56fd68e7dd
CCCL (CUDA C++ Core Libraries) provides: - CUB: device/block/warp-level GPU primitives (reduce, scan, sort, topk) - Thrust: high-level parallel algorithms (transform_reduce, sort, scan) - libcudacxx: CUDA C++ standard library (atomics, barriers, memory) - cudax: experimental features (memory resources, allocators) - Tuning policies: per-SM hardware-specific algorithm parameters Competition optimization vectors mapped to CCCL: - Output TPS (83% weight): warp_reduce, block_reduce, device_topk - Input TPS (14% weight): device_scan, block_load, prefetch - Cache TPS (3% weight): prefix caching strategy patterns - Memory (0.9 util): pooled/cached/buddy allocators Source: https://github.com/NVIDIA/cccl (shallow clone, HEAD only) License: Apache-2.0
228 lines
6.3 KiB
C++
228 lines
6.3 KiB
C++
// SPDX-FileCopyrightText: Copyright (c) 2008-2021, NVIDIA Corporation. All rights reserved.
|
|
// SPDX-License-Identifier: Apache-2.0
|
|
|
|
/*! \file
|
|
* \brief A pointer to an object which resides in memory associated with the
|
|
* \c device system.
|
|
*/
|
|
|
|
#pragma once
|
|
|
|
#include <thrust/detail/config.h>
|
|
|
|
#if defined(_CCCL_IMPLICIT_SYSTEM_HEADER_GCC)
|
|
# pragma GCC system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_CLANG)
|
|
# pragma clang system_header
|
|
#elif defined(_CCCL_IMPLICIT_SYSTEM_HEADER_MSVC)
|
|
# pragma system_header
|
|
#endif // no system header
|
|
|
|
#include <thrust/memory.h>
|
|
|
|
#include <cuda/std/__memory/addressof.h>
|
|
#include <cuda/std/__memory/pointer_traits.h>
|
|
|
|
THRUST_NAMESPACE_BEGIN
|
|
|
|
/*! \addtogroup memory_management Memory Management
|
|
* \{
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
|
|
template <typename T>
|
|
class device_reference;
|
|
|
|
/*! \brief \c device_ptr is a pointer-like object which points to an object that
|
|
* resides in memory associated with the \ref device system.
|
|
*
|
|
* \c device_ptr has pointer semantics: it may be dereferenced safely from
|
|
* anywhere, including the \ref host, and may be manipulated with pointer
|
|
* arithmetic.
|
|
*
|
|
* \c device_ptr can be created with \ref device_new, \ref device_malloc,
|
|
* \ref device_malloc_allocator, \ref device_allocator, or
|
|
* \ref device_pointer_cast, or by explicitly calling its constructor with a
|
|
* raw pointer.
|
|
*
|
|
* The raw pointer contained in a \c device_ptr may be obtained via \c get
|
|
* member function or the \ref raw_pointer_cast free function.
|
|
*
|
|
* \ref algorithms operating on \c device_ptr types will automatically be
|
|
* dispatched to the \ref device system.
|
|
*
|
|
* \note \c device_ptr is not a smart pointer; it is the programmer's
|
|
* responsibility to deallocate memory pointed to by \c device_ptr.
|
|
*
|
|
* \see device_new
|
|
* \see device_malloc
|
|
* \see device_malloc_allocator
|
|
* \see device_allocator
|
|
* \see device_pointer_cast
|
|
* \see raw_pointer_cast
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename T>
|
|
class device_ptr
|
|
: public thrust::pointer<T, thrust::device_system_tag, thrust::device_reference<T>, thrust::device_ptr<T>>
|
|
{
|
|
private:
|
|
using super_t = thrust::pointer<T, thrust::device_system_tag, thrust::device_reference<T>, thrust::device_ptr<T>>;
|
|
|
|
public:
|
|
_CCCL_HIDE_FROM_ABI device_ptr() = default;
|
|
|
|
/*! \brief Construct a null \c device_ptr.
|
|
*
|
|
* This constructor accepts a \c std::nullptr_t value.
|
|
*
|
|
* \post <tt>get() == nullptr</tt>.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
_CCCL_HOST_DEVICE device_ptr(std::nullptr_t)
|
|
: super_t(nullptr)
|
|
{}
|
|
|
|
/*! \brief Construct a \c device_ptr from a raw pointer which is
|
|
* convertible to \c T*.
