sellp.hpp

// SPDX-FileCopyrightText: 2017 - 2024 The Ginkgo authors
//
// SPDX-License-Identifier: BSD-3-Clause
 
#ifndef GKO_PUBLIC_CORE_MATRIX_SELLP_HPP_
#define GKO_PUBLIC_CORE_MATRIX_SELLP_HPP_
 
 
#include <ginkgo/core/base/array.hpp>
#include <ginkgo/core/base/lin_op.hpp>
 
 
namespace gko {
namespace matrix {
 
 
constexpr int default_slice_size = 64;
constexpr int default_stride_factor = 1;
 
 
template <typename ValueType>
class Dense;
 
template <typename ValueType, typename IndexType>
class Csr;
 
template <typename ValueType = default_precision, typename IndexType = int32>
class Sellp : public EnableLinOp<Sellp<ValueType, IndexType>>,
              public ConvertibleTo<Sellp<next_precision<ValueType>, IndexType>>,
              public ConvertibleTo<Dense<ValueType>>,
              public ConvertibleTo<Csr<ValueType, IndexType>>,
              public DiagonalExtractable<ValueType>,
              public ReadableFromMatrixData<ValueType, IndexType>,
              public WritableToMatrixData<ValueType, IndexType>,
              public EnableAbsoluteComputation<
                  remove_complex<Sellp<ValueType, IndexType>>> {
    friend class EnablePolymorphicObject<Sellp, LinOp>;
    friend class Dense<ValueType>;
    friend class Csr<ValueType, IndexType>;
    friend class Sellp<to_complex<ValueType>, IndexType>;
 
public:
    using EnableLinOp<Sellp>::convert_to;
    using EnableLinOp<Sellp>::move_to;
    using ConvertibleTo<
        Sellp<next_precision<ValueType>, IndexType>>::convert_to;
    using ConvertibleTo<Sellp<next_precision<ValueType>, IndexType>>::move_to;
    using ConvertibleTo<Dense<ValueType>>::convert_to;
    using ConvertibleTo<Dense<ValueType>>::move_to;
    using ConvertibleTo<Csr<ValueType, IndexType>>::convert_to;
    using ConvertibleTo<Csr<ValueType, IndexType>>::move_to;
    using ReadableFromMatrixData<ValueType, IndexType>::read;
 
    using value_type = ValueType;
    using index_type = IndexType;
    using mat_data = matrix_data<ValueType, IndexType>;
    using device_mat_data = device_matrix_data<ValueType, IndexType>;
    using absolute_type = remove_complex<Sellp>;
 
    friend class Sellp<next_precision<ValueType>, IndexType>;
 
    void convert_to(
        Sellp<next_precision<ValueType>, IndexType>* result) const override;
 
    void move_to(Sellp<next_precision<ValueType>, IndexType>* result) override;
 
    void convert_to(Dense<ValueType>* other) const override;
 
    void move_to(Dense<ValueType>* other) override;
 
    void convert_to(Csr<ValueType, IndexType>* other) const override;
 
    void move_to(Csr<ValueType, IndexType>* other) override;
 
    void read(const mat_data& data) override;
 
    void read(const device_mat_data& data) override;
 
    void read(device_mat_data&& data) override;
 
    void write(mat_data& data) const override;
 
    std::unique_ptr<Diagonal<ValueType>> extract_diagonal() const override;
 
    std::unique_ptr<absolute_type> compute_absolute() const override;
 
    void compute_absolute_inplace() override;
 
    value_type* get_values() noexcept { return values_.get_data(); }
 
    const value_type* get_const_values() const noexcept
    {
        return values_.get_const_data();
    }
 
    index_type* get_col_idxs() noexcept { return col_idxs_.get_data(); }
 
    const index_type* get_const_col_idxs() const noexcept
    {
        return col_idxs_.get_const_data();
    }
 
    size_type* get_slice_lengths() noexcept
    {
        return slice_lengths_.get_data();
    }
 
    const size_type* get_const_slice_lengths() const noexcept
    {
        return slice_lengths_.get_const_data();
    }
 
    size_type* get_slice_sets() noexcept { return slice_sets_.get_data(); }
 
    const size_type* get_const_slice_sets() const noexcept
    {
        return slice_sets_.get_const_data();
    }
 
    size_type get_slice_size() const noexcept { return slice_size_; }
 
    size_type get_stride_factor() const noexcept { return stride_factor_; }
 
    size_type get_total_cols() const noexcept
    {
        return values_.get_size() / slice_size_;
    }
 
    size_type get_num_stored_elements() const noexcept
    {
        return values_.get_size();
    }
 
    value_type& val_at(size_type row, size_type slice_set,
                       size_type idx) noexcept
    {
        return values_.get_data()[this->linearize_index(row, slice_set, idx)];
    }
 
    value_type val_at(size_type row, size_type slice_set,
                      size_type idx) const noexcept
    {
        return values_
            .get_const_data()[this->linearize_index(row, slice_set, idx)];
    }
 
    index_type& col_at(size_type row, size_type slice_set,
                       size_type idx) noexcept
    {
        return this->get_col_idxs()[this->linearize_index(row, slice_set, idx)];
    }
 
    index_type col_at(size_type row, size_type slice_set,
                      size_type idx) const noexcept
    {
        return this
            ->get_const_col_idxs()[this->linearize_index(row, slice_set, idx)];
    }
 
    static std::unique_ptr<Sellp> create(std::shared_ptr<const Executor> exec,
                                         const dim<2>& size = {},
                                         size_type total_cols = 0);
 
    static std::unique_ptr<Sellp> create(std::shared_ptr<const Executor> exec,
                                         const dim<2>& size,
                                         size_type slice_size,
                                         size_type stride_factor,
                                         size_type total_cols);
 
    Sellp& operator=(const Sellp&);
 
    Sellp& operator=(Sellp&&);
 
    Sellp(const Sellp&);
 
    Sellp(Sellp&&);
 
protected:
    Sellp(std::shared_ptr<const Executor> exec, const dim<2>& size = {},
          size_type total_cols = {});
 
    Sellp(std::shared_ptr<const Executor> exec, const dim<2>& size,
          size_type slice_size, size_type stride_factor, size_type total_cols);
 
    void apply_impl(const LinOp* b, LinOp* x) const override;
 
    void apply_impl(const LinOp* alpha, const LinOp* b, const LinOp* beta,
                    LinOp* x) const override;
 
    size_type linearize_index(size_type row, size_type slice_set,
                              size_type col) const noexcept
    {
        return (slice_set + col) * slice_size_ + row;
    }
 
private:
    array<value_type> values_;
    array<index_type> col_idxs_;
    array<size_type> slice_lengths_;
    array<size_type> slice_sets_;
    size_type slice_size_;
    size_type stride_factor_;
};
 
 
}  // namespace matrix
}  // namespace gko
 
 
#endif  // GKO_PUBLIC_CORE_MATRIX_SELLP_HPP_