Fix tracking of routing permutation in Sabre with disjoint backends #13833
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
If the backing coupling graph is disjoint, and unused components of the coupling graph would not be considered when constructing the complete routing permutation. In practice, Sabre aborts immediately after layout without attempting to route, if it needed to split the DAG across more than one disjoint component, because it can't guarantee correctness of the final routing in the presence of component-spanning barriers or classical communication, so the only way for a component to be forgotten is if the backend is disjoint, but the DAG fits into a single component.
jakelishman
added
stable backport potential
|
One or more of the following people are relevant to this code:
|
ElePT
approved these changes
Feb 13, 2025
There was a problem hiding this comment.
LGTM!! Thanks @jakelishman for diving right into the rabbit hole!
| qc.cx(2, 0) | ||
|
|
||
| disjoint = CouplingMap([(0, 1), (1, 2), (3, 4), (4, 5)]) | ||
| layout_routing_pass = SabreLayout(disjoint, seed=2025_02_12, swap_trials=1, layout_trials=1) |
github-merge-queue bot
pushed a commit
that referenced
this pull request
Feb 13, 2025
…13833) If the backing coupling graph is disjoint, and unused components of the coupling graph would not be considered when constructing the complete routing permutation. In practice, Sabre aborts immediately after layout without attempting to route, if it needed to split the DAG across more than one disjoint component, because it can't guarantee correctness of the final routing in the presence of component-spanning barriers or classical communication, so the only way for a component to be forgotten is if the backend is disjoint, but the DAG fits into a single component.
|
Hmm I removed the PR from the queue because I saw the failed neko test in the merge CI run, but I don't think it's related to this PR. I think it's related to changes in qiskit-aer. |
mergify bot
pushed a commit
that referenced
this pull request
Feb 13, 2025
…13833) If the backing coupling graph is disjoint, and unused components of the coupling graph would not be considered when constructing the complete routing permutation. In practice, Sabre aborts immediately after layout without attempting to route, if it needed to split the DAG across more than one disjoint component, because it can't guarantee correctness of the final routing in the presence of component-spanning barriers or classical communication, so the only way for a component to be forgotten is if the backend is disjoint, but the DAG fits into a single component. (cherry picked from commit b933179)
github-merge-queue bot
pushed a commit
that referenced
this pull request
Feb 13, 2025
…13833) (#13835) If the backing coupling graph is disjoint, and unused components of the coupling graph would not be considered when constructing the complete routing permutation. In practice, Sabre aborts immediately after layout without attempting to route, if it needed to split the DAG across more than one disjoint component, because it can't guarantee correctness of the final routing in the presence of component-spanning barriers or classical communication, so the only way for a component to be forgotten is if the backend is disjoint, but the DAG fits into a single component. (cherry picked from commit b933179) Co-authored-by: Jake Lishman <[email protected]>
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Summary
If the backing coupling graph is disjoint, and unused components of the coupling graph would not be considered when constructing the complete routing permutation. In practice, Sabre aborts immediately after layout without attempting to route, if it needed to split the DAG across more than one disjoint component, because it can't guarantee correctness of the final routing in the presence of component-spanning barriers or classical communication, so the only way for a component to be forgotten is if the backend is disjoint, but the DAG fits into a single component.
Details and comments
Fix #13732, though there's a whole cornucopia of bugs split between very old versions of Qiskit, this commit's parent, and the example in the issue itself that lead from that issue seeing a regression in Qiskit 1.3 to this one.
In actuality, this bug has been around since the disjoint-coupling handling was first added in #9802 (qiskit-terra 0.24), and the test case in this PR would fail that version too (the fact that
TranspileLayout.initial_index_layoutwasn't added til later notwithstanding - the first assertion fails).