🐛 Preserve symbolic Euler chains in target synthesis

[simon1hofmann]

🤖 AI text below 🤖

Description

Compile supported symbolic Euler chains directly into the target basis so that EfficientSU2-style circuits remain bindable after Qiskit export. H/RZ pairs in either order now use one U gate or at most three gates in the other supported bases, without runtime quaternion extraction. This restores the RX/RZ, RX/RY, and R gate-count bounds and fixes the U/H–RZ export failure.

• Reuse direct ZXZ/ZYZ and XZX/XYX identities and the existing Euler emitters. Normalize each evaluated gate operand modulo 4*pi before offsets or phase arithmetic, preserving shared scalar expressions and controlled phase.
• Keep target-aware fusion inside native synthesis: preserve native symbolic runs, require improvement for constant native runs, and leave controlled bodies with their lowering owner. Restore canonicalized controlled U2 to native U when supported.
• Retain the U/unrestricted optimizer until the shared synthesis contracts cover its controlled-body and singleton behavior. Unsupported symbolic shapes retain individual native lowering; this does not make arbitrary runtime Euler expressions exportable.
• Keep numerical/basis coverage in C++ and reduce this PR's Python additions from 19 cases to seven binding/export cases. Move the 100-qubit symbolic workload to the follow-up benchmark specification.

No new dependencies or public API changes are required. Broader consolidation and the remaining Python test audit are tracked in follow-up issue #2614.

Validation

Linux AArch64, optimized release-clang-ipo build:

• 329 focused C++ tests passed: Euler/fusion, rotation merging, native synthesis, and compiler pipelines. Includes both H/RZ orders in all seven bases, phase-sensitive controlled matrices, bounded/shared operands, native preservation, and controlled-U compilation.
• All 97 tests in test/python/test_mlir.py passed with Qiskit and jeff available, including late parameter binding and the U/H–RZ exporter regression.
• Repository lint and whole-file C++ lint: passed with zero findings.

Hosted CI has not been checked for this update. The earlier large-workload timings are not claimed as validation of this revision.

OpenAI Codex assisted with implementation, tests, and this description. Human review of the updated patch remains required before merging.

Checklist

• The pull request only contains commits that are focused and relevant to this change.
• I have added appropriate tests that cover the new/changed functionality.
• I have updated the documentation to reflect these changes.
• The changes follow the project's style guidelines and introduce no new warnings.
• The changes are fully tested and pass the CI checks.
• I have reviewed my own code changes.

If PR contains AI-assisted content:

• Any agent that created, edited, or submitted GitHub content was explicitly authorized for that scope, as required by our AI Usage Guidelines.
• Every agent-authored or agent-edited public text body begins with the visible disclosure 🤖 *AI text below* 🤖 (titles are exempt).
• I have disclosed AI assistance in the PR description.
• I confirm that I have personally reviewed and understood all AI-generated content, and accept full responsibility for it.

[codecov]

Codecov Report

❌ Patch coverage is 96.77419% with 4 lines in your changes missing coverage. Please review.

Files with missing lines	Patch %	Lines
...QCO/Transforms/NativeSynthesis/TargetSynthesis.cpp	86.3%	3 Missing ⚠️
...ms/Optimizations/MergeSingleQubitRotationGates.cpp	98.8%	1 Missing ⚠️

📢 Thoughts on this report? Let us know!

[burgholzer]

Thanks @simon1hofmann 🙏🏼 finally got through this. I pushed a couple of fixes and simplifications and created a follow-up issue (#2614) with further simplification opportunities I identified but did not want to put into this PR in order to not hold it up for too long.
If you are also happy with the changes here, feel free to merge.
The follow-up issue is also up for grabs 🙌🏼

[simon1hofmann]

Thanks @simon1hofmann 🙏🏼 finally got through this. I pushed a couple of fixes and simplifications and created a follow-up issue (#2614) with further simplification opportunities I identified but did not want to put into this PR in order to not hold it up for too long. If you are also happy with the changes here, feel free to merge. The follow-up issue is also up for grabs 🙌🏼

Thanks a lot for the final touches, I can tackle the follow-up issue 👍