.. SPDX-License-Identifier: Apache-2.0

.. currentmodule:: cuda.core.experimental

``cuda.core`` 0.3.0 Release Notes
=================================

Released on MM DD, 2025


Highlights
----------

- Starting this release ``cuda.core`` is licensed under Apache 2.0. The biggest implication of this change is that we are open to external contribution now! Please kindly follow the :ref:`Contributor Guide <contributer_guide>` for detailed instructions.


Breaking Changes
----------------

- The :class:`Buffer` object's ``__init__()`` method is removed, see below.
- The :class:`Buffer` object's :meth:`~Buffer.close` method and destructor now always defer to the underlying memory resource implementation
  to decide the behavior if a stream is not explicitly passed. Previously, in this case it always uses the default stream, which could
  interfere with the memory resource's assumptions.


New features
------------

- :class:`~_module.Kernel` adds :attr:`~_module.Kernel.num_arguments` and :attr:`~_module.Kernel.arguments_info` for introspection of kernel arguments. (#612)
- Add pythonic access to kernel occupancy calculation functions via :attr:`Kernel.occupancy`. (#648)
- Support launching cooperative kernels by setting :attr:`LaunchConfig.cooperative_launch` to `True`.
- A name can be assigned to :class:`ObjectCode` instances generated by both :class:`Program` and :class:`Linker` through their respective options.
- Expose :class:`Buffer`, :class:`DeviceMemoryResource`, :class:`LegacyPinnedMemoryResource`, and :class:`MemoryResource` to the top namespace.
   - Before this release, the internal :class:`Buffer` class had an ``__init__()`` constructor. To align with the design of cuda.core objects,
     this constructor is removed starting this release. Users who still need the old behavior should use the :meth:`~Buffer.from_handle`
     alternative constructor.
- Add a typing annotation for the :attr:`~_stream.IsStreamT.__cuda_stream__` protocol.


New examples
------------

- Add a PyTorch-based example.
- Split the :class:`StridedMemoryView` example into two (CPU/GPU).


Fixes and enhancements
----------------------

- ``cuda.core`` now raises more clear and actionable error messages whenever possible.
- :class:`ObjectCode` can be pickled now.
- Look-up of the :attr:`Event.device` and :attr:`Event.context` (the device and CUDA context where an event was created from) is now possible.
- :class:`Event`-based timing is made more robust (also with better error messages).
- The :func:`launch` function's handling of fp16 scalars was incorrect and is fixed.
- :attr:`ProgramOptions.ptxas_options` can now accept more than one argument.
- The :class:`Device` constructor is made faster.
- The CFFI-based example no longer leaves the intermediate files on disk after it finishes.