tatami_mult.hpp

Go to the documentation of this file.

#ifndef TATAMI_MULT_HPP
#define TATAMI_MULT_HPP
 
#include "dense_row.hpp"
#include "sparse_row.hpp"
#include "dense_column.hpp"
#include "sparse_column.hpp"
 
#include <vector>
 
#include "tatami/tatami.hpp"
#include "sanisizer/sanisizer.hpp"
 
namespace tatami_mult {
 
struct Options {
    int num_threads = 1;
 
    bool prefer_larger = true;
 
    bool column_major_output = true;
};
 
template<typename Value_, typename Index_, typename Right_, typename Output_>
void multiply(const tatami::Matrix<Value_, Index_>& left, const Right_* right, Output_* output, const Options& opt) {
    if (left.sparse()) {
        if (left.prefer_rows()) {
            internal::sparse_row_vector(left, right, output, opt.num_threads);
        } else {
            internal::sparse_column_vector(left, right, output, opt.num_threads);
        }
    } else {
        if (left.prefer_rows()) {
            internal::dense_row_vector(left, right, output, opt.num_threads);
        } else {
            internal::dense_column_vector(left, right, output, opt.num_threads);
        }
    }
}
 
template<typename Left_, typename Value_, typename Index_, typename Output_>
void multiply(const Left_* left, const tatami::Matrix<Value_, Index_>& right, Output_* output, const Options& opt) {
    auto tright = tatami::make_DelayedTranspose(tatami::wrap_shared_ptr(&right));
    if (tright->sparse()) {
        if (tright->prefer_rows()) {
            internal::sparse_row_vector(*tright, left, output, opt.num_threads);
        } else {
            internal::sparse_column_vector(*tright, left, output, opt.num_threads);
        }
    } else {
        if (tright->prefer_rows()) {
            internal::dense_row_vector(*tright, left, output, opt.num_threads);
        } else {
            internal::dense_column_vector(*tright, left, output, opt.num_threads);
        }
    }
}
 
template<typename Value_, typename Index_, typename Right_, typename Output_>
void multiply(const tatami::Matrix<Value_, Index_>& left, const std::vector<Right_*>& right, const std::vector<Output_*>& output, const Options& opt) {
    if (left.sparse()) {
        if (left.prefer_rows()) {
            internal::sparse_row_vectors(left, right, output, opt.num_threads);
        } else {
            internal::sparse_column_vectors(left, right, output, opt.num_threads);
        }
    } else {
        if (left.prefer_rows()) {
            internal::dense_row_vectors(left, right, output, opt.num_threads);
        } else {
            internal::dense_column_vectors(left, right, output, opt.num_threads);
        }
    }
}
 
template<typename Left_, typename Value_, typename Index_, typename Output_>
void multiply(const std::vector<Left_*>& left, const tatami::Matrix<Value_, Index_>& right, const std::vector<Output_*>& output, const Options& opt) {
    auto tright = tatami::make_DelayedTranspose(tatami::wrap_shared_ptr(&right));
    if (tright->sparse()) {
        if (tright->prefer_rows()) {
            internal::sparse_row_vectors(*tright, left, output, opt.num_threads);
        } else {
            internal::sparse_column_vectors(*tright, left, output, opt.num_threads);
        }
    } else {
        if (tright->prefer_rows()) {
            internal::dense_row_vectors(*tright, left, output, opt.num_threads);
        } else {
            internal::dense_column_vectors(*tright, left, output, opt.num_threads);
        }
    }
}
 
namespace internal {
 
template<typename LeftValue_, typename LeftIndex_, typename RightValue_, typename RightIndex_, typename Output_>
void multiply(const tatami::Matrix<LeftValue_, LeftIndex_>& left, const tatami::Matrix<RightValue_, RightIndex_>& right, Output_* output, bool column_major_out, int num_threads) {
    RightIndex_ row_shift;
    LeftIndex_ col_shift;
    if (column_major_out) {
        row_shift = 1;
        col_shift = left.nrow();
    } else {
        row_shift = right.ncol();
        col_shift = 1;
    }
 
    if (left.sparse()) {
        if (left.prefer_rows()) {
            if (right.sparse()) {
                internal::sparse_row_tatami_sparse(left, right, output, row_shift, col_shift, num_threads);
            } else {
                internal::sparse_row_tatami_dense(left, right, output, row_shift, col_shift, num_threads);
            }
        } else {
            if (right.sparse()) {
                internal::sparse_column_tatami_sparse(left, right, output, row_shift, col_shift, num_threads);
            } else {
                internal::sparse_column_tatami_dense(left, right, output, row_shift, col_shift, num_threads);
            }
        }
    } else {
        if (left.prefer_rows()) {
            if (right.sparse()) {
                internal::dense_row_tatami_sparse(left, right, output, row_shift, col_shift, num_threads);
            } else {
                internal::dense_row_tatami_dense(left, right, output, row_shift, col_shift, num_threads);
            }
        } else {
            if (right.sparse()) {
                internal::dense_column_tatami_sparse(left, right, output, row_shift, col_shift, num_threads);
            } else {
                internal::dense_column_tatami_dense(left, right, output, row_shift, col_shift, num_threads);
            }
        }
    }
}
 
}
template<typename LeftValue_, typename LeftIndex_, typename RightValue_, typename RightIndex_, typename Output_>
void multiply(const tatami::Matrix<LeftValue_, LeftIndex_>& left, const tatami::Matrix<RightValue_, RightIndex_>& right, Output_* output, const Options& opt) {
    if (opt.prefer_larger) {
        if (sanisizer::is_less_than(left.nrow(), right.ncol())) {
            auto tright = tatami::make_DelayedTranspose(tatami::wrap_shared_ptr(&right));
            auto tleft = tatami::make_DelayedTranspose(tatami::wrap_shared_ptr(&left));
            internal::multiply(*tright, *tleft, output, !opt.column_major_output, opt.num_threads);
            return;
        }
    }
 
    internal::multiply(left, right, output, opt.column_major_output, opt.num_threads);
}
 
}
 
#endif