DeviceBufferBase

Fully qualified name: isaacsim::core::includes::DeviceBufferBase

template<typename T>
class DeviceBufferBase : public isaacsim::core::includes::Buffer<T>

CUDA device (GPU) memory buffer implementation.

Manages a buffer of memory allocated on a CUDA device. Provides functionality for:

• Memory allocation and deallocation

• Device selection and switching

• Memory copying between host and device

• Debug printing of buffer contents

Note

Uses RAII principles for automatic resource management

Warning

Requires proper CUDA environment setup

Template Parameters:

T – The data type stored in the buffer

Public Functions

inline DeviceBufferBase(const size_t &size = 0, const int device = -1)

Constructs a new device buffer.

Parameters:

• size – [in] Initial size of the buffer in elements (default: 0)

• device – [in] CUDA device ID to allocate on (default: -1 for CPU)

inline virtual ~DeviceBufferBase()

Destructor that ensures proper cleanup of device memory.

inline virtual void setDevice(const int device = -1)

Changes the CUDA device for this buffer.

If the device changes, existing memory is freed on the old device and reallocated on the new device.

Parameters:

device – [in] New CUDA device ID (-1 for CPU)

inline virtual void resize(size_t size)

Resizes the device buffer.

Reallocates memory if the new size is different from the current size. Handles deallocation of existing memory if necessary.

Parameters:

size – [in] New size in number of elements

inline virtual void resizeAsync(

size_t size,

cudaStream_t cudaStream = 0,

)

Asynchronously resizes the device buffer.

Reallocates memory if the new size is different from the current size. Handles deallocation of existing memory if necessary.

Parameters:

• size – [in] New size in number of elements

• cudaStream – [in] CUDA stream to use for asynchronous operation

inline virtual T *data() const

Gets a pointer to the device memory.

Returns:

Raw pointer to the device memory

inline virtual size_t size() const

Gets the current size of the buffer.

Returns:

Number of elements in the buffer

inline virtual void copy(

const void *src,

size_t size,

enum cudaMemcpyKind kind = cudaMemcpyDeviceToHost,

)

Synchronously copies data to the device buffer.

Parameters:

• src – [in] Source pointer to copy from

• size – [in] Number of elements to copy

• kind – [in] Type of memory copy operation

inline virtual void copyAsync(

const void *src,

size_t size,

enum cudaMemcpyKind kind = cudaMemcpyDeviceToHost,

cudaStream_t cudaStream = 0,

)

Asynchronously copies data to the device buffer.

Parameters:

• src – [in] Source pointer to copy from

• size – [in] Number of elements to copy

• kind – [in] Type of memory copy operation

• cudaStream – [in] CUDA stream to use for the copy operation

inline void debugPrint(

const std::string &start,

const std::string &end,

)

Prints the buffer contents for debugging.

Parameters:

• start – [in] String to print before the buffer contents

• end – [in] String to print after the buffer contents

inline size_t sizeofType() const

Gets the size of a single element in bytes.

Returns:

Size of type T in bytes

inline size_t sizeInBytes() const

Gets the total size of the buffer in bytes.

Returns:

Total size of the buffer in bytes

inline MemoryType type() const

Gets the memory type of the buffer.

Returns:

Memory type (Host or Device)

Protected Attributes

MemoryType m_memoryType

Type of memory where the buffer resides.