Resolve Issues in parfor-loops Using MATLAB Copilot

R2026b
Since R2026b

parfor-loops follow specific rules that control how MATLAB® classifies variables and executes loop iterations in parallel. These rules restrict how MATLAB read, writes, and indexes variables across iterations, and restrict certain language features and programming patterns. When you run code that violates these rules, MATLAB generates an error.

If you enable continuous code checking with Code Analyzer, MATLAB displays warnings and error messages about your parfor-loop in the MATLAB Editor and Live Editor before you run your code. To learn more about how Code Analyzer automatically checks code files for errors, see Check Code for Errors and Warnings Using the Code Analyzer.

MATLAB Copilot provides generative AI-powered capabilities specifically for the MATLAB desktop environment. When the Code Analyzer reports parfor warnings or errors, you can use a Copilot action to explain the underlying parfor rule that the code violates and to suggest changes to your code. The Copilot responses use domain-specific material from MathWorks® documentation, information from Code Analyzer, and additional context from the parfor-loop. To use MATLAB Copilot, you must have a MATLAB Copilot license. To learn more about setting up MATLAB Copilot, see Set Up MATLAB Copilot.

Fix parfor-Loop Errors with MATLAB Copilot

If you have a parfor-loop with Code Analyzer warnings and errors that you do not understand, try using Copilot to get a more detailed explanation of the issue.

For example, write a script that runs a parfor-loop with a variable classification error. Copy this code into the MATLAB Editor and save the file.

parfor idx = 1:10
    x(1) = 7;
    x(2) = 8;
    out(idx) = sum(x);
end

In the Editor, the Code Analyzer message indicator bar shows that the file contains an error. The code fragment that causes the error is underlined in red, and the indicator bar shows an error marker at the line of the code fragment.

myParallelCode file in the Editor with the red error symbol displayed in the indicator bar, the error marker in the indicator bar, and the red line under a code fragment.

To view the error message for the code fragment, you can use one of these methods:

  • Click the marker in the indicator bar.

  • Point at the underlined code fragment.

  • Click the underlined code fragment and press Ctrl+M.

The Code Analyzer message on line 2 says that MATLAB cannot classify the x variable. It also includes an Explain button with a symbol, which indicates that you can interact with Copilot. To ask Copilot to explain the variable classification error, click Explain.

myParallelCode file in the Editor after clicking the marker in the indicator bar. The Code Analyzer message displays the error "Line 2: Unable to classify the variable 'x' in the body of the parfor loop.".

Copilot explains the error in the Copilot Chat panel. Typically, the response explains the error, suggests how you can resolve the error, and provides generated code. The exact response depends on the kind of error that you encounter.

Note

Copilot responses can contain errors and inaccuracies. Validate any generated code and text before using it.

Review the response from Copilot. Here, the response explains that MATLAB is unable to classify the x variable because x is assigned to indexed elements in the parfor-loop. The response recommends restructuring the code to make x a temporary variable within each iteration.

Copilot Chat panel, showing a Copilot response to the prompt "Explain 'Unable to classify variable x in the body of the parfor loop'. This code analyzer message originated from the following line of code: 'x(1) = 7;' ".The response explains the variable classification error and generates code. The response also shows that the myParallelCode file is attached to the chat.

If Copilot generates code as part of the explanation, you can insert it into your code file or run it in the Command Window. You can also copy the generated code to the clipboard.

Review and validate the generated code before running it on a remote cluster or with many workers. For example:

  • Confirm that the code has no Code Analyzer warnings or errors.

  • Test the code using serial execution to verify that it produces the expected output.

  • Test the code on a local parallel pool to verify that it runs as expected on the workers.

Editor

If you need additional help to understand the error, you can ask follow-up questions in the Copilot Chat panel. For example, you can ask Copilot to explain the code it just generated. Alternatively, to get a line-by-line explanation of the generated code, select the code, right-click, and select Copilot > Explain Code.

Provide Feedback

MathWorks uses your feedback to improve MathWorks products and services.

There are three buttons at the end of every Copilot response:

  • — Send positive feedback that the answer is helpful.

  • — Send negative feedback that the answer is unhelpful. This button opens a dialog box where you can provide more details. You can choose to send your chat history as part of your feedback.

  • — Copy the entire response to the clipboard.

See Also

Topics