Template Class device_matrix

Defined in File device_matrix.h

Inheritance Relationships

Base Type

private boost::noncopyable

Class Documentation

template<typename T> class device_matrix : private boost::noncopyable

A cuda device buffer object that represents a global memory buffer on a cuda device. Matrix size is fixed at construction and can only be resized using move semantics.

Note

See https://www.quantstart.com/articles/Matrix-Matrix-Multiplication-on-the-GPU-with-Nvidia-CUDA/

Note

See https://forums.developer.nvidia.com/t/guide-cudamalloc3d-and-cudaarrays/23421

Template Parameters:: T – numeric type

Public Functions

inline device_matrix(): Default constructor.

inline device_matrix(device_matrix &&other) noexcept

Move Constructor.

Parameters:: other –

inline device_matrix &operator=(device_matrix &&other) noexcept

Move Assignment Operator.

Parameters:: other –
Returns:: device_matrix&

inline device_matrix(size_t rows, size_t cols, const T *data = nullptr)

Construct a new device matrix object of fixed size and initialized asyncronously if data is provided.

Parameters:

rows – number of rows
cols – number of columns
data – constigous column major data of size rows*cols*sizeof(T) to copy to device

inline explicit device_matrix(const Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic> &data)

template<int Rows, int Cols> inline explicit device_matrix(const Eigen::Matrix<T, Rows, Cols> &data)

inline ~device_matrix()

__host__ inline T *Get()

__host__ inline const T *Get() const

__host__ inline size_t GetRows() const

__host__ inline size_t GetCols() const

__host__ inline size_t GetCount() const

__host__ inline size_t GetSize() const

__host__ inline void SetZeroAsync()

__host__ inline void SetDataSync(const T *data)

Performs a synchronous copy of data into the device buffer.

Parameters:: data –
Returns:: host

__host__ inline void SetDataAsync(const T *data)

Copies data from device to host memory.

Parameters:: data –
Returns:: host

__host__ inline void ToHost(T *out) const

__host__ inline void ToHost(std::vector<T> &out) const

__host__ inline void ToHost(Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic> &out) const

template<int Rows, int Cols> __host__ inline void ToHost(Eigen::Matrix<T, Rows, Cols> &out) const

__host__ inline Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic> ToHost() const

__host__ inline void ToHostAsync(T *out) const

__host__ inline void ToHostAsync(Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic> &out) const

__host__ inline void ToHostVectorAsync(Eigen::Matrix<T, Eigen::Dynamic, 1> &out) const

__host__ inline Eigen::Matrix<T, Eigen::Dynamic, Eigen::Dynamic> ToHostAsync() const