python mpi简单应用

借鉴博客学习一下py下mpi的基本语法

https://blog.csdn.net/u012526003/article/details/80299113

发送

comm.send( data, pid)

接受

data_recv = comm.recv(data)

广播：

comm.bcast(data, 0) (None, 0 )

散播：均分

comm_scatter(data, root)

收集数据：

comm_gather(data, pid)

规约：

comm_reduce(data , pid , op=MPI.SUM)

 

下面让我们来写一个最简单的入门程序——求π值

没找到mpi的函数库，原来c的mpi里面的wtime函数在Python中没有，以此注意n的传值，并且保证主进程最后一个结束输出pi

from mpi4py import MPI

def func(x):
   return (4.0 / (1.0 + x * x))

comm = MPI.COMM_WORLD
comm_rank = comm.Get_rank()
comm_size = comm.Get_size()

PI = 3.141592653589
num = 1
if comm_rank == 0:
   num = 10000
#   starttime = MPI.wtime()
n = comm.bcast(num if comm_rank == 0 else None, root=0)
print("rank %d, n = %d" % (comm_rank, n))

h = 1.0/n
sum = 0.0
for i in range(comm_rank, n, comm_size):
   x = h * (0.5 + i)
   sum += func(x)
npi = h * sum
mypi = comm.reduce(npi, root=0, op=MPI.SUM)

if comm_rank == 0:
   #endtime = MPI.wtime()i
   #waiting other end
   for i in range(1,comm_size):
      x = comm.recv(source = i)
   print("pi is %.16f"% mypi)
else:
   comm.send(n, dest=0)

下面让我们来测试一下程序