⬆️🐍 Update dependency mqt-core to v4

[renovate]

This PR contains the following updates:

Package	Change	Age	Confidence
mqt-core (changelog)	~=3.10.0 → ~=4.0.0

---

Release Notes

munich-quantum-toolkit/core (mqt-core)

v4.0.0

Compare Source

MQT Core 4: a compiler foundation built on MLIR and LLVM

MQT Core 4 is a major architectural release centered on the
MQT Compiler Collection, a quantum-classical compilation framework built on
MLIR and LLVM. It replaces the classic circuit representation with a
compiler infrastructure that connects program import, optimization, hardware
mapping, program exchange, and execution through C++, Python, and mqt-cc.

The new foundation makes quantum operations and classical computation part of
the same structured program. Quantum dialects work with MLIR's classical
arithmetic, functions, and control flow, so the compiler can represent loops,
reusable functions, and measurement feedback throughout supported pipelines.

The central changes are:

• A new program model. QC provides reference-based quantum operations for
  frontend interoperability; QCO uses linear quantum values for transformations.
  QTensor and CBit represent quantum and classical registers. Typed programs,
  builders, and verifiers expose this structure to compiler users and
  developers.
• A shared, extensible compilation pipeline. Import OpenQASM or Qiskit,
  inspect intermediate representations, compose optimization passes, synthesize
  gates, and place and route programs for a device's operations and topology.
  Python, C++, and the command-line driver use the same compiler infrastructure.
• Compilation connected to execution. Exchange structured programs through
  jeff or emit OpenQASM and QIR. QIR lowering produces LLVM text or bitcode for
  Base and Adaptive profiles; DDSIM uses LLVM JIT compilation and the DD runtime
  to execute supported QIR through QDMI. QCO also supports direct DD simulation,
  sampling, and unitary construction.
• Workflows built around the compiler. Typed benchmarks, SDK integrations,
  device-directed compilation, and six executable tutorials connect program
  construction to results. Python wheels include the compiler and local
  simulator; source builds enable the LLVM/MLIR compiler by default.

Upgrading from v3 requires an explicit migration. Classic circuit APIs,
including QuantumComputation, are removed. The low-level DD and QDMI libraries
remain available. Start with the v3-to-v4 upgrade guide for
the new program model, migration paths, API replacements, and build changes
relative to v3.10.0. The entries below retain the contributing PRs and authors.

Added

Compiler APIs and representations

• ✨ Introduce the MLIR/LLVM-based MQT Compiler Collection with typed QC, QCO,
  OpenQASM, jeff, and QIR programs; shared C++ and Python compilation APIs;
  inspection methods; and the mqt-cc driver. ([#​1264], [#​1470], [#​1471],
  [#​1815], [#​1914], [#​2149], [#​2343], [#​2519]) ([@​burgholzer],
  [@​denialhaag], [@​simon1hofmann], [@​taminob])

• ✨ Add QC/QCO program builders with implicit locations, tracked qubits and
  tensors, register names, gate matrices, and checked linear quantum values.
  Preserve caller-owned inputs and reject invalid programs at checked
  boundaries. ([#​1402], [#​1428], [#​1430], [#​1443], [#​1446], [#​1465], [#​1472],
  [#​1474], [#​1510], [#​1542], [#​1580], [#​1602], [#​1620], [#​1623], [#​1626],
  [#​1627], [#​1728], [#​1730], [#​1781], [#​1869], [#​1872], [#​2014], [#​2017],
  [#​2136], [#​2213], [#​2220], [#​2266], [#​2295], [#​2296], [#​2300], [#​2514])
  ([@​burgholzer], [@​denialhaag], [@​li-mingbao],
  [@​MatthiasReumann], [@​simon1hofmann], [@​taminob])

