#include <gtest/gtest.h>
#include "../util/util.h"
#include <cppsim/state.hpp>
#include <cppsim/utility.hpp>
#include <cppsim/state_gpu.hpp>


inline void set_eigen_from_gpu(Eigen::VectorXcd& dst, QuantumStateGpu& src, ITYPE dim) {
	auto ptr = src.duplicate_data_cpp();
	for (ITYPE i = 0; i < dim; ++i) dst[i] = ptr[i];
	free(ptr);
}

inline void set_vector_from_gpu(std::vector<std::complex<double>>& dst, QuantumStateGpu& src, ITYPE dim) {
	auto ptr = src.duplicate_data_cpp();
	for (ITYPE i = 0; i < dim; ++i) dst[i] = ptr[i];
	free(ptr);
}

inline void assert_vector_eq_gpu(std::vector<std::complex<double>>& v1, QuantumStateGpu& v2, ITYPE dim, double eps) {
	auto ptr = v2.duplicate_data_cpp();
	for (UINT i = 0; i < dim; ++i) {
		ASSERT_NEAR(ptr[i].real(), v1[i].real(), eps);
		ASSERT_NEAR(ptr[i].imag(), v1[i].imag(), eps);
	}
	free(ptr);
}

inline void assert_cpu_eq_gpu(QuantumStateCpu& state_cpu, QuantumStateGpu& state_gpu, ITYPE dim, double eps) {
	auto gpu_state_vector = state_gpu.duplicate_data_cpp();
	for (ITYPE i = 0; i < dim; ++i) {
		ASSERT_NEAR(state_cpu.data_cpp()[i].real(), gpu_state_vector[i].real(), eps);
		ASSERT_NEAR(state_cpu.data_cpp()[i].imag(), gpu_state_vector[i].imag(), eps);
	}
	delete gpu_state_vector;
}

// post-selection probability check
// referred to test/csim/test_stat.cpp
TEST(StatOperationTest, ProbTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state(n, idx);
		Eigen::MatrixXcd P0(2, 2), P1(2, 2);
		P0 << 1, 0, 0, 0;
		P1 << 0, 0, 0, 1;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			state.set_Haar_random_state();
			ASSERT_NEAR(state_norm_squared_host(state.data(), dim, state.get_cuda_stream(), state.device_number), 1, eps);
			Eigen::VectorXcd test_state(dim);

			set_eigen_from_gpu(test_state, state, dim);

			for (UINT target = 0; target < n; ++target) {
				double p0 = M0_prob_host(target, state.data(), dim, state.get_cuda_stream(), state.device_number);
				double p1 = M1_prob_host(target, state.data(), dim, state.get_cuda_stream(), state.device_number);
				ASSERT_NEAR((get_expanded_eigen_matrix_with_identity(target, P0, n) * test_state).squaredNorm(), p0, eps);
				ASSERT_NEAR((get_expanded_eigen_matrix_with_identity(target, P1, n) * test_state).squaredNorm(), p1, eps);
				ASSERT_NEAR(p0 + p1, 1, eps);
			}
		}
	}
}

// marginal probability check
// referred to test/csim/test_stat.cpp
TEST(StatOperationTest, MarginalProbTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state(n, idx);
		Eigen::MatrixXcd P0(2, 2), P1(2, 2), Identity(2, 2);
		P0 << 1, 0, 0, 0;
		P1 << 0, 0, 0, 1;
		Identity << 1, 0, 0, 1;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			state.set_Haar_random_state();
			ASSERT_NEAR(state_norm_squared_host(state.data(), dim, state.get_cuda_stream(), state.device_number), 1, eps);
			Eigen::VectorXcd test_state(dim);

			set_eigen_from_gpu(test_state, state, dim);

			for (UINT target = 0; target < n; ++target) {
				// merginal probability check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				std::vector<UINT> index_list, measured_value_list;

				index_list.clear();
				measured_value_list.clear();
				for (UINT i = 0; i < n; ++i) {
					UINT measured_value = rand_int(3);
					if (measured_value != 2) {
						measured_value_list.push_back(measured_value);
						index_list.push_back(i);
					}
					if (measured_value == 0) {
						mat = kronecker_product(P0, mat);
					}
					else if (measured_value == 1) {
						mat = kronecker_product(P1, mat);
					}
					else {
						mat = kronecker_product(Identity, mat);
					}
				}
				double test_marginal_prob = (mat * test_state).squaredNorm();
				double res = marginal_prob_host(index_list.data(), measured_value_list.data(), (UINT)index_list.size(), state.data(), dim, state.get_cuda_stream(), state.device_number);
				ASSERT_NEAR(test_marginal_prob, res, eps);
			}
		}
	}
}

