Specify Your Parallel Preferences

You can access your parallel preferences in any of the following ways:

• On the Home tab in the Environment section, select Parallel > Parallel Preferences

• Select the desktop pool indicator icon, and select Parallel Preferences.

• In the command window, type preferences.

  preferences

  In the navigation tree of the Preferences dialog box, select Parallel Computing Toolbox.

You can control your parallel preference settings as follows:

• Default Profile — Choose the profile you want to use. The default profile is Processes. For more information, see Add and Modify Cluster Profiles.

• Preferred number of workers — Specify the number of workers in your parallel pool. The actual pool size is limited by licensing, cluster size, and cluster profile settings. See Pool Size and Cluster Selection. For the Processes profile, do not choose a preferred number of workers larger than 512. See also Add and Modify Cluster Profiles. Check your access to cloud computing from the Parallel > Discover Clusters menu.

• Automatically create a parallel pool — If a parallel pool is not open, some functionality in Parallel Computing Toolbox™ will automatically create a parallel pool, including:

  • parfor

  • spmd

  • distributed

  • Composite

  • parfeval

  • parfevalOnAll

  • afterEach

  • afterAll

  • gcp

  • mapreduce

  • mapreducer

  • parsim (Simulink)

  When these functions are used, select Automatically create a parallel pool to create a pool automatically. If you select this option, you do not need to open a pool manually using the parpool function.

  If this option is not selected, a pool is not open, and you use:

  • parfeval, parfevalOnAll, afterEach, or afterAll, your code will throw an error.

  • Other Parallel Computing Toolbox functionality, your code will run on the client.

• Shut down and delete a parallel pool — To shut down a parallel pool automatically if the pool has been idle for the specified amount of time, use the IdleTimeout setting. If you use the pool (for example, using parfor or parfeval), the timeout counter is reset. When the timeout is about to expire, a tooltip on the desktop pool indicator warns you and allows you to reset the timer. Note that modifying this setting changes the IdleTimeout of any already started pool.

Related Examples

• Run MATLAB Functions with Automatic Parallel Support
• Scale Up from Desktop to Cluster

More About

• Decide When to Use parfor
• Scale Up parfor-Loops to Cluster and Cloud
• Add and Modify Cluster Profiles