• ✨ Represent classical registers and structured if, for, while, and
  switch control flow across QC and QCO, including break, continue, wide
  register arithmetic, and reusable quantum functions with tracked wire
  correspondence. ([#​1506], [#​1638], [#​1700], [#​1717], [#​1749], [#​1806],
  [#​1808], [#​1824], [#​1925], [#​1927], [#​1935], [#​1936], [#​1938], [#​1975],
  [#​1976], [#​2026], [#​2158], [#​2181], [#​2193], [#​2194], [#​2297], [#​2307],
  [#​2308], [#​2320], [#​2322], [#​2336], [#​2344], [#​2357], [#​2369], [#​2443],
  [#​2448], [#​2464]) ([@​burgholzer], [@​denialhaag], [@​DRovara],
  [@​li-mingbao], [@​MatthiasReumann], [@​simon1hofmann])

• ✨ Canonicalize quantum gates, modifiers, and static qubits while preserving
  phase, wire correspondence, and side effects. Bound angle arithmetic and
  reduce repeated traversal and validation work. ([#​1330], [#​1426], [#​1436],
  [#​1464], [#​1475], [#​1550], [#​1554], [#​1567], [#​1569], [#​1603], [#​1751],
  [#​1762], [#​1782], [#​1807], [#​1985], [#​2041], [#​2210], [#​2281], [#​2282],
  [#​2290], [#​2291], [#​2293], [#​2435], [#​2477], [#​2484], [#​2485], [#​2486],
  [#​2492], [#​2502], [#​2505], [#​2516], [#​2524]) ([@​burgholzer],
  [@​denialhaag], [@​DRovara], [@​Ectras], [@​J4MMlE],
  [@​simon1hofmann], [@​taminob])

Compilation for devices

• ✨ Compile for immutable device targets through C++, Python, and mqt-cc.
  Match ordered operation availability, native gates, and payload capabilities;
  legalize supported control flow; and reuse compiled payloads for explicit
  device submission. ([#​1687], [#​1993], [#​1999], [#​2049], [#​2162], [#​2211],
  [#​2215], [#​2218], [#​2219], [#​2285], [#​2323], [#​2495], [#​2497], [#​2506],
  [#​2510]) ([@​burgholzer], [@​denialhaag], [@​MatthiasReumann],
  [@​simon1hofmann])

• ✨ Place and route QCO programs across sparse and directional device
  topologies, preserving classical feed-forward and quantum wire mappings
  through branches, loops, and switches. ([#​1537], [#​1547], [#​1581], [#​1583],
  [#​1588], [#​1600], [#​1664], [#​1709], [#​1716], [#​1748], [#​1805], [#​1870],
  [#​1904], [#​1911], [#​1951], [#​1956], [#​1997], [#​2004], [#​2007], [#​2016],
  [#​2060], [#​2119], [#​2179], [#​2184], [#​2185], [#​2205], [#​2240], [#​2301],
  [#​2316], [#​2351], [#​2436], [#​2489]) ([@​burgholzer], [@​denialhaag],
  [@​MatthiasReumann], [@​rturrado], [@​simon1hofmann])

• ✨ Evaluate multiple initial layouts, including identity and greedy
  candidates, with CPU-based default trial counts. Reuse routing search storage
  and avoid repeated program-graph and liveness analysis. ([#​1568], [#​1574],
  [#​1930], [#​2180], [#​2488], [#​2499], [#​2500], [#​2517], [#​2535])
  ([@​burgholzer], [@​MatthiasReumann], [@​simon1hofmann])

Optimization and synthesis

• ✨ Optimize single-qubit runs with Hadamard lifting, quaternion-based rotation
  merging, Euler synthesis, and fusion of constant or parameterized unitaries.
  ([#​1407], [#​1605], [#​1672], [#​1674], [#​2002], [#​2038], [#​2228], [#​2478])
  ([@​burgholzer], [@​denialhaag], [@​J4MMlE], [@​lirem101],
  [@​MatthiasReumann], [@​simon1hofmann], [@​taminob])

• ✨ Decompose multi-controlled X, Z, phase, Pauli rotations, and SWAP
  operations. Fuse two-qubit unitaries and synthesize target-native entanglers,
  including RXX, RYY, RZX, RZZ, iSWAP, ECR, and optimal square-root iSWAP
  circuits. ([#​1774], [#​1802], [#​1803], [#​1809], [#​1810], [#​1814], [#​1832],
  [#​1850], [#​1865], [#​1961], [#​1996], [#​1998], [#​2001], [#​2444], [#​2467],
  [#​2468], [#​2478], [#​2507], [#​2531]) ([@​burgholzer], [@​denialhaag],
  [@​simon1hofmann])

