Speed up commutation-aware rewrite searches by taalexander · Pull Request #5298 · NVIDIA/cuda-quantum

taalexander · 2026-08-27T01:17:52Z

Fixes #5268 and #5269. QuakeSimplify currently scans every earlier operation when looking for a gate to combine. For example,

  S(q0)          possible match
  H(q0)          barrier
  H/X(q1) ...    unrelated gap
  S(q0)          search starts here

Previously, the search checked every q1 gate before reaching the H(q0) that stops it (as wires cannot always be followed due to a bunch of other issues noted below). Repeating this across a growing block can produce quadratic scan work.

This change builds a lazy interaction index for each known virtual qubit to operation. This is done on segment boundaries and when I say "virtual" qubit I mean a qubit that is statically knowable as equivalent to itself in the wire-flow chain:

  q0: S, H, S
  q1: H, X, ...

The q0 search now skips the q1 stream and reaches the same H(q0) immediately. Unsupported operations form segment boundaries, while unresolved qubit identities retain the existing block-order search.

All of the other commutation machinery has been kept the same.

Performance:

Reduce performance cost of failed searches in CommutationAwareRewriteMatcher #5268 at 2,048 placements: 1.91 s → 0.038 s
Reduce repeated QuakeSimplify scans across disjoint quantum support #5269 search work no longer grows with the size

I tried to keep it as simple as possible but it was still more LOC than I would have liked. A large portion of the difficulty with all of this work is we really can't "just" follow value form as there are loads of special cases of aliasing, reference form, unknown qubit values, etc., making the wire semantics non-trivial and causing the analysis "segment" within a block. Perhaps the right fix would be to simplify the IR/make the value form more strict, however, I don't think this is possible right now.

To review I recommend looking at it in this order:

find_nearest: unchanged matching rules and the handoff to indexed traversal.
LogicalQubitInteractionIndex: operation classification, segments, and ordered streams.

Build a lazy block-local interaction index keyed by proven logical qubit identity. Segmented streams let find_nearest skip unrelated quantum operations while preserving the existing SSA frontier, endpoint predicates, conservative barriers, and physical fallback for unresolved anchors. Keep index state private to CommutationAnalysis and maintain or discard it through the existing rewrite listener. Reuse normalized anchor support after construction and avoid indexing work on adjacent or otherwise inexpensive searches. Verified with the focused CommutationAwareRewrite unit suite, QuakeSimplify commutation lit coverage, the verifier-enabled linear-value pipeline, public issue reproducers, FTQC hotspots, and NISQ regression guards. Signed-off-by: Thomas Alexander <talexander@nvidia.com>

github-actions · 2026-08-27T01:57:15Z

CI Summary (`push`) — ✅ passed

Run #33127562649 · ✅ 7 · ⏩ 7 · ❌ 0 · ⛔ 0

Top-level jobs (14)

Job	Result
`binaries`	⏩ skipped
`build_and_test`	✅ success
`changes`	✅ success
`config_devdeps`	✅ success
`config_source_build`	⏩ skipped
`config_wheeldeps`	✅ success
`devdeps`	✅ success
`docker_image`	⏩ skipped
`gen_code_coverage`	⏩ skipped
`metadata`	✅ success
`python_metapackages`	⏩ skipped
`python_wheels`	⏩ skipped
`source_build`	⏩ skipped
`wheeldeps`	✅ success

⏩ Skipped jobs (7) — intentionally skipped on PR builds; run on merge_group / workflow_dispatch

Job
`binaries`
`config_source_build`
`docker_image`
`gen_code_coverage`
`python_metapackages`
`python_wheels`
`source_build`

All sub-jobs (43) — every matrix leg, with links

Job	Status	Link
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python)	✅ success	view
Build and test (amd64, llvm, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (amd64, llvm, openmpi) / Dev environment (Python)	✅ success	view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (arm64, llvm, openmpi) / Dev environment (Python)	✅ success	view
CI Summary	❔ in_progress	view
Check for stable CUDA-Q changes	✅ success	view
Configure build (devdeps)	✅ success	view
Configure build (source_build)	⏩ skipped	view
Configure build (wheeldeps)	✅ success	view
Create CUDA Quantum installer	⏩ skipped	view
Create Docker images	⏩ skipped	view
Create Python metapackages	⏩ skipped	view
Create Python wheels	⏩ skipped	view
Gen code coverage	⏩ skipped	view
Load dependencies (amd64, gcc12) / Caching	✅ success	view
Load dependencies (amd64, gcc12) / Finalize	✅ success	view
Load dependencies (amd64, gcc12) / Metadata	✅ success	view
Load dependencies (amd64, llvm) / Caching	✅ success	view
Load dependencies (amd64, llvm) / Finalize	✅ success	view
Load dependencies (amd64, llvm) / Metadata	✅ success	view
Load dependencies (arm64, gcc12) / Caching	✅ success	view
Load dependencies (arm64, gcc12) / Finalize	✅ success	view
Load dependencies (arm64, gcc12) / Metadata	✅ success	view
Load dependencies (arm64, llvm) / Caching	✅ success	view
Load dependencies (arm64, llvm) / Finalize	✅ success	view
Load dependencies (arm64, llvm) / Metadata	✅ success	view
Load source build cache	⏩ skipped	view
Load wheel dependencies (amd64, 12.6) / Caching	✅ success	view
Load wheel dependencies (amd64, 12.6) / Finalize	✅ success	view
Load wheel dependencies (amd64, 12.6) / Metadata	✅ success	view
Load wheel dependencies (amd64, 13.0) / Caching	✅ success	view
Load wheel dependencies (amd64, 13.0) / Finalize	✅ success	view
Load wheel dependencies (amd64, 13.0) / Metadata	✅ success	view
Load wheel dependencies (arm64, 12.6) / Caching	✅ success	view
Load wheel dependencies (arm64, 12.6) / Finalize	✅ success	view
Load wheel dependencies (arm64, 12.6) / Metadata	✅ success	view
Load wheel dependencies (arm64, 13.0) / Caching	✅ success	view
Load wheel dependencies (arm64, 13.0) / Finalize	✅ success	view
Load wheel dependencies (arm64, 13.0) / Metadata	✅ success	view
Prepare cache clean-up	❔ in_progress	view
Retrieve PR info	✅ success	view

