[doc-only] Address review feedback on v0.6.0 release notes

Co-authored-by: Cursor <cursoragent@cursor.com>

Author: Andy-Jost

cuda_core/cuda/core/_memoryview.pyx

@@ -152,33 +152,48 @@ cdef class StridedMemoryView:
 
     @classmethod
     def from_dlpack(cls, obj: object, stream_ptr: int | None=None) -> StridedMemoryView:
+        """Create a view from an object supporting the DLPack_ protocol.
+
+        Parameters
+        ----------
+        obj : object
+            An object implementing the DLPack_ protocol (via ``__dlpack__``).
+        stream_ptr : int, optional
+            Stream pointer for synchronization. If ``None``, no synchronization is performed.
+
+        .. _DLPack: https://dmlc.github.io/dlpack/latest/
+        """
         cdef StridedMemoryView buf = StridedMemoryView.__new__(cls)
         view_as_dlpack(obj, stream_ptr, buf)
         return buf
 
     @classmethod
     def from_cuda_array_interface(cls, obj: object, stream_ptr: int | None=None) -> StridedMemoryView:
-        """Create a view from an object supporting the ``__cuda_array_interface__`` protocol.
+        """Create a view from an object supporting the `__cuda_array_interface__`_ protocol.
 
         Parameters
         ----------
         obj : object
-            An object implementing the ``__cuda_array_interface__`` protocol.
+            An object implementing the `__cuda_array_interface__`_ protocol.
         stream_ptr : int, optional
             Stream pointer for synchronization. If ``None``, no synchronization is performed.
+
+        .. _`__cuda_array_interface__`: https://numba.readthedocs.io/en/stable/cuda/cuda_array_interface.html
         """
         cdef StridedMemoryView buf = StridedMemoryView.__new__(cls)
         view_as_cai(obj, stream_ptr, buf)
         return buf
 
     @classmethod
     def from_array_interface(cls, obj: object) -> StridedMemoryView:
-        """Create a view from an object supporting the ``__array_interface__`` protocol.
+        """Create a view from an object supporting the `__array_interface__`_ protocol.
 
         Parameters
         ----------
         obj : object
-            An object implementing the ``__array_interface__`` protocol (e.g., a numpy array).
+            An object implementing the `__array_interface__`_ protocol (e.g., a numpy array).
+
+        .. _`__array_interface__`: https://numpy.org/doc/stable/reference/arrays.interface.html
         """
         cdef StridedMemoryView buf = StridedMemoryView.__new__(cls)
         view_as_array_interface(obj, buf)
@@ -188,14 +203,17 @@ cdef class StridedMemoryView:
     def from_any_interface(cls, obj: object, stream_ptr: int | None = None) -> StridedMemoryView:
         """Create a view by automatically selecting the best available protocol.
 
-        Tries DLPack first, then falls back to ``__cuda_array_interface__``.
+        Tries DLPack_ first, then falls back to `__cuda_array_interface__`_.
 
         Parameters
         ----------
         obj : object
-            An object implementing DLPack or ``__cuda_array_interface__``.
+            An object implementing DLPack_ or `__cuda_array_interface__`_.
         stream_ptr : int, optional
             Stream pointer for synchronization. If ``None``, no synchronization is performed.
+
+        .. _DLPack: https://dmlc.github.io/dlpack/latest/
+        .. _`__cuda_array_interface__`: https://numba.readthedocs.io/en/stable/cuda/cuda_array_interface.html
         """
         if check_has_dlpack(obj):
             return cls.from_dlpack(obj, stream_ptr)

cuda_core/cuda/core/_program.pyx

@@ -306,12 +306,13 @@ class ProgramOptions:
         Default: False
     extra_sources : list of 2-tuples or tuple of 2-tuples, optional
         Additional NVVM IR modules to compile together with the main program, specified as
-        ``((name1, source1), (name2, source2), ...)``. Each source can be a string (textual
-        LLVM IR), bytes, or bytearray. Only supported for the NVVM backend.
+        ``((name1, source1), (name2, source2), ...)``. Each name is a string identifier used
+        in diagnostic messages. Each source can be a string (textual LLVM IR) or bytes/bytearray
+        (LLVM bitcode). Only supported for the NVVM backend.
         Default: None
     use_libdevice : bool, optional
-        Load NVIDIA's libdevice math builtins library during NVVM compilation. Only
-        supported for the NVVM backend.
+        Load NVIDIA's `libdevice <https://docs.nvidia.com/cuda/libdevice-users-guide/>`_
+        math builtins library. Only supported for the NVVM backend.
         Default: False
     """
 

cuda_core/docs/source/api.rst

@@ -23,8 +23,6 @@ CUDA runtime
    Graph
    GraphBuilder
    launch
-   LEGACY_DEFAULT_STREAM
-   PER_THREAD_DEFAULT_STREAM
 
    :template: autosummary/cyclass.rst
 
@@ -51,6 +49,18 @@ CUDA runtime
    LaunchConfig
    VirtualMemoryResourceOptions
 
+.. data:: LEGACY_DEFAULT_STREAM
+
+   The legacy default CUDA stream. All devices share the same legacy default
+   stream, and work launched on it is not concurrent with work on any other
+   stream.
+
+.. data:: PER_THREAD_DEFAULT_STREAM
+
+   The per-thread default CUDA stream. Each host thread has its own per-thread
+   default stream, and work launched on it can execute concurrently with work
+   on other non-blocking streams.
+
 
 CUDA compilation toolchain
 --------------------------

cuda_core/docs/source/release/0.6.0-notes.rst

@@ -35,7 +35,6 @@ New features
 
   - Added ``from_array_interface`` constructor for creating views from numpy arrays.
   - Improved structured dtype array support.
-  - Relaxed the power-of-two itemsize check in ``StridedLayout``.
   - Added bfloat16 dlpack support when the optional ``ml_dtypes`` package is installed.
 
 - Added public access to default CUDA streams via module-level constants
@@ -56,32 +55,30 @@ New features
   ``cuda.core`` and the installed ``cuda-bindings`` version.
 
 
-New examples
-------------
-
-None.
-
-
 Fixes and enhancements
 ----------------------
 
-- Reduced wheel and installed package sizes by excluding Cython source files and build
-  artifacts from distribution packages.
+- Eliminated spurious CUDA driver errors during interpreter shutdown by ensuring
+  resources are destroyed in the correct order.
+
+- Fixed a bug preventing weak references to core API objects.
+
+- Fixed zero-sized allocations in legacy memory resources, which previously failed on
+  certain platforms.
 
 - Improved performance by Cythonizing :class:`Program` and :class:`ObjectCode` internals.
 
 - Reduced :class:`~utils.StridedMemoryView` construction overhead.
 
-- Legacy and per-thread default streams are now singletons, ensuring consistent identity
-  across the application.
-
 - ``__hash__`` and ``__eq__`` on core API classes no longer require a CUDA context.
 
 - Device attribute queries now gracefully handle unsupported attributes on older CUDA
   drivers, returning sensible defaults instead of raising errors.
 
-- Fixed zero-sized allocations in legacy memory resources, which previously failed on
-  certain platforms.
-
 - Added a warning when :class:`ManagedMemoryResource` is created on platforms without
   concurrent managed access support.
+
+- Reduced wheel and installed package sizes by excluding Cython source files and build
+  artifacts from distribution packages.
+
+- Slightly improved typing support.