test_thrust.cpp
53b9b1ac6cc4master

test_thrust.cpp
View Options

			#include "loop.hh"
			#include "grid.hh"
			#include <thrust/for_each.h>
			#include <thrust/iterator/zip_iterator.h>
			#include <algorithm>

			using namespace tamaas;

			template <typename Functor>
			class ApplyFunctor {
			public:
			ApplyFunctor(const Functor & functor):functor(functor) {}

			template <typename Tuple>
			__host__ __device__
			void operator()(Tuple&& t) {
			thrust::system::cuda::detail::bulk_::detail::apply_from_tuple(functor, t);
			}

			private:
			const Functor & functor;
			};

			template <typename Functor, typename... Containers>
			void loop(Functor&& func, Containers&&... containers) {
			auto begin = thrust::make_zip_iterator(thrust::make_tuple(containers.begin()...));
			auto end = thrust::make_zip_iterator(thrust::make_tuple(containers.end()...));

			thrust::for_each(begin, end, ApplyFunctor<Functor>(func));
			}


			int main() {
			Grid<Real, 1> grid({20}, 2);
			Grid<Real, 1> other({20}, 2);
			std::iota(grid.begin(), grid.end(), 0);
			other = 2;

			loop([] CUDA_LAMBDA (Real & x, Real & y) { x += y; }, grid, other);
			cudaDeviceSynchronize();
			std::cout << grid << std::endl;
			return 0;
			}