Program Listing for File state_gpu.hpp

↰ Return to documentation for file (/home/docs/checkouts/readthedocs.org/user_builds/qulacs-rtd/checkouts/latest/src/cppsim/state_gpu.hpp)

#pragma once

#include "exception.hpp"
#include "state.hpp"

#ifdef _USE_GPU

#include <gpusim/memory_ops.h>
#include <gpusim/stat_ops.h>
#include <gpusim/update_ops_cuda.h>
#include <gpusim/util_func.h>

class QuantumStateGpu : public QuantumStateBase {
private:
    void* _state_vector;  // void* is assumed as GTYPE*
    Random random;

public:
    explicit QuantumStateGpu(UINT qubit_count_)
        : QuantumStateBase(qubit_count_, true, 0) {
        set_device(0);
        this->_cuda_stream = allocate_cuda_stream_host(1, 0);
        this->_state_vector =
            reinterpret_cast<void*>(allocate_quantum_state_host(this->_dim, 0));
        initialize_quantum_state_host(
            this->data(), _dim, _cuda_stream, device_number);
    }

    QuantumStateGpu(UINT qubit_count_, UINT device_number_)
        : QuantumStateBase(qubit_count_, true, device_number_) {
        int num_device = get_num_device();
        assert(device_number_ < num_device);
        set_device(device_number_);
        this->_cuda_stream = allocate_cuda_stream_host(1, device_number_);
        this->_state_vector = reinterpret_cast<void*>(
            allocate_quantum_state_host(this->_dim, device_number_));
        initialize_quantum_state_host(
            this->data(), _dim, _cuda_stream, device_number_);
    }

    virtual ~QuantumStateGpu() {
        release_quantum_state_host(this->data(), device_number);
        release_cuda_stream_host(this->_cuda_stream, 1, device_number);
    }
    virtual void set_zero_state() override {
        initialize_quantum_state_host(
            this->data(), _dim, _cuda_stream, device_number);
    }
    virtual void set_zero_norm_state() override {
        throw NotImplementedException(
            "set_zero_norm_state for QuantumStateGpu is not implemented "
            "yet");
    }
    virtual void set_computational_basis(ITYPE comp_basis) override {
        set_computational_basis_host(
            comp_basis, _state_vector, _dim, _cuda_stream, device_number);
    }
    virtual void set_Haar_random_state() override {
        initialize_Haar_random_state_with_seed_host(
            this->data(), _dim, random.int32(), _cuda_stream, device_number);
    }
    virtual void set_Haar_random_state(UINT seed) override {
        initialize_Haar_random_state_with_seed_host(
            this->data(), _dim, seed, _cuda_stream, device_number);
    }
    virtual double get_zero_probability(
        UINT target_qubit_index) const override {
        return M0_prob_host(target_qubit_index, this->data(), _dim,
            _cuda_stream, device_number);
    }
    virtual double get_marginal_probability(
        std::vector<UINT> measured_values) const override {
        std::vector<UINT> target_index;
        std::vector<UINT> target_value;
        for (UINT i = 0; i < measured_values.size(); ++i) {
            UINT measured_value = measured_values[i];
            if (measured_value == 0 || measured_value == 1) {
                target_index.push_back(i);
                target_value.push_back(measured_value);
            }
        }
        return marginal_prob_host(target_index.data(), target_value.data(),
            (UINT)target_index.size(), this->data(), _dim, _cuda_stream,
            device_number);
    }
    virtual double get_entropy() const override {
        return measurement_distribution_entropy_host(
            this->data(), _dim, _cuda_stream, device_number);
    }

    virtual double get_squared_norm() const override {
        return state_norm_squared_host(
            this->data(), _dim, _cuda_stream, device_number);
    }

    virtual double get_squared_norm_single_thread() const override {
        return state_norm_squared_host(
            this->data(), _dim, _cuda_stream, device_number);
    }

    virtual void normalize(double squared_norm) override {
        normalize_host(
            squared_norm, this->data(), _dim, _cuda_stream, device_number);
    }

    virtual void normalize_single_thread(double squared_norm) override {
        normalize_host(
            squared_norm, this->data(), _dim, _cuda_stream, device_number);
    }

    virtual QuantumStateBase* allocate_buffer() const override {
        QuantumStateGpu* new_state =
            new QuantumStateGpu(this->_qubit_count, device_number);
        return new_state;
    }
    virtual QuantumStateGpu* copy() const override {
        QuantumStateGpu* new_state =
            new QuantumStateGpu(this->_qubit_count, device_number);
        copy_quantum_state_from_device_to_device(new_state->data(),
            _state_vector, _dim, _cuda_stream, device_number);
        for (UINT i = 0; i < _classical_register.size(); ++i)
            new_state->set_classical_value(i, _classical_register[i]);
        return new_state;
    }
    virtual void load(const QuantumStateBase* _state) override {
        if (!_state->is_state_vector()) {
            throw InoperatableQuantumStateTypeException(
                "Error: QuantumStateGpu::load(const QuantumStateBase*): "
                "cannot load DensityMatrix to StateVector");
        }
        if (_state->get_device_name() == "gpu") {
            copy_quantum_state_from_device_to_device(
                this->data(), _state->data(), dim, _cuda_stream, device_number);
        } else {
            this->load(_state->data_cpp());
        }
        this->_classical_register = _state->classical_register;
    }

    virtual void load(const std::vector<CPPCTYPE>& _state) override {
        copy_quantum_state_from_cppstate_host(
            this->data(), _state.data(), dim, _cuda_stream, device_number);
    }

