用mpi4py库计算pi值

以下是一个用mpi4py库计算pi值的示例代码：

from mpi4py import MPI
import random

def compute_pi(num_samples):
    comm = MPI.COMM_WORLD
    rank = comm.Get_rank()
    size = comm.Get_size()
    count = 0
    for i in range(num_samples//size):
        x = random.random()
        y = random.random()
        if x*x + y*y < 1:
            count += 1
    count = comm.gather(count, root=0)
    if rank == 0:
        total_count = sum(count)
        pi = 4 * total_count / num_samples
        print("Approximated pi value:", pi)

if __name__ == "__main__":
    num_samples = 1000000 # change this value to increase/decrease accuracy
    compute_pi(num_samples)

此代码使用Monte Carlo方法来估算pi值，它在每个MPI进程中生成num_samples/size个随机点，并计算这些点中有多少在圆内。最后，MPI进程将计数发送到根进程，根进程汇总计数并计算最终的pi值。注意，代码中的num_samples应该至少是MPI进程数的倍数，以确保每个进程都有相同数量的样本点。