|
|
*
|
|
* \tparam U A type whose pointer is convertible to \c T*.
|
|
* \param ptr A raw pointer to a \c U in device memory to construct from.
|
|
*
|
|
* \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
|
|
*
|
|
* \pre \c ptr points to a location in device memory.
|
|
*
|
|
* \post <tt>get() == nullptr</tt>.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename U>
|
|
_CCCL_HOST_DEVICE explicit device_ptr(U* ptr)
|
|
: super_t(ptr)
|
|
{}
|
|
|
|
/*! \brief Copy construct a \c device_ptr from another \c device_ptr whose
|
|
* pointer type is convertible to \c T*.
|
|
*
|
|
* \tparam U A type whose pointer is convertible to \c T*.
|
|
* \param other A \c device_ptr to a \c U to construct from.
|
|
*
|
|
* \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
|
|
*
|
|
* \post <tt>get() == other.get()</tt>.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename U>
|
|
_CCCL_HOST_DEVICE device_ptr(device_ptr<U> const& other)
|
|
: super_t(other)
|
|
{}
|
|
|
|
/*! \brief Set this \c device_ptr to point to the same object as another
|
|
* \c device_ptr whose pointer type is convertible to \c T*.
|
|
*
|
|
* \tparam U A type whose pointer is convertible to \c T*.
|
|
* \param other A \c device_ptr to a \c U to assign from.
|
|
*
|
|
* \pre <tt>std::is_convertible_v<U*, T*> == true</tt>.
|
|
*
|
|
* \post <tt>get() == other.get()</tt>.
|
|
*
|
|
* \return \c *this.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename U>
|
|
_CCCL_HOST_DEVICE device_ptr& operator=(device_ptr<U> const& other)
|
|
{
|
|
super_t::operator=(other);
|
|
return *this;
|
|
}
|
|
|
|
/*! \brief Set this \c device_ptr to null.
|
|
*
|
|
* This operator accepts a \c std::nullptr_t value.
|
|
*
|
|
* \post <tt>get() == nullptr</tt>.
|
|
*
|
|
* \return \c *this.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
_CCCL_HOST_DEVICE device_ptr& operator=(std::nullptr_t)
|
|
{
|
|
super_t::operator=(nullptr);
|
|
return *this;
|
|
}
|
|
};
|
|
|
|
#ifdef _CCCL_DOXYGEN_INVOKED
|
|
/*! Write the address that a \c device_ptr points to to an output stream.
|
|
*
|
|
* \param os The output stream.
|
|
* \param dp The \c device_ptr to output.
|
|
*
|
|
* \return \c os.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename T, typename CharT, typename Traits>
|
|
_CCCL_HOST std::basic_ostream<CharT, Traits>& operator<<(std::basic_ostream<CharT, Traits>& os, device_ptr<T> const& dp);
|
|
#endif
|
|
|
|
/*! \brief Create a \c device_ptr from a raw pointer.
|
|
*
|
|
* \tparam T Any type.
|
|
* \param ptr A raw pointer to a \c T in device memory.
|
|
*
|
|
* \pre \c ptr points to a location in device memory.
|
|
*
|
|
* \return A \c device_ptr<T> pointing to \c ptr.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename T>
|
|
_CCCL_HOST_DEVICE device_ptr<T> device_pointer_cast(T* ptr);
|
|
|
|
/*! \brief Create a \c device_ptr from another \c device_ptr.
|
|
*
|
|
* \tparam T Any type.
|
|
* \param dptr A \c device_ptr to a \c T.
|
|
*
|
|
* \verbatim embed:rst:leading-asterisk
|
|
* .. versionadded:: 2.2.0
|
|
* \endverbatim
|
|
*/
|
|
template <typename T>
|
|
_CCCL_HOST_DEVICE device_ptr<T> device_pointer_cast(device_ptr<T> const& dptr);
|
|
|
|
/*! \} // memory_management
|
|
*/
|
|
|
|
THRUST_NAMESPACE_END
|
|
|
|
#include <thrust/detail/device_ptr.inl>
|
|
#include <thrust/detail/raw_pointer_cast.h>
|