Merge branch 'main' into cythonize_more

Author: leofang

.pre-commit-config.yaml

@@ -33,10 +33,11 @@ repos:
       - id: no-markdown-in-docs-source
         name: Prevent markdown files in docs/source directories
         entry: bash -c
-        args: ['if find . -path "*/docs/source/*.md" -not -path "./docs/README.md" | grep -q .; then echo "ERROR: Markdown files found in docs/source/ directories. Use reST (.rst) instead."; exit 1; fi']
+        args:
+          - 'for file in "$@"; do >&2 echo "error: markdown file "$file" found in a docs/source directory. Only reST (.rst) is allowed"; done && exit 1'
+          - "_" # fake script name, because bash considers $0 (the first argument) to be the script name
         language: system
-        pass_filenames: false
-        always_run: true
+        files: '^.*/docs/source/.*\.md$'
 
   # Standard hooks
   - repo: https://github.com/pre-commit/pre-commit-hooks

cuda_bindings/cuda/bindings/_version.py

@@ -1,4 +1,4 @@
 # SPDX-FileCopyrightText: Copyright (c) 2024-2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 # SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
 
-__version__ = "13.0.1"
+__version__ = "13.0.2"

…ngs/docs/source/release/12.9.X-notes.rst …ngs/docs/source/release/12.9.3-notes.rstcuda_bindings/docs/source/release/12.9.X-notes.rst renamed to cuda_bindings/docs/source/release/12.9.3-notes.rst cuda_bindings/docs/source/release/12.9.X-notes.rst renamed to cuda_bindings/docs/source/release/12.9.3-notes.rst

@@ -16,7 +16,7 @@ Highlights
 * The Python overhead of calling functions in CUDA bindings in ``driver``, ``runtime`` and ``nvrtc`` has been reduced by approximately 30%.
 * Updated the ``cuda.bindings.runtime`` module to statically link against the CUDA Runtime library from CUDA Toolkit 12.9.1.
 * ``cyruntime.getLocalRuntimeVersion`` now uses pathfinder to find the CUDA runtime.
-
+* Experimental free-threaded builds are available on PyPI. More details are available in our :ref:`support` docs.
 
 Known issues
 ------------

…ngs/docs/source/release/13.X.Y-notes.rst …ngs/docs/source/release/13.0.2-notes.rstcuda_bindings/docs/source/release/13.X.Y-notes.rst renamed to cuda_bindings/docs/source/release/13.0.2-notes.rst cuda_bindings/docs/source/release/13.X.Y-notes.rst renamed to cuda_bindings/docs/source/release/13.0.2-notes.rst

@@ -19,6 +19,7 @@ Highlights
 * On Windows, the ``pywin32`` dependency has been removed. The necessary Windows API functions are now accessed directly.
 * Updated the ``cuda.bindings.runtime`` module to statically link against the CUDA Runtime library from CUDA Toolkit 13.0.1.
 * ``cyruntime.getLocalRuntimeVersion`` now uses pathfinder to find the CUDA runtime.
+* Experimental free-threaded builds are available on PyPI. More details are available in our :ref:`support` docs.
 
 
 Bug fixes

cuda_bindings/docs/source/support.rst

@@ -1,6 +1,8 @@
 .. SPDX-FileCopyrightText: Copyright (c) 2025 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
 .. SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE
 
+.. _support:
+
 ``cuda.bindings`` Support Policy
 ================================
 
@@ -23,9 +25,20 @@ The ``cuda.bindings`` module has the following support policy:
    module could require Cython layer users to rebuild their projects and update their pinning to
    this module.
 
+Free-threading Build Support
+----------------------------
+As of cuda-bindings 13.0.2/12.9.3, wheels for the `free-threaded interpreter`_ are shipped to PyPI.
+
+1. Support for these builds is best effort, due to heavy use of `built-in
+   modules that are known to be thread-unsafe`_, such as ``ctypes``.
+2. For now, you are responsible for making sure that calls into ``cuda-bindings``
+   libraries are thread-safe. This is subject to change.
+
 The NVIDIA CUDA Python team reserves rights to amend the above support policy. Any major changes,
 however, will be announced to the users in advance.
 
 
 .. _CUDA minor version compatibility: https://docs.nvidia.com/deploy/cuda-compatibility/#minor-version-compatibility
 .. _CPython EOL schedule: https://devguide.python.org/versions/
+.. _built-in modules that are known to be thread-unsafe: https://github.com/python/cpython/issues/116738
+.. _free-threaded interpreter: https://docs.python.org/3/howto/free-threading-python.html

cuda_core/build_hooks.py

@@ -45,7 +45,7 @@ def _get_proper_cuda_bindings_major_version() -> str:
         m = re.search(r"CUDA Version:\s*([\d\.]+)", out.stdout.decode())
         if m:
             return m.group(1).split(".")[0]
-    except FileNotFoundError:
+    except (FileNotFoundError, subprocess.CalledProcessError):
         # the build machine has no driver installed
         pass
 

cuda_core/docs/source/install.rst

@@ -42,6 +42,18 @@ and likewise use ``[cu13]`` for CUDA 13.
 Note that using ``cuda.core`` with NVRTC installed from PyPI via ``pip install`` requires
 ``cuda.bindings`` 12.8.0+. Likewise, with nvJitLink it requires 12.8.0+.
 
+Free-threading Build Support
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+As of cuda-core 0.4.0, wheels for the `free-threaded interpreter`_ are shipped to PyPI.
+
+1. Support for these builds is best effort, due to heavy use of `built-in
+   modules that are known to be thread-unsafe`_, such as ``ctypes``.
+2. For now, you are responsible for making sure that calls into ``cuda-core``
+   libraries are thread-safe. This is subject to change.
+
+.. _built-in modules that are known to be thread-unsafe: https://github.com/python/cpython/issues/116738
+.. _free-threaded interpreter: https://docs.python.org/3/howto/free-threading-python.html
+
 Installing from Conda (conda-forge)
 -----------------------------------