• ✨ Normalize global phases and expand multi-operation quantum modifiers while
  preserving full-unitary semantics. ([#​1986], [#​1995], [#​2006], [#​2015],
  [#​2189]) ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

• ✨ Add measurement lifting, classical-control replacement, explicit dead-gate
  elimination, quantum-loop unrolling, qubit reuse, and Pauli twirling.
  ([#​1705], [#​1718], [#​1755], [#​1756], [#​1923], [#​1924], [#​2039], [#​2118],
  [#​2216], [#​2224]) ([@​burgholzer], [@​denialhaag], [@​DRovara],
  [@​MatthiasReumann], [@​simon1hofmann])

Program import and exchange

• ✨ Import OpenQASM 2 and 3 into QC and emit OpenQASM 3 with reusable gates,
  structured control flow, fixed-width angles, affine qubit indices, and
  dynamically bounded loops. Diagnose unsupported inputs and bound frontend and
  export resource use. ([#​1780], [#​1910], [#​1987], [#​1994], [#​2003], [#​2050],
  [#​2169], [#​2203], [#​2304], [#​2338], [#​2456], [#​2465], [#​2482], [#​2483],
  [#​2515]) ([@​burgholzer], [@​denialhaag], [@​J4MMlE],
  [@​simon1hofmann])

• ✨ Import and export Qiskit circuits, reusable custom gates, captured
  classical expressions, and structured control flow. Preserve measurements,
  wide register comparisons, and target-aware gate definitions. ([#​2031],
  [#​2133], [#​2140], [#​2175], [#​2176], [#​2223], [#​2342], [#​2348], [#​2398],
  [#​2439], [#​2452], [#​2461], [#​2508]) ([@​burgholzer], [@​denialhaag],
  [@​MatthiasReumann], [@​simon1hofmann])

• ✨ Preserve symbolic Qiskit parameters and parameter-vector provenance through
  supported compiler conversions. ([#​2150], [#​2178]) ([@​burgholzer],
  [@​simon1hofmann])

• ✨ Exchange jeff programs as bytes or files, preserving supported gates,
  phase, scalar control-flow state, arrays, and reusable functions. Use native
  integer arithmetic and report recoverable deserialization failures. ([#​1479],
  [#​1548], [#​1565], [#​1637], [#​1676], [#​1706], [#​1776], [#​1836], [#​1934],
  [#​1939], [#​2000], [#​2018], [#​2105], [#​2212], [#​2339], [#​2451], [#​2457],
  [#​2525]) ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

• ✨ Generate QIR 2.1 Base and Adaptive profiles as LLVM text or bitcode. Derive
  capability and resource metadata, lower reusable functions and classical
  computation, preserve output order, and reject unsupported Base control flow
  during conversion. ([#​1446], [#​1513], [#​1521], [#​1548], [#​1567], [#​1569],
  [#​1570], [#​1572], [#​1580], [#​1620], [#​1624], [#​1626], [#​1648], [#​1710],
  [#​1751], [#​1787], [#​1815], [#​1823], [#​1933], [#​1978], [#​1979], [#​2026],
  [#​2030], [#​2066], [#​2217], [#​2294], [#​2302], [#​2340], [#​2446], [#​2447],
  [#​2449], [#​2463], [#​2526]) ([@​burgholzer], [@​denialhaag],
  [@​li-mingbao], [@​MatthiasReumann], [@​simon1hofmann])

Decision diagrams and execution

• ✨ Construct functionality, simulate, and sample QCO programs through C++ and
  Python, with dense-array helpers for supported compiler inputs. Support
  classical arithmetic, structured control flow, dynamic quantum data, and
  statically sized entry-block registers for unitary construction. ([#​1915],
  [#​1973], [#​2077], [#​2078], [#​2079], [#​2334], [#​2455], [#​2474], [#​2518],
  [#​2526]) ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

