如何用openmp优化矩阵向量乘法？

#pragma omp parallel for
for (int i = 0; i < matrix_rows; i++) {
    double sum = 0;
    for (int j = 0; j < matrix_cols; j++) {
        sum += matrix[i][j] * vector[j];
    }
    result[i] = sum;
}

#include <omp.h>

// 设置线程数为4
omp_set_num_threads(4);

#pragma omp parallel for
for (int i = 0; i < matrix_rows; i++) {
    double sum = 0;
    for (int j = 0; j < matrix_cols; j++) {
        sum += matrix[i][j] * vector[j];
    }
    result[i] = sum;
}

#pragma omp parallel private(sum)
{
    int tid = omp_get_thread_num();
    double sum = 0;
    #pragma omp for
    for (int i = 0; i < matrix_rows; i++) {
        for (int j = 0; j < matrix_cols; j++) {
            sum += matrix[i][j] * vector[j];
        }
        result[i] = sum;
    }
}