Python中MPI消息传递接口

MPI(Message-Passing-Interface)消息传递接口

1.MPI安装

python mpi安装mpi4py的python库

pip install mpi4py

下载MicrosoftWindowsMPI安装包网址：https://github.com/icepoint666/MPI中的MSMpiSetup.rar

安装后得到Bin文件夹：

将MicrosoftMPI/Bin文件夹路径添加到环境变量：

命令行输入mpiexec运行，出现帮助表示安装成功

2.试验代码

from mpi4py import MPI

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

print 'I'm the %d process of %d processes" % (comm_rank, comm_size)

2.点对点传输

点对点通信.其实就是最简单的进程A向进程B发送信息,而进程B向进程A接收信息.这是关于两个进程之间的通信.

示例代码：

#mpip2p.py
from mpi4py import MPI

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

data = [comm_rank]*5
comm.send(data,dest=(comm_rank+1)%comm_size)
data_recv =comm.recv(source=(comm_rank-1)%comm_size)
print "my rank is %d, and Ireceived:" %comm_rank
print data_recv

在命令行中输入命令

mpiexec -n 5 python mpip2p.py

执行结果：

my rank is 4, and Ireceived:
[3, 3, 3, 3, 3]
my rank is 3, and Ireceived:
[2, 2, 2, 2, 2]
my rank is 2, and Ireceived:
[1, 1, 1, 1, 1]
my rank is 0, and Ireceived:
[4, 4, 4, 4, 4]
my rank is 1, and Ireceived:
[0, 0, 0, 0, 0]

指定启动5个mpi进程来执行后面的程序。相当于对脚本拷贝了5份，每个进程运行一份，互不干扰。在运行的时候代码里面唯一的不同，就是各自的rank也就是ID不一样。

Get_rank()函数：获取当前进程rank值

Get_size()函数：获取总共的进程数

send()函数：将数据送给rank为dest的值的进程

rec