to fix the issue? 🙂