✅ Required checks (6/6) — declared in .github/required-checks.yml for push

Required check	Status	Link
Build and test (amd64, llvm, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (amd64, llvm, openmpi) / Dev environment (Python)	✅ success	view
Build and test (arm64, llvm, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (arm64, llvm, openmpi) / Dev environment (Python)	✅ success	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Debug)	✅ success	view
Build and test (amd64, gcc12, openmpi) / Dev environment (Python)	✅ success	view

Run rotation legalization between two commutation-aware simplification phases. This preserves pre-legalization rotation combining and post-legalization cleanup without invalidating the interaction index after every expanding rewrite. Add focused coverage for cleanup between a legalized rotation sequence and an existing adjacent gate. Verified with the QuakeSimplify and commutation-aware focused suites on the combined FTQC candidate. Signed-off-by: Thomas Alexander <talexander@nvidia.com>

atgeller

I'd like some more clarification about the details of the interaction index, there were some confusing/unexpected things that I saw.

atgeller · 2026-08-27T18:54:04Z

+}
+
+// Build one ordered interaction stream for each logical qubit. A search on `q0`
+// can then skip an `X(q1)` and visit only `S(q0), H(q0)`. Hard barriers split


Could you clarify here - what is a hard barrier? Is there such a thing as a "soft" barrier?

atgeller · 2026-08-27T18:59:37Z

+// takes the erased operation's position. The rewrite listener removes erased
+// pointers and rebuilds the index after insertions that cannot be placed
+// safely.
+class cudaq::quake::detail::LogicalQubitInteractionIndex {


I think it would be helpful to have an explanation of how the streams are used for the analysis, even if briefly, here.

atgeller · 2026-08-27T19:02:52Z

+  bool hasPendingInsertions() const { return !pendingInsertions.empty(); }
+
+  PriorInteractionWalkResult
+  walk(Operation *anchor, Operation *inclusiveUpperBound,


I'm lost by what walk is doing. Can you add a description/contract?

It seems to return conclusive in any case it can find the walk bounds, regardless of the visitor function which is not what I would expect.

atgeller · 2026-08-27T19:03:22Z

+  }
+
+  PriorInteractionWalkResult
+  walkSingleQubitStream(const Segment &segment, QubitId qubitId,


It seems this method always returns PriorInteractionWalkResult::conclusive?

atgeller · 2026-08-27T19:10:50Z

+                                  std::move(*replacementSupport)};
+    removeInteractions(operation, position->second);
+    positions.erase(position);
+    auto [inserted, didInsert] =


Nit: on 831 this is [position, inserted], but here it is [inserted, didInsert]. Inconsistent naming

Rename the index around the operations stored for each logical qubit and define segment boundaries as unconditional indexed-search stops. Replace the two-value walk result with a boolean completion contract, and document traversal order, fallback behavior, query-specific stops, and ordinal maintenance. Verified by building cudaq-opt and OptimizerUnitTests. All 41 optimizer unit tests and both focused lit tests pass. The issue reproducer output hashes remain unchanged. Signed-off-by: Thomas Alexander <talexander@nvidia.com>

atgeller

👍 thanks for the clarifications @taalexander!

taalexander requested a review from atgeller August 27, 2026 01:17

github-actions Bot added the core compiler label Aug 27, 2026

taalexander force-pushed the perf/general-commutation-interaction-index branch from 95790e1 to d4aace7 Compare August 27, 2026 01:22

taalexander force-pushed the perf/general-commutation-interaction-index branch from 1568ec9 to e80d6e7 Compare August 27, 2026 15:20

atgeller reviewed Aug 27, 2026

View reviewed changes

taalexander requested a review from atgeller August 27, 2026 23:45

atgeller approved these changes Aug 28, 2026

View reviewed changes

Navigation Menu

Uh oh!

Speed up commutation-aware rewrite searches - #5298

Speed up commutation-aware rewrite searches#5298
taalexander wants to merge 3 commits into
NVIDIA:mainfrom
taalexander:perf/general-commutation-interaction-index

taalexander commented Aug 27, 2026 •

edited

Loading

Uh oh!

github-actions Bot commented Aug 27, 2026 •

edited

Loading

Uh oh!

atgeller left a comment

Uh oh!

atgeller Aug 27, 2026

Uh oh!

atgeller Aug 27, 2026

Uh oh!

atgeller Aug 27, 2026

Uh oh!

atgeller Aug 27, 2026

Uh oh!

atgeller Aug 27, 2026

Uh oh!

atgeller left a comment •

edited

Loading

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Conversation

taalexander commented Aug 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

github-actions Bot commented Aug 27, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

CI Summary (push) — ✅ passed

Uh oh!

atgeller left a comment

Choose a reason for hiding this comment

Uh oh!

atgeller Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

atgeller Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

atgeller Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

atgeller Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

atgeller Aug 27, 2026

Choose a reason for hiding this comment

Uh oh!

atgeller left a comment • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

taalexander commented Aug 27, 2026 •

edited

Loading

github-actions Bot commented Aug 27, 2026 •

edited

Loading

CI Summary (`push`) — ✅ passed

atgeller left a comment •

edited

Loading