This is machine translation

Translated by Microsoft
Mouseover text to see original. Click the button below to return to the English verison of the page.

Note: This page has been translated by MathWorks. Please click here
To view all translated materals including this page, select Japan from the country navigator on the bottom of this page.

Customizing the Settings for the Examples in the Parallel Computing Toolbox™

This example shows how to change the behavior of the examples in the Parallel Computing Toolbox™. There are at least two versions of each example in the Parallel Computing Toolbox: a sequential version and a distributed version. The distributed versions will not run unless they can submit jobs to a cluster, so the most important setting explained here is the cluster profile that the examples use to identify the cluster to submit to.

If you are not familiar with the concepts of distributed computing, please read the documentation for the Parallel Computing Toolbox in the MATLAB® help browser.

For further reading, see:

Accessing and Changing the Default Profile

The examples all use the default profile when creating and submitting jobs. To be more specific, the examples use the cluster identified by the default profile. On the Home tab, in the Environment section, click Parallel > Default Cluster to view and change the default profile, or you can use the parallel.defaultClusterProfile function to get and set the default profile:

ans =


The profiles documentation contains the complete guide to managing your profiles.

Accessing the Current Example Settings

All example-related settings other than the default profile can be obtained and changed via paralleldemoconfig. Running the paralleldemoconfig command without any input arguments gives us a structure with the current example settings. We will examine these settings one at a time. Any changes that we make to these settings remain in effect for the duration of the current MATLAB session. If we want some changes to be applied every time MATLAB starts, we should put them into our startup.m file. First, we look at the current settings:

orgconf = paralleldemoconfig();
disp( orgconf )
      NumTasks: 4
    NetworkDir: [1x1 struct]
    Difficulty: 1

Changing the Number of Tasks

The number of tasks the examples create is stored in the NumTasks field. We can change it to a different value by using paralleldemoconfig.

paralleldemoconfig('NumTasks', 3); % Use 3 tasks.

Changing the Example Difficulty Level

The total run time of most of the examples can be changed by a constant factor through the Difficulty field of paralleldemoconfig. The default value is 1.0, and the examples attempt to adapt their problem sizes so that they run in half the regular time when the Difficulty is set to 0.5 and twice the regular time when the Difficulty is set to 2.0. The Difficulty field can be set to any value greater than 0.

It should be noted that some of the examples perform a fixed amount of computation, therefore their run time cannot be controlled by the Difficulty field, and those examples issue a warning if they are run with a Difficulty value other than 1.0. Let's see how the sequential blackjack example responds to changes in the difficulty level:

paralleldemoconfig('Difficulty', 0.5);
Elapsed time is 17.4 seconds

By looking at the graph of the total winnings, we can see how many people were playing in the simulation. Observe that when we set the Difficulty field to 1.0, we simulate twice the number of players we simulated when using a difficulty level of 0.5. Also notice that the Blackjack example executes in half the time when run with Difficulty set to 0.5 when compared to the default value of 1.0.

paralleldemoconfig('Difficulty', 1.0);
Elapsed time is 35.1 seconds

Customizing the Network Directory

The NetworkDir field stores the name of a directory on a shared file system that the examples can use when they want to transfer input/output data between the client and the workers via the file system. The directory is specified in two different formats, one for the Windows® platform and one for UNIX® platforms. If our client and workers are only using one platform, we need to concern ourselves with only that field and we can safely ignore the other. Additionally, the directory permissions must be such that MATLAB can write to it, both from the client and the workers.

Assuming that the following two paths refer to the same directory on our file server, we can inform the Parallel Computing Toolbox of that correspondence:

windowsdir = '\\mycomputer\user\subdir';
unixdir = '/home/user/subdir';
paralleldemoconfig('NetworkDir', ...
    struct('windows', windowsdir, 'unix', unixdir));

Notice that we specified the Windows path as a UNC path and not in terms of mapped network drives.

If we do not have a shared network file system, we cannot run any of the examples that try to make use of it. The same applies if there is no directory that all the Windows machines recognize by one name and all the UNIX machines can recognize by another name.

Restoring the Original Settings

We do not want this tutorial to change the default example settings, so we restore their original values.

Was this topic helpful?