[simon1hofmann]

🤖 AI text below 🤖

Description

Fixes #2588. Stacked on #2565 (codex/bench-controlled-gates).

Target compilation and synthesis currently reject a controlled composite when an inverse or power modifier wraps its body. Reuse the existing inverse and power-unrolling helpers inside eligible controls, then simplify the resulting modifiers in the existing multi-control decomposition pass.

The shared unrolling helpers require verified linear QCO IR. This avoids rebuilding wire identities, copying inverse bodies, and maintaining a separate control wrapper. Inverse bodies retain reverse gate order. Constant integer powers are distributed only when body operations act on disjoint wires. Width thresholds and native-target support remain respected. No new pipeline pass, early rotation merging, dependencies, or frontend-specific handling are added.

Usage and limitations

The existing compile_for_target(...) and synthesize_for_target(...) APIs now accept supported nested forms such as ctrl @ inv @ composite and integer powers of disjoint composite bodies.

Fractional powers, runtime exponents, and powers of overlapping body operations are not distributed. Such inputs still need an existing simplification or native synthesis path; this does not add arbitrary large-unitary synthesis or repeated expansion of circuit bodies.

Validation

• Rebased onto upstream main c92ffa7a6 and reviewed together with the stacked PR.
• Combined-stack validation passes: 1,524 native tests (compiler 233, decomposition 315, MQT transforms 31, QC IR 368, QCO IR 575, and two native-synthesis checks) and all 428 Qiskit translation tests.
• Consolidated pipeline tests compare full unitaries for plain, inverse, and positive/negative integer-power bodies through both target APIs. Retained boundary cases cover width, native support, control scope, unsupported powers, and failed unrolling.
• Removed 203 net lines of test code across the stack: MCMT cases reduced from 36 to 6 and unsupported-power combinations from 7 to 3. No production code changed in this test/documentation cleanup.
• Before consolidation, twelve additional Qiskit-to-target exact-unitary checks passed for nested modifiers and open MCMT inverses.
• Clang 23/ThinLTO native build, repository lint, and full changed-file C++ lint pass. Python validation uses Qiskit 2.5.2.

Hosted checks for the updated heads are pending.

Codex assisted with implementation, tests, documentation, and this description.

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

✅ All modified and coverable lines are covered by tests.

📢 Thoughts on this report? Let us know!

[burgholzer]

This one also looks good! Just trimmed down tests and docs again. Let's get this in. Two more PRs merged, one more issue closed