如何使用python mpi4py将所有可能的列对分散/发送到子进程，并找到列之间的一致性？并行计算

from mpi4py import MPI
import numpy as np

def find_consistency(columns):
    # 这里假设一致性检查是一个简单的比较操作
    return np.allclose(columns[0], columns[1])

def main():
    comm = MPI.COMM_WORLD
    rank = comm.Get_rank()
    size = comm.Get_size()

    # 假设我们有一个矩阵
    matrix = np.array([[1.0, 2.0, 3.0], [4.0, 5.0, 6.0], [7.0, 8.0, 9.0]])
    num_cols = matrix.shape[1]

    # 计算每个进程需要处理的列对数量
    num_pairs = num_cols * (num_cols - 1) // 2
    pairs_per_process = num_pairs // size
    remainder = num_pairs % size

    start_pair = rank * pairs_per_process + min(rank, remainder)
    end_pair = start_pair + pairs_per_process + (1 if rank < remainder else 0)

    for i in range(start_pair, end_pair):
        col1 = i // (num_cols - 1)
        col2 = i % (num_cols - 1) + (1 if col1 == i // (num_cols - 1) else 0)
        column_pair = (matrix[:, col1], matrix[:, col2])

        # 分散列对到子进程
        if rank == 0:
            for dest in range(1, size):
                comm.send(column_pair, dest=dest)
        else:
            column_pair = comm.recv(source=0)

        # 检查一致性
        is_consistent = find_consistency(column_pair)
        print(f"Process {rank}: Columns {col1} and {col2} are consistent: {is_consistent}")

if __name__ == "__main__":
    main()