Make VF2Layout and VF2PostLayout work with control flow #8804
Conversation
The subgraph isomorphism problem we are solving here can fairly easily be extended to account for the available types of control flow in Qiskit right now. We just treat each control flow operation as "transparent" in the pass, i.e. as if each of its blocks were simply inserted verbatim (with the requisite wire mapping) sequentially. We use the same weighting heuristics for control-flow loop bodies in the scoring as elsewhere; for-loops count for the number of iterations, while if-else and while loops count (all of) their bodies once. The testing in this particular commit is slightly light, because some of the general strategies involve a `transpile` call, which isn't fully ready yet with control flow.
|
Thank you for opening a new pull request. Before your PR can be merged it will first need to pass continuous integration tests and be reviewed. Sometimes the review process can be slow, so please be patient. While you're waiting, please feel free to review other open PRs. While only a subset of people are authorized to approve pull requests for merging, everyone is encouraged to review open pull requests. Doing reviews helps reduce the burden on the core team and helps make the project's code better for everyone. One or more of the the following people are requested to review this:
|
Pull Request Test Coverage Report for Build 3154181627
💛 - Coveralls |
There was a problem hiding this comment.
I think this LGTM, it is fairly straightforward change to make the interaction graph built with recursing into the control flow blocks. I think we can expand the testing after the fact when we have a full pipeline working. The other thing I think we'll want tests for on VF2PostLayout is the case where strict_direction=True since that is one place where VF2PostLayout and VF2Layout differ in behavior.
jakelishman
added
the
Changelog: New Feature
Summary
The subgraph isomorphism problem we are solving here can fairly easily be extended to account for the available types of control flow in Qiskit right now. We just treat each control flow operation as "transparent" in the pass, i.e. as if each of its blocks were simply inserted verbatim (with the requisite wire mapping) sequentially.
We use the same weighting heuristics for control-flow loop bodies in the scoring as elsewhere; for-loops count for the number of iterations, while if-else and while loops count (all of) their bodies once.
The testing in this particular commit is slightly light, because some of the general strategies involve a
transpilecall, which isn't fully ready yet with control flow.Details and comments
Getting VF2 working with control flow was very easy - this sort of "block transparency" lends itself very well to recursive closures - but filling out the tests, especially of
VF2PostLayoutwas more challenging. I'm somewhat open to more things we might want to do, given that right now, we can't yet rely on being able to route a circuit with control flow, which stymies efforts to usetranspile. Hopefully #8418 will make this available, however.Adding this simplifies the work needed to get the control-flow modifications to the preset pass managers working.