open3d::core::cuda Namespace Reference

Functions
int	DeviceCount ()
bool	IsAvailable ()
void	ReleaseCache ()
	Releases CUDA memory manager cache. This is typically used for debugging. More...
void	Synchronize ()
void	Synchronize (const Device &device)
void	AssertCUDADeviceAvailable (int device_id)
void	AssertCUDADeviceAvailable (const Device &device)
bool	SupportsMemoryPools (const Device &device)

Function Documentation

AssertCUDADeviceAvailable() [1/2]

void open3d::core::cuda::AssertCUDADeviceAvailable

(

const Device &

device

)

Checks if the CUDA device-ID is available and throws error if not. The CUDA device-ID must be between 0 to device count - 1.

Parameters

device

The device to be checked.

AssertCUDADeviceAvailable() [2/2]

void open3d::core::cuda::AssertCUDADeviceAvailable

(

int

device_id

)

Checks if the CUDA device-ID is available and throws error if not. The CUDA device-ID must be between 0 to device count - 1.

Parameters

device_id

The cuda device id to be checked.

DeviceCount()

int open3d::core::cuda::DeviceCount

(

)

Returns the number of available CUDA devices. Returns 0 if Open3D is not compiled with CUDA support.

IsAvailable()

bool open3d::core::cuda::IsAvailable

(

)

Returns true if Open3D is compiled with CUDA support and at least one compatible CUDA device is detected.

ReleaseCache()

void open3d::core::cuda::ReleaseCache

(

)

Releases CUDA memory manager cache. This is typically used for debugging.

SupportsMemoryPools()

bool open3d::core::cuda::SupportsMemoryPools

(

const Device &

device

)

Checks if the CUDA device support Memory Pools used by the Stream Ordered Memory Allocator, see https://docs.nvidia.com/cuda/cuda-runtime-api/group__CUDART__MEMORY__POOLS.html

Parameters

device

The device to be checked.

Synchronize() [1/2]

void open3d::core::cuda::Synchronize

(

)

Calls cudaDeviceSynchronize() for all CUDA devices. If Open3D is not compiled with CUDA this function has no effect.

Synchronize() [2/2]

void open3d::core::cuda::Synchronize

(

const Device &

device

)

Calls cudaDeviceSynchronize() for the specified device. If Open3D is not compiled with CUDA or if device is not a CUDA device, this function has no effect.

Parameters

device

The device to be synchronized.