0.7.0-notes.rst

.. currentmodule:: cuda.core

cuda.core 0.7.0 Release Notes

Highlights

• Introduced support for explicit graph construction. CUDA graphs can now be built programmatically by adding nodes and edges, and their topology can be modified after construction.
• Added CUDA-OpenGL interoperability support, enabling zero-copy sharing of GPU memory between CUDA compute kernels and OpenGL renderers.
• Added :class:`TensorMapDescriptor` for Hopper+ TMA (Tensor Memory Accelerator) bulk data movement, with automatic kernel argument integration.
• :class:`~utils.StridedMemoryView` now supports DLPack export via from_dlpack() array API.

New features

• Added the :mod:`cuda.core.graph` public module containing :class:`~graph.GraphDef` for explicit graph construction, typed node subclasses, and supporting types. :class:`~graph.GraphBuilder` (stream capture) also moves into this module.
• Added :meth:`~graph.GraphBuilder.callback` for CPU callbacks during stream capture, mirroring the existing :meth:`~graph.GraphDef.callback` API.
• Added :class:`GraphicsResource` for CUDA-OpenGL interoperability. Factory classmethods :meth:`~GraphicsResource.from_gl_buffer` and :meth:`~GraphicsResource.from_gl_image` register OpenGL objects for CUDA access, and mapping returns a :class:`Buffer` for zero-copy kernel use.
• Added :class:`TensorMapDescriptor` wrapping the CUDA driver's CUtensorMap for Hopper+ TMA (Tensor Memory Accelerator) bulk data movement. :class:`~utils.StridedMemoryView` gains an :meth:`~utils.StridedMemoryView.as_tensor_map` method for convenient descriptor creation, with automatic dtype inference, stride computation, and first-class kernel argument integration.
• Added DLPack export support to :class:`~utils.StridedMemoryView` via __dlpack__ and __dlpack_device__, complementing the existing import path.
• Added the DLPack C exchange API (__dlpack_c_exchange_api__) to :class:`~utils.StridedMemoryView`.
• Added NVRTC precompiled header (PCH) support (CUDA 12.8+). :class:`ProgramOptions` gains pch, create_pch, use_pch, pch_dir, and related options. :attr:`Program.pch_status` reports the PCH creation outcome, and :meth:`~Program.compile` automatically resizes the NVRTC PCH heap and retries when PCH creation fails due to heap exhaustion.
• Added NUMA-aware managed memory pool placement. :class:`ManagedMemoryResourceOptions` gains a preferred_location_type option ("device", "host", or "host_numa"), and :attr:`ManagedMemoryResource.preferred_location` queries the resolved location. The existing preferred_location parameter retains full backwards compatibility.
• Added NUMA-aware pinned memory pool placement. :class:`PinnedMemoryResourceOptions` gains a numa_id option, and :attr:`PinnedMemoryResource.numa_id` queries the host NUMA node ID used for pool placement. When ipc_enabled=True and numa_id is not set, the NUMA node is automatically derived from the current CUDA device.
• Added support for CUDA 13.2.

New examples

• gl_interop_plasma.py: Real-time plasma effect demonstrating CUDA-OpenGL interoperability via :class:`GraphicsResource`.
• tma_tensor_map.py: TMA bulk data movement using :class:`TensorMapDescriptor` on Hopper+ GPUs.

Fixes and enhancements

• Fixed managed memory buffers being misclassified as kDLCUDAHost in DLPack device mapping. They are now correctly reported as kDLCUDAManaged. (#1863)
• Fixed IPC-enabled pinned memory pools using a hardcoded NUMA node ID of 0 instead of the NUMA node closest to the active CUDA device. On multi-NUMA systems where the device is attached to a non-zero host NUMA node, this could cause pool creation or allocation failures. (#1603)
• Fixed :attr:`DeviceMemoryResource.peer_accessible_by` returning stale results when wrapping a non-owned (default) memory pool. The property now always queries the CUDA driver for non-owned pools, so multiple wrappers around the same pool see consistent state. (#1720)
• Fixed a bare except clause in stream acceptance that silently swallowed all exceptions, including KeyboardInterrupt and SystemExit. Only the expected "protocol not supported" case is now caught. (#1631)
• :class:`~utils.StridedMemoryView` now validates strides at construction time so unsupported layouts fail immediately instead of on first metadata access. (#1429)
• IPC file descriptor cleanup now uses a C++ shared_ptr with a POSIX deleter, avoiding cryptic errors when a :class:`DeviceMemoryResource` is destroyed during Python shutdown.
• Improved error message when :class:`ManagedMemoryResource` is called without options on platforms that lack a default managed memory pool (e.g. WSL2). (#1617)
• Handle properties on core API objects now return None during Python shutdown instead of crashing.
• Reduced Python overhead in :class:`Program` and :class:`Linker` by moving compilation and linking operations to the C level and releasing the GIL during backend calls. This benefits workloads that create many programs or linkers, and enables concurrent compilation in multithreaded applications.
• Error enum explanations are now derived from cuda-bindings docstrings when available (bindings 12.9.6+ or 13.2.0+), with frozen tables as fallback for older versions.
• Improved optional dependency handling for NVVM and nvJitLink imports so that only genuinely missing optional modules are treated as unavailable; unrelated import failures now surface normally, and cuda.core now depends directly on cuda-pathfinder.