inline typename std::enable_if<
std::is_convertible<
typename std::iterator_traits<InputIterator>::iterator_category
, host_random_access_iterator_tag
>::value
&& std::is_convertible<
typename std::iterator_traits<OutputIterator>::iterator_category
, host_random_access_iterator_tag
>::value
&& std::is_same<
typename std::iterator_traits<InputIterator>::value_type
, typename std::iterator_traits<OutputIterator>::value_type
>::value
, OutputIterator>::type copy(InputIterator first, InputIterator last, OutputIterator result, stream& stream)
{
typename std::iterator_traits<InputIterator>::difference_type size = last - first;
CUDA_CALL( cudaMemcpyAsync(
&*result
, &*first
, size * sizeof(typename std::iterator_traits<InputIterator>::value_type)
, cudaMemcpyHostToHost
, stream.data()
) );
return result + size;
}
void copy(vector<T> const& src, symbol<T[]>& dst, stream& stream)
{
assert(src.size() == dst.size());
CUDA_CALL(cudaMemcpyToSymbolAsync(reinterpret_cast<char const*>(dst.data()), src.data(), src.size() * sizeof(T), 0, cudaMemcpyDeviceToDevice, stream.data()));
}
void copy(vector<T> const& src, host::vector<T>& dst, size_t size, stream& stream)
{
assert(size <= src.capacity());
assert(size <= dst.capacity());
CUDA_CALL(cudaMemcpyAsync(dst.data(), src.data(), size * sizeof(T), cudaMemcpyDeviceToHost, stream.data()));
}
void copy(symbol<T[]> const& src, host::vector<T>& dst, stream& stream)
{
assert(src.size() == dst.size());
CUDA_CALL(cudaMemcpyFromSymbolAsync(dst.data(), reinterpret_cast<char const*>(src.data()), src.size() * sizeof(T), 0, cudaMemcpyDeviceToHost, stream.data()));
}
void copy(vector<T> const& src, host::vector<T>& dst, stream& stream)
{
assert(src.size() == dst.size());
CUDA_CALL(cudaMemcpyAsync(dst.data(), src.data(), src.size() * sizeof(T), cudaMemcpyDeviceToHost, stream.data()));
}