    virtual void load(const CPPCTYPE* _state) override {
        copy_quantum_state_from_cppstate_host(
            this->data(), _state, dim, _cuda_stream, device_number);
    }
    virtual const std::string get_device_name() const override { return "gpu"; }

    virtual CPPCTYPE* data_cpp() const override {
        throw NotImplementedException(
            "Cannot reinterpret state vector on GPU to cpp complex "
            "vector. Use duplicate_data_cpp instead.");
    }

    virtual CTYPE* data_c() const override {
        throw NotImplementedException(
            "Cannot reinterpret state vector on GPU to C complex vector. "
            "Use duplicate_data_cpp instead.");
    }

    virtual void* data() const override {
        return reinterpret_cast<void*>(this->_state_vector);
    }

    virtual CTYPE* duplicate_data_c() const override {
        CTYPE* _copy_state = (CTYPE*)malloc(sizeof(CTYPE) * dim);
        get_quantum_state_host(
            this->_state_vector, _copy_state, dim, _cuda_stream, device_number);
        return _copy_state;
    }

    virtual CPPCTYPE* duplicate_data_cpp() const override {
        CPPCTYPE* _copy_state = (CPPCTYPE*)malloc(sizeof(CPPCTYPE) * dim);
        get_quantum_state_host(
            this->_state_vector, _copy_state, dim, _cuda_stream, device_number);
        return _copy_state;
    }

    virtual void add_state(const QuantumStateBase* state) override {
        if (!state->is_state_vector()) {
            throw InoperatableQuantumStateTypeException(
                "Error: QuantumStateGpu::add_state(const QuantumStateBase*): "
                "cannot add DensityMatrix to StateVector");
        }
        state_add_host(state->data(), this->data(), this->dim, _cuda_stream,
            device_number);
    }

    virtual void add_state_with_coef(
        CPPCTYPE coef, const QuantumStateBase* state) override {
        if (!state->is_state_vector()) {
            throw InoperatableQuantumStateTypeException(
                "Error: QuantumStateGpu::add_state_with_coef(CPPCTYPE, "
                "const QuantumStateBase*): "
                "cannot add DensityMatrix to StateVector");
        }
        state_multiply_host(
            coef, this->data(), this->dim, _cuda_stream, device_number);
        state_add_host(state->data(), this->data(), this->dim, _cuda_stream,
            device_number);
        state_multiply_host(CPPCTYPE(1) / coef, this->data(), this->dim,
            _cuda_stream, device_number);
    }

    virtual void add_state_with_coef_single_thread(
        CPPCTYPE coef, const QuantumStateBase* state) override {
        if (!state->is_state_vector()) {
            throw InoperatableQuantumStateTypeException(
                "Error: "
                "QuantumStateGpu::add_state_with_coef_single_thread(CPPCTYPE, "
                "const QuantumStateBase*): "
                "cannot add DensityMatrix to StateVector");
        }
        state_multiply_host(CPPCTYPE(1) / coef, this->data(), this->dim,
            _cuda_stream, device_number);
        state_add_host(state->data(), this->data(), this->dim, _cuda_stream,
            device_number);
        state_multiply_host(
            coef, this->data(), this->dim, _cuda_stream, device_number);
    }

    virtual void multiply_coef(CPPCTYPE coef) override {
        state_multiply_host(
            coef, this->data(), this->dim, _cuda_stream, device_number);
    }

    virtual void multiply_elementwise_function(
        const std::function<CPPCTYPE(ITYPE)>& func) override {
        std::vector<CPPCTYPE> diagonal_matrix(dim);
        for (ITYPE i = 0; i < dim; ++i) {
            diagonal_matrix[i] = func(i);
        }
        multi_qubit_diagonal_matrix_gate_host(diagonal_matrix.data(),
            this->data(), dim, _cuda_stream, device_number);
    }

    virtual std::vector<ITYPE> sampling(UINT sampling_count) override {
        std::vector<double> stacked_prob;
        std::vector<ITYPE> result;
        double sum = 0.;
        auto ptr = this->duplicate_data_cpp();
        stacked_prob.push_back(0.);
        for (UINT i = 0; i < this->dim; ++i) {
            sum += norm(ptr[i]);
            stacked_prob.push_back(sum);
        }

        for (UINT count = 0; count < sampling_count; ++count) {
            double r = random.uniform();
            auto ite =
                std::lower_bound(stacked_prob.begin(), stacked_prob.end(), r);
            auto index = std::distance(stacked_prob.begin(), ite) - 1;
            result.push_back(index);
        }
        free(ptr);
        return result;
    }

    virtual std::vector<ITYPE> sampling(
        UINT sampling_count, UINT random_seed) override {
        random.set_seed(random_seed);
        return this->sampling(sampling_count);
    }

    virtual std::string to_string() const override {
        std::stringstream os;
        ComplexVector eigen_state(this->dim);
        auto data = this->duplicate_data_cpp();
        for (UINT i = 0; i < this->dim; ++i) eigen_state[i] = data[i];
        os << " *** Quantum State ***" << std::endl;
        os << " * Qubit Count : " << this->qubit_count << std::endl;
        os << " * Dimension   : " << this->dim << std::endl;
        os << " * State vector : \n" << eigen_state << std::endl;
        free(data);
        return os.str();
    }

    virtual boost::property_tree::ptree to_ptree() const override {
        throw NotImplementedException(
            "to_ptree for QuantumStateGpu is not implemented "
            "yet");
    }
};

namespace state {
CPPCTYPE DllExport inner_product(
    const QuantumStateGpu* state_bra, const QuantumStateGpu* state_ket);
}  // namespace state

#endif  // _USE_GPU