登录  
 加关注
查看详情
   显示下一条  |  关闭
温馨提示!由于新浪微博认证机制调整,您的新浪微博帐号绑定已过期,请重新绑定!立即重新绑定新浪微博》  |  关闭

银河里的星星

落在人间

 
 
 

日志

 
 

openMPI测试 编译 运行 及MCA框架(zz)  

2010-04-18 00:18:23|  分类: 高性能计算 |  标签: |举报 |字号 订阅

  下载LOFTER 我的照片书  |

测试Open MPI的安装

命令"ompi_info"可用来检测Open MPI的安装状态。该命令会返回 一个关于你的Open MPI安装的大概信息。

 

Note that the ompi_info command is extremely helpful in determining which components are installed as well as listing all the run-time settable parameters that are available in each component (as well as their default values).

 

下列选项可能会有用:

--all       Show a *lot* of information about your Open MPI installation.

--parsable  Display all the information in an easily grep/cut/awk/sed-able format.

--param <framework> <component>

            A <framework> of "all" and a <component> of "all" will show all parameters to all components.  Otherwise, the parameters of all the components in a specific framework, or just the parameters of a specific component can be displayed by using an appropriate <framework> and/or <component> name.

 

Changing the values of these parameters is explained in the "The Modular Component Architecture (MCA)" section, below.

编译Open MPI应用

Open MPI提供了可用来编译MPI应用的"wrapper"编译器:

C:          mpicc

C++:        mpiCC (or mpic++ if your filesystem is case-insensitive)

Fortran 77: mpif77

Fortran 90: mpif90

 

例如:

shell$ mpicc hello_world_mpi.c -o hello_world_mpi -g

 

All the wrapper compilers do is add a variety of compiler and linker flags to the command line and then invoke a back-end compiler.  To be specific: the wrapper compilers do not parse source code at all; they are solely command-line manipulators, and have nothing to do with the actual compilation or linking of programs.  The end result is an MPI executable that is properly linked to all the relevant libraries.

运行Open MPI应用

 

Open MPI提供了mpirun mpiexec(他们是相同的)。例如:

shell$ mpirun -np 2 hello_world_mpi

shell$ mpiexec -np 1 hello_world_mpi : -np 1 hello_world_mpi

是相同的。某些mpiexecswitches(比如-host –archare not yet functional, 尽管在你使用它们时并不会发生错误。

 

The rsh launcher接受一个-hostfile参数(也可以使用选项"-machinefile",两者是等价的);你可以指定一个-hostfile参数来标志一个标准的mpirun-style hostfile(每行一个主机名):

shell$ mpirun -hostfile my_hostfile -np 2 hello_world_mpi

 

如果你想在一个节点上运行多个进程,那么hostfile 可以使用 "slots" 属性。如果没有指定"slots",那么将假设其数目为1。例如,使用如下的hostfile

---------------------------------------------------------------------------

node1.example.com

node2.example.com

node3.example.com slots=2

node4.example.com slots=4

---------------------------------------------------------------------------

 

shell$ mpirun -hostfile my_hostfile -np 8 hello_world_mpi

 

will launch MPI_COMM_WORLD rank 0 on node1, rank 1 on node2, ranks 2 and 3 on node3, and ranks 4 through 7 on node4.

 

Other starters, such as the batch scheduling environments, do not require hostfiles (and will ignore the hostfile if it is supplied). They will also launch as many processes as slots have been allocated by the scheduler if no "-np" argument has been provided.  For example, running an interactive SLURM job with 8 processors:

 

shell$ srun -n 8 -A

shell$ mpirun a.out

 

The above command will launch 8 copies of a.out in a single MPI_COMM_WORLD on the processors that were allocated by SLURM.

 

Note that the values of component parameters can be changed on the mpirun / mpiexec command line.  This is explained in the section below, "The Modular Component Architecture (MCA)".

The Modular Component Architecture (MCA)

The MCA is the backbone of Open MPI -- most services and functionality are implemented through MCA components.  Here is a list of all the component frameworks in Open MPI:

 

---------------------------------------------------------------------------

MPI 组件框架:

-------------------------

 

allocator - Memory allocator

bml       - BTL management layer

btl       - MPI point-to-point byte transfer layer, used for MPI point-to-point messages on some types of networks

coll      - MPI collective algorithms

io        - MPI-2 I/O

mpool     - Memory pooling

mtl       - Matching transport layer, used for MPI point-to-point messages on some types of networks

osc       - MPI-2 one-sided communications

pml       - MPI point-to-point management layer

rcache    - Memory registration cache

topo      - MPI topology routines

 

Back-end run-time environment component frameworks:

---------------------------------------------------

 

errmgr    - RTE error manager

gpr       - General purpose registry

iof       - I/O forwarding

ns        - Name server

odls      - OpenRTE daemon local launch subsystem

oob       - Out of band messaging

pls       - Process launch system

ras       - Resource allocation system

rds       - Resource discovery system

rmaps     - Resource mapping system

rmgr      - Resource manager

rml       - RTE message layer

schema    - Name schemas

sds       - Startup / discovery service

smr       - State-of-health monitoring subsystem

 

Miscellaneous frameworks:

-------------------------

 

backtrace - Debugging call stack backtrace support

maffinity - Memory affinity

memory    - Memory subsystem hooks

memcpy    - Memopy copy support

memory    - Memory management hooks

paffinity - Processor affinity

timer     - High-resolution timers

 

---------------------------------------------------------------------------

 

Each framework typically has one or more components that are used at run-time.  For example, the btl framework is used by MPI to send bytes across underlying networks.  The tcp btl, for example, sends messages across TCP-based networks; the gm btl sends messages across GM Myrinet-based networks.

 

Each component typically has some tunable parameters that can be changed at run-time.  Use the ompi_info command to check a component to see what its tunable parameters are.  For example:

 

shell$ ompi_info --param btl tcp

 

shows all the parameters (and default values) for the tcp btl component.

 

These values can be overridden at run-time in several ways.  At run-time, the following locations are examined (in order) for new values of parameters:

 

1. <prefix>/etc/openmpi-mca-params.conf

 

   This file is intended to set any system-wide default MCA parameter values -- it will apply, by default, to all users who use this Open MPI installation.  The default file that is installed contains many comments explaining its format.

 

2. $HOME/.openmpi/mca-params.conf

 

   If this file exists, it should be in the same format as <prefix>/etc/openmpi-mca-params.conf.  It is intended to provide per-user default parameter values.

 

3. environment variables of the form OMPI_MCA_<name> set equal to a <value>

 

   Where <name> is the name of the parameter.  For example, set the variable named OMPI_MCA_btl_tcp_frag_size to the value 65536 (Bourne-style shells):

 

   shell$ OMPI_MCA_btl_tcp_frag_size=65536

   shell$ export OMPI_MCA_btl_tcp_frag_size

 

4. the mpirun command line: --mca <name> <value>

 

   Where <name> is the name of the parameter.  For example:

 

   shell$ mpirun --mca btl_tcp_frag_size 65536 -np 2 hello_world_mpi

 

These locations are checked in order.  For example, a parameter value passed on the mpirun command line will override an environment variable; an environment variable will override the system-wide defaults.

 

  评论这张
 
阅读(3140)| 评论(0)

历史上的今天

评论

<#--最新日志,群博日志--> <#--推荐日志--> <#--引用记录--> <#--博主推荐--> <#--随机阅读--> <#--首页推荐--> <#--历史上的今天--> <#--被推荐日志--> <#--上一篇,下一篇--> <#-- 热度 --> <#-- 网易新闻广告 --> <#--右边模块结构--> <#--评论模块结构--> <#--引用模块结构--> <#--博主发起的投票-->
 
 
 
 
 
 
 
 
 
 
 
 
 
 

页脚

网易公司版权所有 ©1997-2018