Prepare 2.5.0

Cargo.toml

@@ -3,7 +3,7 @@ members = ["crates/*"]
 resolver = "2"
 
 [workspace.package]
-version = "2.5.0-rc1"
+version = "2.5.0"
 edition = "2024"
 rust-version = "1.87"  # Keep in sync with README.md, rust-toolchain.toml, and tools/install_rust_msrv.sh
 license = "Apache-2.0"

crates/bindgen/include/qiskit/version.h

@@ -21,11 +21,11 @@
 #define QISKIT_VERSION_MAJOR 2
 #define QISKIT_VERSION_MINOR 5
 #define QISKIT_VERSION_PATCH 0
-#define QISKIT_RELEASE_LEVEL QISKIT_RELEASE_LEVEL_RC
+#define QISKIT_RELEASE_LEVEL QISKIT_RELEASE_LEVEL_FINAL
 // For the final release, set the below to 0.
-#define QISKIT_RELEASE_SERIAL 1
+#define QISKIT_RELEASE_SERIAL 0
 
-#define QISKIT_VERSION "2.5.0-rc1"
+#define QISKIT_VERSION "2.5.0"
 
 #define QISKIT_GET_VERSION_HEX(major, minor, patch, level, serial)                                 \
     (((major) & 0xff) << 24 | ((minor) & 0xff) << 16 | ((patch) & 0xff) << 8 |                     \

docs/release_notes.rst

@@ -21,4 +21,4 @@ Qiskit |version| release notes
       though, which is typically `rc1` tags.
 
 .. release-notes::
-   :earliest-version: 2.4.0rc1
+   :earliest-version: 2.5.0rc1

qiskit/VERSION.txt

@@ -1 +1 @@
-2.5.0.rc1
+2.5.0

releasenotes/notes/2.4/fix-composition-of-virtual-permutations-7f1d92e9c45dfd3a.yaml

@@ -1,5 +1,5 @@
 ---
 fixes:
   - |
-    Fixed a bug in :class:`ElidePermutations` and :class:`StarPreRouting` when composing
+    Fixed a bug in :class:`.ElidePermutations` and :class:`.StarPreRouting` when composing
     virtual permutation layouts (implicit permutations applied at the end of circuit).

releasenotes/notes/2.4/fix-controlled-rotations-commutationchecker-73b6e758be26b5dc.yaml

@@ -4,7 +4,6 @@ fixes:
     Fixed a bug in :class:`.CommutationChecker` when evaluating commutativity relations between
     controlled-rotation gates :class:`.CRXGate`, :class:`.CRYGate` and :class:`.CRZGate` and other
     standard gates. In particular, fixed the behavior of :class:`.CommutationAnalysis`,
-    :class:`.CommutativeCancellation`, `CommutativeOptimization` and
-    `qk_transpiler_pass_standalone_commutative_cancellation` passes in the presence of such gates.
-
+    :class:`.CommutativeCancellation`, :class:`CommutativeOptimization` and
+    :c:func:`qk_transpiler_pass_standalone_commutative_cancellation` passes in the presence of such gates.
     See `#16164 <https://github.com/Qiskit/qiskit/issues/16164>`__.

releasenotes/notes/2.5/add-support-commutation-x-gates-03757fcd282434a9.yaml

@@ -1,7 +1,7 @@
 ---
 features_transpiler:
   - The :class:`.CommutativeCancellation` transpiler pass will now emit an
-    :class:`.XGate` or :class:`.SXGate` if the tracked X rotation is an angle
-    of $\pi$ or a multiple of $\frac{\pi}{2}$ respectively and either the
-    :class:`.Target` supports class:`.XGate` or :class:`.SXGate`, or the
+    :class:`.XGate` or :class:`.SXGate` if the tracked X rotation is a multiple
+    of :math:`\pi` or a multiple of :math:`\pi/2` respectively and either the
+    :class:`.Target` supports :class:`.XGate` or :class:`.SXGate`, or the
     circuit contains the gates.

releasenotes/notes/2.5/deprecation_15004-27fdb0281d08d502.yaml

@@ -3,6 +3,6 @@ deprecations_synthesis:
   - |
     Some of the arguments in the function :func:`.qs_decomposition` are now deprecated:
 
-     * `opt_a1` and `opt_a2` are no longer needed, as the funcion automatically determines if the optimizations should run
-     * `decomposer_1q` won't accept any callable anymore, only instances of :class:`.OneQubitBasisDecomposer`
-     * `decomposer_2q` won't accept any callable anymore, only instances of :class:`.TwoQubitBasisDecomposer`
+     * ``opt_a1`` and ``opt_a2`` are no longer needed, as the funcion automatically determines if the optimizations should run
+     * ``decomposer_1q`` won't accept any callable anymore, only instances of :class:`.OneQubitEulerDecomposer`
+     * ``decomposer_2q`` won't accept any callable anymore, only instances of :class:`.TwoQubitBasisDecomposer`

releasenotes/notes/2.5/fix-ppm-serialization-33203db9d45fa662.yaml

@@ -1,7 +1,7 @@
 fixes:
   - |
     Fixed an issue with the serialization to QPY of 
-    :class:`.PauliProductMeasurement` and :class:`.PauliProductRotation`
+    :class:`.PauliProductMeasurement` and :class:`.PauliProductRotationGate`
     where the paulis were treated incorrectly, breaking compatibility with
     previous qiskit versions.
     See `#16099 <https://github.com/Qiskit/qiskit/issues/16099>`__.

releasenotes/notes/2.5/fix-template-optimization-global-phase-14537.yaml

@@ -1,7 +1,7 @@
 ---
 fixes:
   - |
-    Fixed global phase handling in :class:`.TemplateSubstitution`, which previously
+    Fixed global phase handling in :class:`.TemplateOptimization`, which previously
     silently reset to input circuit's global phase to zero and didn't correct for global
     phases of the templates.
 