// entropy
TEST(StatOperationTest, EntropyTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		// CTYPE* state = allocate_quantum_state(dim);
		void* state = allocate_quantum_state_host(dim, idx);
		CTYPE* state_cpp = (CTYPE*)malloc(dim * sizeof(CTYPE));

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state, state_cpp, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state_cpp[i];

			for (UINT target = 0; target < n; ++target) {
				double ent = 0;
				for (ITYPE ind = 0; ind < dim; ++ind) {
					double prob = norm(test_state[ind]);
					if (prob > eps)
						ent += -prob * log(prob);
				}
				ASSERT_NEAR(ent, measurement_distribution_entropy_host(state, dim, stream_ptr, idx), eps);
			}
		}
		free(state_cpp);
		release_quantum_state_host(state, idx);
		release_cuda_stream_host(stream_ptr, 1, idx);
	}
}

// inner product
TEST(StatOperationTest, InnerProductTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		CTYPE* buffer = allocate_quantum_state(dim);
		void* buffer_gpu = allocate_quantum_state_host(dim, idx);
		for (UINT rep = 0; rep < max_repeat; ++rep) {

			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			for (UINT target = 0; target < n; ++target) {
				initialize_Haar_random_state_host(buffer_gpu, dim, stream_ptr, idx);
				get_quantum_state_host(buffer_gpu, buffer, dim, stream_ptr, idx);
				CPPCTYPE inp = inner_product_host(buffer_gpu, state_gpu, dim, stream_ptr, idx);
				Eigen::VectorXcd test_buffer(dim);
				for (ITYPE i = 0; i < dim; ++i) test_buffer[i] = buffer[i];
				std::complex<double> test_inp = (test_buffer.adjoint() * test_state);
				ASSERT_NEAR(inp.real(), test_inp.real(), eps);
				ASSERT_NEAR(inp.imag(), test_inp.imag(), eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
		release_quantum_state(buffer);
		release_quantum_state_host(buffer_gpu, idx);
	}
}


// single qubit expectation value
TEST(StatOperationTest, SingleQubitExpectationValueTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			for (UINT target = 0; target < n; ++target) {
				// single qubit expectation value check
				target = rand_int(n);
				UINT pauli = rand_int(3) + 1;
				if (pauli == 0) pauli_op = Identity;
				else if (pauli == 1) pauli_op = X;
				else if (pauli == 2) pauli_op = Y;
				else if (pauli == 3) pauli_op = Z;
				std::complex<double> value = (test_state.adjoint() * get_expanded_eigen_matrix_with_identity(target, pauli_op, n) * test_state);
				ASSERT_NEAR(value.imag(), 0, eps);
				double test_expectation = value.real();
				double expectation = expectation_value_single_qubit_Pauli_operator_host(pauli, target, state_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(expectation, test_expectation, eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitExpectationValueWholeTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation whole list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				std::vector<UINT> pauli_whole;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(4);
					if (pauli == 0) pauli_op = Identity;
					else if (pauli == 1) pauli_op = X;
					else if (pauli == 2) pauli_op = Y;
					else if (pauli == 3) pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					pauli_whole.push_back(pauli);
				}
				std::complex<double> value = (test_state.adjoint() * mat * test_state);
				ASSERT_NEAR(value.imag(), 0, eps);
				double test_expectation = value.real();
				double expectation = expectation_value_multi_qubit_Pauli_operator_whole_list_host(pauli_whole.data(), n, state_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(expectation, test_expectation, eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitExpectationValueZopWholeTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation whole list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				std::vector<UINT> pauli_whole;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(2);
					if (pauli == 1) pauli = 3;
					if (pauli == 0) pauli_op = Identity;
					else pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					pauli_whole.push_back(pauli);
				}
				std::complex<double> value = (test_state.adjoint() * mat * test_state);
				ASSERT_NEAR(value.imag(), 0, eps);
				double test_expectation = value.real();
				double expectation = expectation_value_multi_qubit_Pauli_operator_whole_list_host(pauli_whole.data(), n, state_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(expectation, test_expectation, eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
	}
}


// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitExpectationValuePartialTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const double eps = 1e-14;
	const UINT max_repeat = 10;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
			Identity << 1, 0, 0, 1;
			X << 0, 1, 1, 0;
			Z << 1, 0, 0, -1;
			Y << 0, -1.i, 1.i, 0;

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation partial list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				Eigen::MatrixXcd pauli_op;

				std::vector<UINT> pauli_partial, pauli_index;
				std::vector<std::pair<UINT, UINT>> pauli_partial_pair;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(4);
					if (pauli == 0) pauli_op = Identity;
					else if (pauli == 1) pauli_op = X;
					else if (pauli == 2) pauli_op = Y;
					else if (pauli == 3) pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					if (pauli != 0) {
						pauli_partial_pair.push_back(std::make_pair(i, pauli));
					}
				}
				std::random_shuffle(pauli_partial_pair.begin(), pauli_partial_pair.end());
				for (auto val : pauli_partial_pair) {
					pauli_index.push_back(val.first);
					pauli_partial.push_back(val.second);
				}
				std::complex<double> value = (test_state.adjoint() * mat * test_state);
				ASSERT_NEAR(value.imag(), 0, eps);
				double test_expectation = value.real();
				double expectation = expectation_value_multi_qubit_Pauli_operator_partial_list_host(pauli_index.data(), pauli_partial.data(), (UINT)pauli_index.size(), state_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(expectation, test_expectation, eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitExpectationValueZopPartialTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const double eps = 1e-14;
	const UINT max_repeat = 10;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		CTYPE* state = allocate_quantum_state(dim);
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_gpu, dim, stream_ptr, idx), 1, eps);
			Eigen::VectorXcd test_state(dim);
			get_quantum_state_host(state_gpu, state, dim, stream_ptr, idx);
			for (ITYPE i = 0; i < dim; ++i) test_state[i] = state[i];

			Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
			Identity << 1, 0, 0, 1;
			X << 0, 1, 1, 0;
			Z << 1, 0, 0, -1;
			Y << 0, -1.i, 1.i, 0;

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation partial list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				Eigen::MatrixXcd pauli_op;

				std::vector<UINT> pauli_partial, pauli_index;
				std::vector<std::pair<UINT, UINT>> pauli_partial_pair;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(2);
					if (pauli == 1) pauli = 3;
					if (pauli == 0) pauli_op = Identity;
					else pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					if (pauli != 0) {
						pauli_partial_pair.push_back(std::make_pair(i, pauli));
					}
				}
				std::random_shuffle(pauli_partial_pair.begin(), pauli_partial_pair.end());
				for (auto val : pauli_partial_pair) {
					pauli_index.push_back(val.first);
					pauli_partial.push_back(val.second);
				}
				std::complex<double> value = (test_state.adjoint() * mat * test_state);
				ASSERT_NEAR(value.imag(), 0, eps);
				double test_expectation = value.real();
				double expectation = expectation_value_multi_qubit_Pauli_operator_partial_list_host(pauli_index.data(), pauli_partial.data(), (UINT)pauli_index.size(), state_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(expectation, test_expectation, eps);
			}
		}
		release_quantum_state(state);
		release_quantum_state_host(state_gpu, idx);
	}
}

TEST(StatOperationTest, MultiQubitTransitionAmplitudeWholeTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		CTYPE* state_ket = allocate_quantum_state(dim);
		CTYPE* state_bra = allocate_quantum_state(dim);
		void* state_ket_gpu = allocate_quantum_state_host(dim, idx);
		void* state_bra_gpu = allocate_quantum_state_host(dim, idx);

		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_ket_gpu, dim, stream_ptr, idx);
			initialize_Haar_random_state_host(state_bra_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_ket_gpu, dim, stream_ptr, idx), 1, eps);
			ASSERT_NEAR(state_norm_squared_host(state_bra_gpu, dim, stream_ptr, idx), 1, eps);

