ema::cuda::CUDADeviceManager Class Reference

Public Member Functions

int get_device_count () const
 Returns the number of available CUDA capable compute devices.
void sync_stream (cudaStream_t stream)
 perform synchronization on a stream
int get_best_device () const
 return sthe best possible device from multiple GPUs
void get_memory_statistics (CUDAMemStats &cms, const int K=0) const
 Returns the memory structure associated with device K.
void update_memory_statistics ()
 Updates the 'device_stats' structures with the most current memory usage statistics Please note that it is possible memory can be allocated from other sources, this is simply to provide a metric for checking relative memory availability.
void print_device_info () const
 output a list of all available CUDA hardware and compute capabilities

Static Public Member Functions

static CUDADeviceManagerself ()
 Return reference to the device manager object.

Detailed Description

Definition at line 77 of file CudaMemoryAllocator.h.

The documentation for this class was generated from the following file: