Install and configure MPI - Revision history

Jprenaud: /* Compilation */

2008-02-14T14:29:26Z

Compilation

Jprenaud at 10:07, 22 August 2007

2007-08-22T10:07:23Z

Jprenaud: /* Test the ring */

2007-08-22T10:05:04Z

Test the ring

Jprenaud: /* Trouble shooting */

2007-08-22T10:04:20Z

Trouble shooting

Jprenaud at 10:04, 22 August 2007

2007-08-22T10:04:09Z

Jprenaud: /* Close the ring */

2007-08-22T10:00:43Z

Close the ring

Jprenaud: /* Test the ring */

2007-08-22T10:00:33Z

Test the ring

Jprenaud: /* Using an MPI ring */

2007-08-22T09:59:39Z

Using an MPI ring

Jprenaud: /* Password in `.mpd.conf` */

2007-08-22T09:43:39Z

Password in <code>.mpd.conf</code>

Jprenaud: /* Configuration of MPI */

2007-08-22T09:42:29Z

Configuration of MPI

@@ Line 38: / Line 38: @@
 </pre></code>
-This will install MPI in <code>/path/tp/mpi/</code>.
+This will install MPI in <code>/path/to/mpi/</code>.
 Note that you will need to install MPI on all the nodes in your network that will be used for MPI jobs. As they probably have the same computer architecture, you could just copy the <code>/path/to/mpi</code> accross.

@@ Line 1: / Line 1: @@
 [[Category:Estel]]
-This article explains how to install and configure MPI to be able to run the '''ESTEL''' model in parallel on a network of computers. Note that this article is merely a quick run through the [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI Installer's Guide] at http://www-unix.mcs.anl.gov/mpi/mpich2/.
+This article explains how to install and configure MPI to be able to run the '''[[Estel | ESTEL]]''' model in parallel on a network of computers. Note that this article is merely a quick run through the [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI Installer's Guide] at http://www-unix.mcs.anl.gov/mpi/mpich2/.
 = Pre-requesites=
@@ Line 135: / Line 137: @@
 </pre></code>
-If you can do this, '''congratulations'''! You should be able to run parallel ESTEL jobs on your ring now!
+If you can do this, '''congratulations'''! You should be able to run parallel [[Estel | ESTEL]] jobs on your ring now!
 Just remember to start a <code>mpd</code> ring before running your parallel job.

@@ Line 135: / Line 135: @@
 </pre></code>
-If you can do this, congratulations, you should be able to run parallel programs on your ring now! Just remember to start a <code>mpd</code> ring before running your parallel job.
+If you can do this, '''congratulations'''! You should be able to run parallel ESTEL jobs on your ring now!
 == Close the ring ==

← Older revision		Revision as of 10:04, 22 August 2007
Line 149:		Line 149:
	* read the [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI manual]!		* read the [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI manual]!
	* take your time...		* take your time...
−	~~/etc/hosts~~

← Older revision		Revision as of 10:04, 22 August 2007
Line 144:		Line 144:

	= Trouble shooting =		= Trouble shooting =
		+	The [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI Installer's Guide] has a good section about troubleshooting (Appendix A). To summarise things that might be helpful:
		+	* make sure you can start <code>mpd</code> on each host separately first
		+	* check that the <code>/etc/hosts</code> on each host is correct and has an entry for each host with the right IP address.
		+	* read the [http://www-unix.mcs.anl.gov/mpi/mpich2/downloads/mpich2-doc-install.pdf MPI manual]!
		+	* take your time...
	/etc/hosts		/etc/hosts

@@ Line 138: / Line 138: @@
 == Close the ring ==
-The command <code>mpdallexit</code> is used to terminate the <code>mpd</code>ring:
+The command <code>mpdallexit</code> is used to terminate the <code>mpd</code> ring:
 <code><pre>
 master $ mpdallexit

@@ Line 96: / Line 96: @@
 This file should be created on the master node, i.e. the one that you will use to launch MPI jobs. All the hosts in the list need to have a working installation of MPI.
-= Using an MPI ring =
+= Using an <code>mpd</code> ring =
 == Start the ring ==
 <code><pre>
 master $ mpd -n 3 -f ~/mpd.hosts
 </pre></code>
-Where 3 is the number of nodes to include in the ring. By default, the master is always included in the ring.
+In the example above, "3" is the number of nodes to include in the ring. By default, the master is always included in the ring.
 == Test the ring ==
-the command <code>mpdtrace</code> can be used to test which machines are in the ring:
+The command <code>mpdtrace</code> can be used to list which machines are in the ring:
 <code><pre>
 master $ mpdtrace
@@ Line 114: / Line 118: @@
 </pre></code>
-You can also send a command in parallel using <code>mpiexec</code>. The option <code>-l</code> will append the node number before the outoput of the command. The option <code>-n 5</code> request the job to be sent to 5 processors, as we only have 3, it will wrap around as shown in the listing below where master and slave2 have been used twice:
+To test the ring, use <code>mpdringtest</code>. That will send a message to circle around the ring (a loop) and tell you how long it took. One loop is very fast so make the message circle round a few times:
 <code><pre>
 master $ mpiexec -l -n 5 hostname
@@ Line 123: / Line 134: @@
 : slave2
 </pre></code>
 == Close the ring ==
-The command <code>mpdallexit</code> is used to close the MPI ring:
+The command <code>mpdallexit</code> is used to terminate the <code>mpd</code>ring:
 <code><pre>
 master $ mpdallexit

@@ Line 82: / Line 82: @@
 </pre></code>
-It is interesting to note that we asked for two processes with the <code>mpiexec</code> command with the arguments <code>-n 2</code> although we have only one machine in our network yet. The <code>mpd</code> is intelligent enough to just wrap around. This is useful to test parallel programs on a single machine.
+It is interesting to note that we asked for two processes with the <code>mpiexec</code> command with the arguments <code>-n 2</code> although we have only one machine in our network yet. The <code>mpd</code> is intelligent enough to just wrap around. In the output above, the <code>hostname</code> is used and we can see that processes 0 and 1 return the sdame value so thay are definitely run on the same host. This is very useful to test parallel programs on a single machine.
 == Hosts in <code>mpd.hosts</code> ==

@@ Line 48: / Line 48: @@
 </pre></code>
-In particular, you should be able to run the MPI Fortran 90 wrapper <code>mpif90</code> now:
+Now, you should be able to see the MPI Fortran 90 wrapper <code>mpif90</code> and the MPI multiprocessor daemon <code>mpd></code>:
 <code><pre>
-which mpif90
+$ which mpif90
 /path/to/mpi/bin/mpif90
 </pre></code>
-This needs to be done on each node in the network. This is straightforward if your home directory is shared for all the nodes or you need to do it manually othwerwise.
+Note that the <code>PATH</code> needs to be set on each node which will be used for MPI simulations. This is straightforward if your home directory is shared for all the nodes or you need to do it manually otherwise.
 == Password in <code>.mpd.conf</code> ==
@@ Line 66: / Line 68: @@
 '''Check if this needs to be done on all nodes.... and if the secret thing needs to be the same. Probably yes.'''
 == Hosts in <code>mpd.hosts</code> ==
-Create a file in your home directory called <code>mpd.hosts</code> which contain a list of the nodes to be used by MPI, one per line. If the network consists of the master and two slaves slave1 and slave2, <code>mpd.conf</code> would contain:
+To be able to send the processes to other hosts on the network, create a file in your home directory called <code>mpd.hosts</code> which contain a list of the nodes to be used by MPI, one per line. If the network consists of the master and two slaves slave1 and slave2, <code>mpd.conf</code> would contain:
 <code><pre>
@@ Line 77: / Line 94: @@
 </pre></code>
-This file should be created on the master node, i.e. the one that you will use to launch MPI jobs.
+This file should be created on the master node, i.e. the one that you will use to launch MPI jobs. All the hosts in the list need to have a working installation of MPI.
 = Using an MPI ring =

Install and configure MPI - Revision history

Jprenaud: /* Compilation */

Jprenaud at 10:07, 22 August 2007

Jprenaud: /* Test the ring */

Jprenaud: /* Trouble shooting */

Jprenaud at 10:04, 22 August 2007

Jprenaud: /* Close the ring */

Jprenaud: /* Test the ring */

Jprenaud: /* Using an MPI ring */

Jprenaud: /* Password in .mpd.conf */

Jprenaud: /* Configuration of MPI */

Jprenaud: /* Password in `.mpd.conf` */