			Eigen::VectorXcd test_state_ket(dim);
			Eigen::VectorXcd test_state_bra(dim);
			get_quantum_state_host(state_ket_gpu, state_ket, dim, stream_ptr, idx);
			get_quantum_state_host(state_bra_gpu, state_bra, dim, stream_ptr, idx);

			for (ITYPE i = 0; i < dim; ++i) test_state_ket[i] = state_ket[i];
			for (ITYPE i = 0; i < dim; ++i) test_state_bra[i] = state_bra[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation whole list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				std::vector<UINT> pauli_whole;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(4);
					if (pauli == 0) pauli_op = Identity;
					else if (pauli == 1) pauli_op = X;
					else if (pauli == 2) pauli_op = Y;
					else if (pauli == 3) pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					pauli_whole.push_back(pauli);
				}
				std::complex<double> test_transition_amplitude = (test_state_bra.adjoint() * mat * test_state_ket);
				CPPCTYPE transition_amplitude = transition_amplitude_multi_qubit_Pauli_operator_whole_list_host(pauli_whole.data(), n, state_bra_gpu, state_ket_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(transition_amplitude.real(), test_transition_amplitude.real(), eps);
				ASSERT_NEAR(transition_amplitude.imag(), test_transition_amplitude.imag(), eps);
			}
		}
		release_quantum_state(state_ket);
		release_quantum_state(state_bra);
		release_quantum_state_host(state_ket_gpu, idx);
		release_quantum_state_host(state_bra_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitTransitionAmplitudeZopWholeTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		CTYPE* state_ket = allocate_quantum_state(dim);
		CTYPE* state_bra = allocate_quantum_state(dim);
		void* state_ket_gpu = allocate_quantum_state_host(dim, idx);
		void* state_bra_gpu = allocate_quantum_state_host(dim, idx);

		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_ket_gpu, dim, stream_ptr, idx);
			initialize_Haar_random_state_host(state_bra_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_ket_gpu, dim, stream_ptr, idx), 1, eps);
			ASSERT_NEAR(state_norm_squared_host(state_bra_gpu, dim, stream_ptr, idx), 1, eps);

			Eigen::VectorXcd test_state_ket(dim);
			Eigen::VectorXcd test_state_bra(dim);
			get_quantum_state_host(state_ket_gpu, state_ket, dim, stream_ptr, idx);
			get_quantum_state_host(state_bra_gpu, state_bra, dim, stream_ptr, idx);

			for (ITYPE i = 0; i < dim; ++i) test_state_ket[i] = state_ket[i];
			for (ITYPE i = 0; i < dim; ++i) test_state_bra[i] = state_bra[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation whole list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				std::vector<UINT> pauli_whole;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(2);
					if (pauli == 1) pauli = 3;
					if (pauli == 0) pauli_op = Identity;
					else pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					pauli_whole.push_back(pauli);
				}
				std::complex<double> test_transition_amplitude = (test_state_bra.adjoint() * mat * test_state_ket);
				CPPCTYPE transition_amplitude = transition_amplitude_multi_qubit_Pauli_operator_whole_list_host(pauli_whole.data(), n, state_bra_gpu, state_ket_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(transition_amplitude.real(), test_transition_amplitude.real(), eps);
				ASSERT_NEAR(transition_amplitude.imag(), test_transition_amplitude.imag(), eps);
			}
		}
		release_quantum_state(state_ket);
		release_quantum_state(state_bra);
		release_quantum_state_host(state_ket_gpu, idx);
		release_quantum_state_host(state_bra_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitTransitionAmplitudePartialTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		CTYPE* state_ket = allocate_quantum_state(dim);
		CTYPE* state_bra = allocate_quantum_state(dim);
		void* state_ket_gpu = allocate_quantum_state_host(dim, idx);
		void* state_bra_gpu = allocate_quantum_state_host(dim, idx);

		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_ket_gpu, dim, stream_ptr, idx);
			initialize_Haar_random_state_host(state_bra_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_ket_gpu, dim, stream_ptr, idx), 1, eps);
			ASSERT_NEAR(state_norm_squared_host(state_bra_gpu, dim, stream_ptr, idx), 1, eps);

			Eigen::VectorXcd test_state_ket(dim);
			Eigen::VectorXcd test_state_bra(dim);
			get_quantum_state_host(state_ket_gpu, state_ket, dim, stream_ptr, idx);
			get_quantum_state_host(state_bra_gpu, state_bra, dim, stream_ptr, idx);

			for (ITYPE i = 0; i < dim; ++i) test_state_ket[i] = state_ket[i];
			for (ITYPE i = 0; i < dim; ++i) test_state_bra[i] = state_bra[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation partial list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				Eigen::MatrixXcd pauli_op;

				std::vector<UINT> pauli_partial, pauli_index;
				std::vector<std::pair<UINT, UINT>> pauli_partial_pair;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(4);
					if (pauli == 0) pauli_op = Identity;
					else if (pauli == 1) pauli_op = X;
					else if (pauli == 2) pauli_op = Y;
					else if (pauli == 3) pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					if (pauli != 0) {
						pauli_partial_pair.push_back(std::make_pair(i, pauli));
					}
				}
				std::random_shuffle(pauli_partial_pair.begin(), pauli_partial_pair.end());
				for (auto val : pauli_partial_pair) {
					pauli_index.push_back(val.first);
					pauli_partial.push_back(val.second);
				}
				std::complex<double> test_transition_amplitude = (test_state_bra.adjoint() * mat * test_state_ket);
				CPPCTYPE transition_amplitude = transition_amplitude_multi_qubit_Pauli_operator_partial_list_host(pauli_index.data(), pauli_partial.data(), (UINT)pauli_index.size(), state_bra_gpu, state_ket_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(transition_amplitude.real(), test_transition_amplitude.real(), eps);
				ASSERT_NEAR(transition_amplitude.imag(), test_transition_amplitude.imag(), eps);
			}
		}
		release_quantum_state(state_ket);
		release_quantum_state(state_bra);
		release_quantum_state_host(state_ket_gpu, idx);
		release_quantum_state_host(state_bra_gpu, idx);
	}
}

// multi qubit expectation value whole
TEST(StatOperationTest, MultiQubitTransitionAmplitudeZopPartialTest) {
	const UINT n = 6;
	const ITYPE dim = 1ULL << n;
	const UINT max_repeat = 10;
	const double eps = 1e-14;

	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		CTYPE* state_ket = allocate_quantum_state(dim);
		CTYPE* state_bra = allocate_quantum_state(dim);
		void* state_ket_gpu = allocate_quantum_state_host(dim, idx);
		void* state_bra_gpu = allocate_quantum_state_host(dim, idx);

		Eigen::MatrixXcd Identity(2, 2), X(2, 2), Y(2, 2), Z(2, 2);
		Eigen::MatrixXcd pauli_op;
		Identity << 1, 0, 0, 1;
		X << 0, 1, 1, 0;
		Z << 1, 0, 0, -1;
		Y << 0, -1.i, 1.i, 0;

		for (UINT rep = 0; rep < max_repeat; ++rep) {
			initialize_Haar_random_state_host(state_ket_gpu, dim, stream_ptr, idx);
			initialize_Haar_random_state_host(state_bra_gpu, dim, stream_ptr, idx);
			ASSERT_NEAR(state_norm_squared_host(state_ket_gpu, dim, stream_ptr, idx), 1, eps);
			ASSERT_NEAR(state_norm_squared_host(state_bra_gpu, dim, stream_ptr, idx), 1, eps);

			Eigen::VectorXcd test_state_ket(dim);
			Eigen::VectorXcd test_state_bra(dim);
			get_quantum_state_host(state_ket_gpu, state_ket, dim, stream_ptr, idx);
			get_quantum_state_host(state_bra_gpu, state_bra, dim, stream_ptr, idx);

			for (ITYPE i = 0; i < dim; ++i) test_state_ket[i] = state_ket[i];
			for (ITYPE i = 0; i < dim; ++i) test_state_bra[i] = state_bra[i];

			for (UINT target = 0; target < n; ++target) {
				// multi qubit expectation partial list value check
				Eigen::MatrixXcd mat = Eigen::MatrixXcd::Identity(1, 1);
				Eigen::MatrixXcd pauli_op;

				std::vector<UINT> pauli_partial, pauli_index;
				std::vector<std::pair<UINT, UINT>> pauli_partial_pair;
				for (UINT i = 0; i < n; ++i) {
					UINT pauli = rand_int(2);
					if (pauli == 1) pauli = 3;
					if (pauli == 0) pauli_op = Identity;
					else pauli_op = Z;
					mat = kronecker_product(pauli_op, mat);
					if (pauli != 0) {
						pauli_partial_pair.push_back(std::make_pair(i, pauli));
					}
				}
				std::random_shuffle(pauli_partial_pair.begin(), pauli_partial_pair.end());
				for (auto val : pauli_partial_pair) {
					pauli_index.push_back(val.first);
					pauli_partial.push_back(val.second);
				}
				std::complex<double> test_transition_amplitude = (test_state_bra.adjoint() * mat * test_state_ket);
				CPPCTYPE transition_amplitude = transition_amplitude_multi_qubit_Pauli_operator_partial_list_host(pauli_index.data(), pauli_partial.data(), (UINT)pauli_index.size(), state_bra_gpu, state_ket_gpu, dim, stream_ptr, idx);
				ASSERT_NEAR(transition_amplitude.real(), test_transition_amplitude.real(), eps);
				ASSERT_NEAR(transition_amplitude.imag(), test_transition_amplitude.imag(), eps);
			}
		}
		release_quantum_state(state_ket);
		release_quantum_state(state_bra);
		release_quantum_state_host(state_ket_gpu, idx);
		release_quantum_state_host(state_bra_gpu, idx);
	}
}

TEST(StateTest, GenerateAndRelease) {    
    UINT n = 10;
    double eps = 1e-14;
    QuantumState state(n);
    ASSERT_EQ(state.qubit_count, n);
    ASSERT_EQ(state.dim, 1ULL << n);
    state.set_zero_state();

	auto ptr = state.duplicate_data_cpp();
	for (UINT i = 0; i < state.dim; ++i) {
        if (i == 0) ASSERT_NEAR(abs(ptr[i] - 1.), 0, eps);
        else ASSERT_NEAR(abs(ptr[i]), 0, eps);
    }
	delete ptr;

	Random random;
    for (UINT repeat = 0; repeat < 10; ++repeat) {
        ITYPE basis = random.int64()%state.dim;
        state.set_computational_basis(basis);
		auto ptr = state.duplicate_data_cpp();
		for (UINT i = 0; i < state.dim; ++i) {
			if (i == basis) ASSERT_NEAR(abs(ptr[i] - 1.), 0, eps);
            else ASSERT_NEAR(abs(ptr[i]), 0, eps);
        }
		delete ptr;
	}
    for (UINT repeat = 0; repeat < 10; ++repeat) {
        state.set_Haar_random_state();
        ASSERT_NEAR(state.get_squared_norm(),1.,eps);
    }
}

TEST(StateTest, Sampling) {
    UINT n = 10;
    QuantumState state(n);
    state.set_Haar_random_state();
    state.set_computational_basis(100);
    auto res1 = state.sampling(1024);
    state.set_computational_basis(100);
    auto res2 = state.sampling(1024);
}


TEST(StateTest, SetState) {
	const double eps = 1e-10;
	const UINT n = 10;
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state(n, idx);
		const ITYPE dim = 1ULL << n;
		std::vector<std::complex<double>> state_vector(dim);
		for (UINT i = 0; i < dim; ++i) {
			double d = (double)i;
			state_vector[i] = d + std::complex<double>(0, 1) * (d + 0.1);
		}
		state.load(state_vector);

		assert_vector_eq_gpu(state_vector, state, dim, eps);
	}
}

TEST(StateTest, CopyState) {
	const double eps = 1e-10;
	const UINT n = 7;
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);
		QuantumState state(n);
		state.set_Haar_random_state();
		const ITYPE dim = 1ULL << n;
		void* state_gpu = allocate_quantum_state_host(dim, idx);
		copy_quantum_state_from_host_to_device(state_gpu, state.data(), dim, stream_ptr, idx);
		CPPCTYPE* state_cpu_copy = (CPPCTYPE*)malloc(sizeof(CPPCTYPE) * dim);
		get_quantum_state_host(state_gpu, state_cpu_copy, dim, stream_ptr, idx);

		for (UINT i = 0; i < dim; ++i) {
			ASSERT_NEAR(state.data_cpp()[i].real(), state_cpu_copy[i].real(), eps);
			ASSERT_NEAR(state.data_cpp()[i].imag(), state_cpu_copy[i].imag(), eps);
		}
		release_quantum_state_host(state_gpu, idx);
	}
}

TEST(StateTest, CopyStateDevice) {
	const double eps = 1e-10;
	const UINT n = 8;
	const ITYPE dim = 1ULL << n;
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		set_device(idx);
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		void* state1_gpu = allocate_quantum_state_host(dim, idx);
		void* state2_gpu = allocate_quantum_state_host(dim, idx);

		initialize_Haar_random_state_host(state1_gpu, dim, stream_ptr, idx);

		CPPCTYPE* state1_cpu_copy = (CPPCTYPE*)malloc(sizeof(CPPCTYPE) * dim);
		CPPCTYPE* state2_cpu_copy = (CPPCTYPE*)malloc(sizeof(CPPCTYPE) * dim);
		get_quantum_state_host(state1_gpu, state1_cpu_copy, dim, stream_ptr, idx);
		copy_quantum_state_from_device_to_device(state2_gpu, state1_gpu, dim, stream_ptr, idx);
		get_quantum_state_host(state2_gpu, state2_cpu_copy, dim, stream_ptr, idx);

		for (UINT i = 0; i < dim; ++i) {
			ASSERT_NEAR(state1_cpu_copy[i].real(), state2_cpu_copy[i].real(), eps);
			ASSERT_NEAR(state1_cpu_copy[i].imag(), state2_cpu_copy[i].imag(), eps);
		}
		release_quantum_state_host(state1_gpu, idx);
		release_quantum_state_host(state2_gpu, idx);
	}
}

TEST(StateTest, AddState) {
	const double eps = 1e-10;
	const UINT n = 10;
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		auto stream_ptr = allocate_cuda_stream_host(1, idx);

		QuantumStateGpu state1(n, idx);
		QuantumStateGpu state2(n, idx);
		state1.set_Haar_random_state();
		state2.set_Haar_random_state();

		const ITYPE dim = 1ULL << n;
		std::vector<std::complex<double>> state_vector1(dim);
		std::vector<std::complex<double>> state_vector2(dim);

		set_vector_from_gpu(state_vector1, state1, dim);
		set_vector_from_gpu(state_vector2, state2, dim);

		state1.add_state(&state2);
		for (ITYPE i = 0; i < dim; ++i) {
			state_vector1[i] += state_vector2[i];
		}
		assert_vector_eq_gpu(state_vector1, state1, dim, eps);
		assert_vector_eq_gpu(state_vector2, state2, dim, eps);
	}
}

TEST(StateTest, MultiplyCoef) {
	const double eps = 1e-10;
	const UINT n = 10;
	const std::complex<double> coef(0.5, 0.2);
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state(n, idx);
		state.set_Haar_random_state();

		const ITYPE dim = 1ULL << n;
		std::vector<std::complex<double>> state_vector(dim);

		auto ptr = state.duplicate_data_cpp();
		for (ITYPE i = 0; i < dim; ++i) {
			state_vector[i] = ptr[i] * coef;
		}
		delete ptr;
		state.multiply_coef(coef);

		assert_vector_eq_gpu(state_vector, state, dim, eps);
	}
}


TEST(StateTest, StateCPUtoGPU) {
	const double eps = 1e-10;
	const UINT n = 10;
	const ITYPE dim = 1ULL << n;
	const std::complex<double> coef(0.5, 0.2);
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state_gpu(n, idx);
		QuantumState state_cpu(n);
		state_cpu.set_Haar_random_state();
		state_gpu.load(&state_cpu);

		assert_cpu_eq_gpu(state_cpu, state_gpu, dim, eps);
	}
}

TEST(StateTest, StateGPUtoCPU) {
	const double eps = 1e-10;
	const UINT n = 10;
	const ITYPE dim = 1ULL << n;
	const std::complex<double> coef(0.5, 0.2);
	int ngpus = get_num_device();
	for (int idx = 0; idx < ngpus; ++idx) {
		QuantumStateGpu state_gpu(n, idx);
		QuantumState state_cpu(n);
		state_gpu.set_Haar_random_state();
		state_cpu.load(&state_gpu);

		assert_cpu_eq_gpu(state_cpu, state_gpu, dim, eps);
	}
}