mdspan to DLPack

docs/libcudacxx/extended_api/mdspan.rst

@@ -10,6 +10,7 @@ Mdspan
    mdspan/host_device_accessor
    mdspan/restrict_accessor
    mdspan/shared_memory_accessor
+   mdspan/mdspan_to_dlpack
 
 .. list-table::
    :widths: 25 45 30 30
@@ -34,3 +35,8 @@ Mdspan
      - ``mdspan`` and accessor for CUDA shared memory
      - CCCL 3.2.0
      - CUDA 13.2
+
+   * - :ref:`mdspan to dlpack <libcudacxx-extended-api-mdspan-mdspan-to-dlpack>`
+     - Convert a ``mdspan`` to a ``DLTensor``
+     - CCCL 3.2.0
+     - CUDA 13.2

docs/libcudacxx/extended_api/mdspan/mdspan_to_dlpack.rst

@@ -0,0 +1,134 @@
+.. _libcudacxx-extended-api-mdspan-mdspan-to-dlpack:
+
+``mdspan`` to DLPack
+====================
+
+This functionality provides a conversion from ``cuda::host_mdspan``, ``cuda::device_mdspan``, and ``cuda::managed_mdspan`` to `DLPack <https://dmlc.github.io/dlpack/latest/>`__ ``DLTensor`` view.
+
+Defined in the ``<cuda/mdspan>`` header.
+
+Conversion functions
+--------------------
+
+.. code:: cuda
+
+   namespace cuda {
+
+   template <typename T, typename Extents, typename Layout, typename Accessor>
+   [[nodiscard]] dlpack_tensor<Extents::rank()>
+   to_dlpack(const cuda::host_mdspan<T, Extents, Layout, Accessor>& mdspan);
+
+   template <typename T, typename Extents, typename Layout, typename Accessor>
+   [[nodiscard]] dlpack_tensor<Extents::rank()>
+   to_dlpack(const cuda::device_mdspan<T, Extents, Layout, Accessor>& mdspan,
+                    cuda::device_ref device = cuda::device_ref{0});
+
+   template <typename T, typename Extents, typename Layout, typename Accessor>
+   [[nodiscard]] dlpack_tensor<Extents::rank()>
+   to_dlpack(const cuda::managed_mdspan<T, Extents, Layout, Accessor>& mdspan);
+
+   } // namespace cuda
+
+Types
+-----
+
+.. code:: cuda
+
+  namespace cuda {
+
+  template <size_t Rank>
+  class dlpack_tensor {
+  public:
+      dlpack_tensor();
+      dlpack_tensor(const dlpack_tensor&) noexcept;
+      dlpack_tensor(dlpack_tensor&&) noexcept;
+      dlpack_tensor& operator=(const dlpack_tensor&) noexcept;
+      dlpack_tensor& operator=(dlpack_tensor&&) noexcept;
+      ~dlpack_tensor() noexcept = default;
+
+      DLTensor&       get() noexcept;
+      const DLTensor& get() const noexcept;
+  };
+
+  } // namespace cuda
+
+``cuda::dlpack_tensor`` stores a ``DLTensor`` and owns the backing storage for its ``shape`` and ``strides`` pointers. The class does not use any heap allocation.
+
+.. note:: **Lifetime**
+
+  The ``DLTensor`` associated with ``cuda::dlpack_tensor`` must not outlive the wrapper. If the wrapper is destroyed, the returned ``DLTensor::shape`` and ``DLTensor::strides`` pointers will dangle.
+
+.. note:: **Const-correctness**
+
+  ``DLTensor::data`` points at ``mdspan.data_handle()`` (or is ``nullptr`` if ``mdspan.size() == 0``). If ``T`` is ``const``, the pointer is ``const_cast``'d because ``DLTensor::data`` is unqualified.
+
+Semantics
+---------
+
+The conversion produces a non-owning DLPack view of the ``mdspan`` data and metadata:
+
+- ``DLTensor::ndim`` is ``mdspan.rank()``.
+- For rank > 0, ``DLTensor::shape[i]`` is ``mdspan.extent(i)``.
+- For rank > 0, ``DLTensor::strides[i]`` is ``mdspan.stride(i)``.
+- ``DLTensor::byte_offset`` is always ``0``.
+- ``DLTensor::device`` is:
+
+  - ``{kDLCPU, 0}`` for ``cuda::host_mdspan``
+  - ``{kDLCUDA, device.get()}`` for ``cuda::device_mdspan``
+  - ``{kDLCUDAManaged, 0}`` for ``cuda::managed_mdspan``
+
+Element types are mapped to ``DLDataType`` according to the DLPack conventions, including:
+
+- ``bool``.
+- Signed and unsigned integers.
+- IEEE-754 Floating-point and extended precision floating-point, including ``__half``, ``__nv_bfloat16``, ``__float128``, FP8, FP6, FP4 when available.
+- Complex: ``cuda::std::complex<__half>``, ``cuda::std::complex<float>``, and ``cuda::std::complex<double>``.
+- `CUDA built-in vector types <https://docs.nvidia.com/cuda/cuda-programming-guide/05-appendices/cpp-language-extensions.html#built-in-types>`__, such as ``int2``, ``float4``, etc.
+- Vector types for extended floating-point, such as ``__half2``, ``__nv_fp8x4_e4m3``, etc.
+
+Constraints
+-----------
+
+- The accessor ``data_handle_type`` must be a pointer type.
+
+Runtime errors
+--------------
+
+- If any ``extent(i)`` or ``stride(i)`` cannot be represented in ``int64_t``, the conversion raises an exception.
+
+Availability notes
+------------------
+
+- This API is available only when DLPack header is present, namely ``<dlpack/dlpack.h>`` is found in the include path.
+
+References
+----------
+
+- `DLPack C API <https://dmlc.github.io/dlpack/latest/c_api.html>`__ documentation.
+
+Example
+-------
+
+.. code:: cuda
+
+  #include <dlpack/dlpack.h>
+  #include <cuda/mdspan>
+  #include <cuda/std/cassert>
+  #include <cuda/std/cstdint>
+
+  int main() {
+    using extents_t = cuda::std::extents<size_t, 2, 3>;
+
+    int data[6] = {0, 1, 2, 3, 4, 5};
+    cuda::host_mdspan<int, extents_t> md{data, extents_t{}};
+
+    auto dl              = cuda::to_dlpack(md);
+    const auto& dltensor = dl.get();
+
+    // `dl` owns the shape/stride storage; `dltensor.data` is a non-owning pointer to `data`.
+    assert(dltensor.device.device_type == kDLCPU);
+    assert(dltensor.ndim == 2);
+    assert(dltensor.shape[0] == 2 && dltensor.shape[1] == 3);
+    assert(dltensor.strides[0] == 3 && dltensor.strides[1] == 1);
+    assert(dltensor.data == data);
+  }