如何启动IPython MPI集群?

时间:2012-10-18 18:40:45

标签: mpi ipython

我正在尝试使用MPI引擎在本地计算机上启动IPython。我需要MPI。 The IPython documentation建议只运行这个:

$ ipcluster start -n 4 --engines=MPIEngineSetLauncher

然而,在此之后它并没有说明该怎么做。对我来说导致:

$ ipcluster start -n 4 --engines=MPIEngineSetLauncher
2012-10-18 11:07:14,340.340 [IPClusterStart] Using existing profile dir: u'/Users/enos/.ipython/profile_default'
2012-10-18 11:07:14.342 [IPClusterStart] Starting ipcluster with [daemon=False]
2012-10-18 11:07:14.342 [IPClusterStart] Creating pid file: /Users/enos/.ipython/profile_default/pid/ipcluster.pid
2012-10-18 11:07:14.342 [IPClusterStart] Starting Controller with LocalControllerLauncher
2012-10-18 11:07:15.343 [IPClusterStart] Starting 4 Engines with MPIEngineSetLauncher
2012-10-18 11:07:45.348 [IPClusterStart] Engines appear to have started successfully

它就在那里。我如何获得IPython命令提示符?

我在OSX 10.8上使用4核机器在IPython 0.13上尝试这个。

1 个答案:

答案 0 :(得分:0)

我在评论中找到了答案,但我认为将它放在这里会更有用。

“你只需在另一个终端启动ipython。然后示例代码创建一个连接到正在运行的引擎的Client实例。 - Greg Inozemtsev”