• ✨ Return ordered DDSIM QIR shots and matching histograms, including
  variable-length recorded outputs. ([#​2368]) ([@​burgholzer])

• ✨ Extract statevectors from eligible Adaptive QIR with classical control
  flow, dynamic allocations, and direct helpers. Retain uncollapsed states after
  eligible OpenQASM and QIR terminal sampling, with lazy dense and sparse result
  queries. ([#​2491], [#​2494], [#​2512]) ([@​burgholzer])

• ✨ Capture framed, typed QIR output records through DDSIM with custom2=True
  and result CUSTOM1, alongside ordinary shots and counts. ([#​2526])
  ([@​burgholzer], [@​simon1hofmann])

Structured benchmarks

• ✨ Add typed quantum benchmarks with versioned instance specifications,
  analytic references, deterministic manifests, and C++, Python, and CLI
  interfaces. Families cover BV, GHZ, Grover, QFT, QPE, multiplexers,
  teleportation, QFT adders, modular multiplication, and repeat-until-success
  programs. ([#​2135], [#​2299], [#​2315], [#​2324], [#​2337], [#​2380], [#​2402],
  [#​2404], [#​2409], [#​2410], [#​2493]) ([@​burgholzer], [@​denialhaag])

Documentation and development

• 📝 Connect compilation, execution, device access, and program exchange in the
  user guide. Add six executable compiler tutorials and checked Python/C++
  examples with downloadable notebooks. ([#​1555], [#​1635], [#​1773], [#​1899],
  [#​1959], [#​2058], [#​2165], [#​2462], [#​2503], [#​2509], [#​2526])
  ([@​burgholzer], [@​denialhaag], [@​MatthiasReumann],
  [@​simon1hofmann], [@​ystade])

• 📝 Consolidate development policy and agent guidance, including compiler
  contracts, terminology, validation, and reproducible benchmarks. ([#​1905],
  [#​1907], [#​1908], [#​1909], [#​1960], [#​2072], [#​2120], [#​2128], [#​2256],
  [#​2272], [#​2354], [#​2490], [#​2496], [#​2498], [#​2501], [#​2523], [#​2526])
  ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

• 🐳 Add a dev container for local development. ([#​1786]) ([@​denialhaag])

Changed

Builds and supported configurations

• 🔧 Build the LLVM/MLIR 23.1.1 compiler infrastructure and DDSIM device by
  default. Support LLVM builds without exceptions or RTTI, and use
  BUILD_MQT_CORE_MLIR=OFF for DD/QDMI builds that omit the compiler and DDSIM.
  ([#​1356], [#​1549], [#​1953], [#​2125], [#​2127], [#​2284], [#​2298], [#​2538])
  ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

• 💥 Require CMake 3.28 or newer for source builds and embedded projects, with
  native dependency exclusions and system include handling. ([#​2421])
  ([@​burgholzer])

• ⚡ Reduce build, documentation, and test-discovery overhead. Share Ninja
  presets and generate C++ lint prerequisites without a full build while
  honoring disabled interprocedural optimization. ([#​1944], [#​1954], [#​1988],
  [#​2047], [#​2075], [#​2083], [#​2426], [#​2459], [#​2470]) ([@​burgholzer],
  [@​denialhaag])

• 📦 Use vcs-versioning and standard dynamic metadata for SCM-derived Python
  package versions, with lower bounds for build dependencies. ([#​1544], [#​2065],
  [#​2145], [#​2163]) ([@​burgholzer], [@​denialhaag])

• ⬆️ Update clang-tidy to version 23 and adapt the C++ sources to its
  diagnostics. Format TableGen files and align LLVM include and namespace
  conventions. ([#​1573], [#​1673], [#​1675], [#​1765], [#​2028], [#​2328])
  ([@​burgholzer], [@​denialhaag], [@​simon1hofmann])

