Determine Default Pool Size for Local Parallel Environments

If you have Parallel Computing Toolbox™, you can start a parallel pool of workers to take advantage of the cores in your multicore machine.

A parallel pool is a group of MATLAB^® workers on which you can run code. By default, each worker is limited to a single computational thread. To learn more about parallel pools, see Run Code on Parallel Pools.

Parallel Computing Toolbox provides the parallel environment profiles Processes and Threads for running parallel code on your local desktop machine. The NumWorkers property determines the maximum number of workers that the Processes and Threads profiles use. By default, NumWorkers is equal to the number of high-performance physical CPU cores on your machine.

Physical Cores and Logical Cores

A physical CPU core is a hardware processing unit with its own floating-point unit (FPU), capable of performing calculations and running instructions. Technologies like simultaneous multithreading (SMT) [1] or Intel^® Hyper-Threading Technology [2] allow a single physical core to appear as multiple logical cores to the operating system. However, logical cores often share FPUs, which can cause performance bottlenecks in compute-bound tasks. The default value of NumWorkers is based on the number of physical cores, not logical cores. This setting provides each worker with exclusive access to an FPU.

Machines with Performance and Efficiency Cores

Some CPUs have both Performance-cores (P-cores) and Efficient-cores (E-cores) [3]. P-cores have similar computational capability to traditional physical cores, while E-cores are optimized for power efficiency and less intensive tasks. Workers can use both types of cores, relying on the operating system to schedule tasks across the available cores. However, for parallel pools, a best practice is to start with a number of workers equal to the number of physical P-cores. So, if your CPU has both P- and E-cores, the default value of NumWorkers is equal to the number of physical P-cores only.

Before R2024a: The default number of workers is equal to the total number of physical cores, including both P- and E-cores.

View Additional Pool Size Details in Cluster Profile Manager

Since R2026a

The Cluster Profile Manager displays how MATLAB determines the default value of the NumWorkers property for the Processes and Threads profiles. For example, in the figure below, the NumWorkers property displays: 10, corresponds to number of performant cores (default). This means MATLAB detects 10 high-performance cores and uses those 10 as the default value for NumWorkers.

You can also view the total number of available physical cores, including the number of P-cores if applicable. To view this information, point to the information button next to the NumWorkers property value.

If you change the NumWorkers value, the Cluster Profile Manager displays only the modified property value, without the explanatory text or the "(default)" label. The information icon and tooltip are also not shown. If you restore NumWorkers to its default value, the Cluster Profile Manager returns to displaying the default NumWorkers value, explanatory text, and the information button.

Edit Maximum Number of Workers

You can override the default NumWorkers property value in the Processes profile programmatically or using the Cluster Profile Manager. You can change the NumWorkers property value in the Threads profile only by using the Cluster Profile Manager.

As a best practice, start with NumWorkers equal to the number of physical cores or P-cores on your machine. Then adjust the value gradually while monitoring performance and stability.

• You might further improve performance by including E-cores, but scaling behavior can vary. Compute-bound mathematical operations typically complete faster and more efficiently when they run exclusively on physical cores or physical P-cores where available.

• If your code is not computationally intensive, for example, code that is input and output (I/O) intensive, then you can consider using up to two workers per physical core.

Note

Ensure that your machine has sufficient memory and disk space to support the number of workers you configure. To learn about disk and memory requirements for Parallel Computing Toolbox, see Product Requirements & Platform Availability for Parallel Computing Toolbox.

Using the Cluster Profile Manager

1. Open the Cluster Profile Manager. On the Home tab, in the Environment section, select Parallel > Create and Manage Clusters.

2. In the Cluster Profile tab, select Processes or Threads.

3. Modify the NumWorkers profile property. In the Manage Profile section, click Edit. Enter the new value in the NumWorkers text box.

   

   .

Programmatically

You can programmatically change the NumWorkers property value in the Processes profile using the parcluster and saveProfile functions. Programmatically changing the NumWorkers property value in the Threads profile is not supported.

1. Create a cluster object in your workspace using the Processes profile. 

   c = parcluster("Processes")

   Before R2022b:

   c = parcluster("local") 

2. Specify a new value for the NumWorkers property.

   c.NumWorkers = 10;

3. Save the modified profile of the cluster by using the saveProfile function.

   saveProfile(c) 

The next time you start a pool using the updated profile, MATLAB attempts to create a pool with the number of workers specified in the NumWorkers property.

References

[1] Wikipedia. “Simultaneous multithreading.” August 4, 2025. https://wikipedia.org/w/index.php?title=Simultaneous_multithreading&oldid=1304464531.

[2] Wikipedia. “Hyper-threading.” August 4, 2025. https://wikipedia.org/w/index.php?title=Hyper-threading&oldid=1304462351.

[3] Intel “How Intel® Core™ Processors Work” August 4, 2025. https://www.intel.com/content/www/us/en/gaming/resources/how-hybrid-design-works.html