.. SPDX-FileCopyrightText: Copyright (c) 2025-2026 NVIDIA CORPORATION & AFFILIATES. All rights reserved.
.. SPDX-License-Identifier: LicenseRef-NVIDIA-SOFTWARE-LICENSE

.. module:: cuda.bindings

``cuda-bindings`` 12.9.6 Release notes
======================================

Highlights
----------

* ``cuda.bindings.nvml`` has graduated from experimental (``cuda.bindings._nvml``)
  to a fully supported public module with extensive handwritten Pythonic API
  coverage spanning ~170 functions across system queries, device discovery,
  memory, power, clocks, utilization, thermals, NVLink, and device configuration.
  (`PR #1524 <https://github.com/NVIDIA/cuda-python/pull/1524>`_,
  `PR #1548 <https://github.com/NVIDIA/cuda-python/pull/1548>`_)
* Add ``nvFatbin`` bindings.
  (`PR #1467 <https://github.com/NVIDIA/cuda-python/pull/1467>`_)
* Performance improvement: ``cuda.bindings`` now uses a faster ``enum``
  implementation, rather than the standard library's ``enum.IntEnum``.
  This leads to much faster import times, and slightly faster attribute access
  times.
  (`PR #1581 <https://github.com/NVIDIA/cuda-python/pull/1581>`_)
* Multiple performance improvements cumulatively reducing Python-to-C call
  overhead through faster ``void *`` conversion, faster result returning,
  optimized enum-to-vector conversion, and stack-allocated small arrays.

Bugfixes
--------

* Fixed an issue where the ``CU_POINTER_ATTRIBUTE_DEVICE_ORDINAL`` attribute was
  retrieved as an unsigned int, rather than a signed int.
  (`PR #1336 <https://github.com/NVIDIA/cuda-python/pull/1336>`_)
* Fixed a use-after-free in ``_HelperInputVoidPtr`` properties when backed by
  Python buffer objects.
  (`PR #1629 <https://github.com/NVIDIA/cuda-python/pull/1629>`_)

Miscellaneous
-------------

* Faster ``void *`` conversion using stack-allocated buffers instead of heap
  allocation.
  (`PR #1616 <https://github.com/NVIDIA/cuda-python/pull/1616>`_)
* Faster returning of results from driver, runtime, and NVRTC bindings.
  (`PR #1647 <https://github.com/NVIDIA/cuda-python/pull/1647>`_,
  `PR #1656 <https://github.com/NVIDIA/cuda-python/pull/1656>`_)
* Faster conversion of enum sequences to vectors by eliminating temporary
  Python objects.
  (`PR #1667 <https://github.com/NVIDIA/cuda-python/pull/1667>`_)
* Stack-allocated small numeric arrays in driver bindings, reducing heap
  allocation overhead.
  (`PR #1545 <https://github.com/NVIDIA/cuda-python/pull/1545>`_)
* NVML bindings now use ``cuda_pathfinder`` for library discovery, consistent
  with other CUDA libraries.
  (`PR #1661 <https://github.com/NVIDIA/cuda-python/pull/1661>`_)
* ``CUDA_HOME`` is no longer required at metadata resolution time (e.g.
  ``pip install --dry-run``, ``uv lock``); it is only needed at actual build time.
  (`PR #1652 <https://github.com/NVIDIA/cuda-python/pull/1652>`_)

Known issues
------------

* Updating from older versions (v12.6.2.post1 and below) via ``pip install -U cuda-python`` might not work. Please do a clean re-installation by uninstalling ``pip uninstall -y cuda-python`` followed by installing ``pip install cuda-python``.