Runtime and client behavior

• 💥 Align QIR execution with QIR 2.1 runtime and QIS signatures, isolated
  per-job state, deterministic seeded sampling, and checked state extraction.
  Optimize eligible terminal sampling and output preparation; verify Base output
  records against QIR-Runner. ([#​2034], [#​2035], [#​2036], [#​2044], [#​2466],
  [#​2513]) ([@​burgholzer])

• ⚡ Reduce QDMI discovery, result-decoding, and Slurm overhead. Release the
  Python GIL during native QDMI calls and avoid repeated capability and result
  queries. ([#​2440], [#​2460], [#​2472], [#​2475], [#​2481], [#​2511])
  ([@​burgholzer])

• ⚡ Reduce DD cache clearing, element lookup, export, and allocation overhead,
  and avoid repeated QC/QCO conversion work. ([#​2441], [#​2453], [#​2473],
  [#​2474], [#​2479]) ([@​burgholzer])

Fixed

Decision diagrams

• 🐛 Correct DD cache invalidation after garbage collection, compressed-matrix
  entries, partial traces, Kronecker products, numeric-table reuse, and logical
  Python vector indexing. Reject malformed paths and ragged dense matrices.
  ([#​2441], [#​2453], [#​2479]) ([@​burgholzer])

• 🐛 Use deterministic, collision-free DOT node IDs, define signed
  complex-weight hashing, and preserve real-number collection flags during
  relinking. ([#​2518]) ([@​burgholzer])

Device access

• 🐛 Preserve QDMI provider lifetimes and target metadata; validate job
  parameters and decoded results; and propagate execution failures through SDK
  integrations. ([#​2440], [#​2454], [#​2458], [#​2511], [#​2512])
  ([@​burgholzer])

Removed

• 💥 Remove qc::QuantumComputation, MQT::CoreIR, MQT::CoreQASM, classic
  Python circuit APIs, and circuit-taking DD helpers. Use compiler-backed QC/QCO
  APIs or remain on the v3 release series. ([#​2054], [#​2288])
  ([@​burgholzer], [@​simon1hofmann])

• 💥 Remove the standalone QIR runner and make the QIR runtime and JIT internal
  DDSIM implementation details. Execute QIR through QDMI. ([#​2246])
  ([@​burgholzer], [@​denialhaag])

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • "every weekend"
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR becomes conflicted, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about these updates again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.

[renovate]

⚠️ Artifact update problem

Renovate failed to update an artifact related to this branch. You probably do not want to merge this PR as-is.

♻ Renovate will retry this branch, including artifacts, only when one of the following happens:

• any of the package files in this branch needs updating, or
• the branch becomes conflicted, or
• you click the rebase/retry checkbox if found above, or
• you rename this PR's title to start with "rebase!" to trigger it manually

The artifact failure details are included below:

File name: uv.lock

Command failed: uv lock --upgrade-package mqt-core --upgrade-package mqt-core
Using CPython 3.14.7 interpreter at: /opt/containerbase/tools/python/3.14.7/bin/python3
  × No solution found when resolving dependencies for split (markers:
  │ python_full_version >= '3.15' and sys_platform == 'win32'):
  ╰─▶ Because mqt-qcec>=3.10.0 depends on mqt-core>=3.10.0,<3.11.dev0
      and mqt-qmap:dev depends on mqt-qcec>=3.10, we can conclude that
      mqt-qmap:dev depends on mqt-core>=3.10.0,<3.11.dev0.
      And because mqt-qmap:dev depends on mqt-core>=4.0.0 and your project
      requires mqt-qmap:dev, we can conclude that your project's requirements
      are unsatisfiable.

hint: While the active Python version is 3.14, the resolution failed for other Python versions supported by your project. Consider limiting your project's supported Python versions using `requires-python`.

[denialhaag]

MQT QMAP will not be compatible with MQT Core 4 for the foreseeable future. Closing this.

[renovate]

Renovate Ignore Notification

Because you closed this PR without merging, Renovate will ignore this update. You will not get PRs for any future 4.x releases. But if you manually upgrade to 4.x then Renovate will re-enable minor and patch updates automatically.

If you accidentally closed this PR, or if you changed your mind: rename this PR to get a fresh replacement PR.