allocate_free.hpp

Go to the documentation of this file.

#ifndef PARALUTION_UTILS_ALLOCATE_FREE_HPP_
#define PARALUTION_UTILS_ALLOCATE_FREE_HPP_

#include <iostream>

// When CUDA backend is available the host memory allocation
// can use cudaMallocHost() function for pinned memory
// thus the memory transfers to the GPU are faster
// and this also enables the async transfers.
// Uncomment to use pinned memory
//
// To force standard CPU allocation comment the following line 
// #define PARALUTION_CUDA_PINNED_MEMORY

namespace paralution {

template <typename DataType>
void allocate_host(const int size, DataType **ptr);

template <typename DataType>
void free_host(DataType **ptr);

template <typename DataType>
void set_to_zero_host(const int size, DataType *ptr);

}

#endif // PARALUTION_UTILS_ALLOCATE_FREE_HPP_