releasenotes/notes/2.5/mimalloc-403d3300aa698fae.yaml

@@ -1,8 +1,9 @@
 ---
-features_misc:
+performance:
   - The precompiled wheels on PyPI are now built using `mimalloc <https://github.com/microsoft/mimalloc>`__
     as the global allocator. This improves the runtime performance of Qiskit and also should
-    reduce the resident set size when running Qiskit.
+    reduce the resident set size when running Qiskit. For details see
+    `#16024 <https://github.com/Qiskit/qiskit/pull/16024>`__
 build:
   - Qiskit can now be optionally built using `mimalloc <https://github.com/microsoft/mimalloc>`__
     as the global allocator to improve the runtime performance of Qiskit.

releasenotes/notes/2.5/reset_duration_genericbackend-6a3be7c7e43ea2d4.yaml

@@ -1,5 +1,5 @@
 ---
 fixes:
   - |
-    Circuits with reset instruction now can be scheduled using :class:`~GenericBackendV2`. The problem was that 
-    :class:`~GenericBackendV2` did not have a duration for `reset`.
+    Circuits with reset instruction now can be scheduled using :class:`.GenericBackendV2`. The problem was that 
+    :class:`.GenericBackendV2` did not have a duration for `reset`.

releasenotes/notes/2.5/sabre-lookahead-layers-c00ad07aebfc751c.yaml

@@ -1,6 +1,6 @@
 ---
 features_transpiler:
-  - During Sabre layout and routing (:class:`.SabreLayout` and :class:`.SabreRouting`), the
+  - During Sabre layout and routing (:class:`.SabreLayout` and :class:`.SabreSwap`), the
     ``lookahead`` heuristic is now defined in terms of numbers of circuit layers, rather than numbers
     of circuit gates.  This reduces bias in the heuristic; for narrow circuits the previous
     heuristic could consider very far-off gates at equal weight to nearly routable gates, while for

releasenotes/notes/2.5/two_qubit_peephole-de6d3438ed7df6a9.yaml

@@ -30,7 +30,7 @@ features_transpiler:
         optimized.draw("mpl")
 
     This functionality could previously have been perfomed by running
-    :class:`.ConsolidateBlocks`, and:class:`.UnitarySynthesis` sequentially
+    :class:`.ConsolidateBlocks`, and :class:`.UnitarySynthesis` sequentially
     in your pass manager. However, this new pass offers improved runtime
     performance by performing the synthesis in parallel. It also has improved
     heuristics enabled by doing the optimization in a single step which can

releasenotes/notes/prepare-2.5.0-019f9e4ce2df3e42.yaml

@@ -0,0 +1,45 @@
+---
+prelude: |
+   Qiskit v2.5.0 is a new feature release of the Qiskit SDK.
+   The highlights of this release are:
+
+   * **C API functions for inspecting control flow instructions and classical expressions:**
+     This release introduces a set of new C API functions for inspecting control-flow instructions 
+     and classical expressions. These additions provide low-level access to the structure and 
+     semantics of control-flow constructs, enabling external tools and integrations to analyze, 
+     traverse, and reason about classical logic embedded within quantum circuits.
+
+   * **A multi-IR staged pass manager support:**
+     This release introduces a staged pass manager capable of handling workflows across multiple 
+     intermediate representations (IRs) through the new :class:`.MultiStagePassManager` . 
+     This design enables multi-IR compiler pipelines by allowing each stage to either preserve 
+     the current IR or lower into a different one.
+
+   * **A Pauli Based Computation transpilation pipeline:**
+     This release introduces a new utility function :func:`~.generate_preset_pbc_pass_manager`, 
+     for Pauli-based compilation as an incremental step toward fault-tolerant support. 
+     It enables the construction of preset pass managers targeting the Pauli Based Computation (PBC) 
+     intermediate representation. The generated pipelines compile circuits into a form expressed in terms 
+     of :class:`.PauliProductRotationGate` and :class:`.PauliProductMeasurement`, 
+     while preserving standard operations and control flow. 
+     This provides a convenient entry point for workflows that operate on or optimize within the PBC representation.
+
+   * **Multi-threaded tranpiler passes:**
+     This release introduces a new multithreaded transpiler pass, :class:`.TwoQubitPeepholeOptimization`, 
+     designed to perform local two-qubit unitary optimizations. The pass identifies two-qubit subcircuits, 
+     computes their unitary representation, and replaces them with more efficient synthesized equivalents, 
+     reducing overall gate count or circuit fidelity. 
+     In addition, several existing passes, :class:`.ConsolidateBlocks`, :class:`.CommutationAnalysis`, 
+     and :class:`.Optimize1qGatesDecomposition` have been enhanced with multithreading support, 
+     improving performance and scalability of transpilation on larger circuits.
+
+   * **Improved tranpiler performance:**
+     This release delivers substantial transpiler performance improvements, 
+     with average speedups of 2-3x across typical workloads, driven by a broad set of optimizations 
+     throughout the stack. Key contributions include algorithmic improvements to the SABRE layout and routing methods 
+     snd the multithreaded transpiler passes mentioned above. 
+     Additional gains come from switching to `mimalloc <https://github.com/microsoft/mimalloc>__` as the global allocator, 
+     adopting `nalgebra <https://docs.rs/nalgebra/latest/nalgebra/>`__ for matrix operations in decomposers, 
+     using `foldhash <https://github.com/orlp/foldhash>`__ in place of previous hashing strategies, 
+     and incorporating performance improvements in `rustworkx <https://www.rustworkx.org/>`__, 
+     together significantly enhancing both runtime efficiency and scalability.