Highly Efficient FFT for Exascale: HeFFTe v2.4
|
The CUDA backend uses a CUDA stream. More...
#include <heffte_backend_cuda.h>
Public Types | |
using | stream_type = cudaStream_t |
The type for the internal stream. | |
using | stream_type = std::reference_wrapper<sycl::queue> |
The type for the internal stream. | |
using | stream_type = hipStream_t |
The type for the internal stream. | |
Public Member Functions | |
device_instance (cudaStream_t new_stream=nullptr) | |
Constructor, sets up the stream. | |
cudaStream_t | stream () |
Returns the nullptr. | |
cudaStream_t | stream () const |
Returns the nullptr (const case). | |
void | synchronize_device () const |
Syncs the execution with the queue, no-op in the CPU case. | |
device_instance () | |
Empty constructor. | |
device_instance (sycl::queue &new_stream) | |
Constructor assigning the queue. | |
device_instance (std::reference_wrapper< sycl::queue > &new_stream) | |
Constructor assigning from an existing wrapper. | |
sycl::queue & | stream () |
Returns the nullptr. | |
sycl::queue & | stream () const |
Returns the nullptr. | |
void | synchronize_device () const |
Syncs the execution with the queue. | |
device_instance (hipStream_t new_stream=nullptr) | |
Constructor, sets up the stream. | |
hipStream_t | stream () |
Returns the nullptr. | |
hipStream_t | stream () const |
Returns the nullptr (const case). | |
void | synchronize_device () const |
Syncs the execution with the queue. | |
Public Attributes | |
cudaStream_t | _stream |
The CUDA stream to be used in all operations. | |
std::reference_wrapper< sycl::queue > | _stream |
The sycl::queue, either user provided or created by heFFTe. | |
hipStream_t | _stream |
The CUDA stream to be used in all operations. | |
The CUDA backend uses a CUDA stream.
The ROCm backend uses a HIP stream.
Specialization that contains the sycl::queue needed for the DPC++ backend.