Jake Lishman

Now rebased over #12410.

I've force-pushed a major new commit that properly refactored a bunch of the parallelism utilities to better support `should_run_in_parallel` at the level of robustness we'd expect from the public interface....

Given how major the changes I made here were, and that we're after the feature freeze deadline, I'm fine if we choose to leave this for Qiskit 2.0.

This is rebased over `main` now.

Should be resurrected now: I already rebased it a couple of weeks ago, so the last conflicts here were just super minor ones in a couple of test files that...

I'm not denying the overhead is too heavy, but I'm not certain that `Target` being in individual passes will be having any effect: assuming it's the exact same `Target` object...

I'd also be very interested in finding where the waste is in our transpiler-pass setup and working out a way to reduce it - I suspect that doing that would...

Hahaha, I posted that, then realised that I'd used Torino while you're using Sherbrooke. I just re-ran the same thing with Sherbrooke, and Sherbrooke's `Target` is 6 MB compared to...

Ok, so I'm really not very familiar with the pulse data structures at all, but there appear to be three major components to why this is so expensive: ### No...

Naoki and I returned briefly to this today for other reasons, and realised that `dill` serialisation, which the pass managers use, is about 10x slower than raw `pickle`, in large...