'parpool(2)' fails

Hi, I am testing some MATLAB code via GitHub Action. Strange bug occurs when I was testing `parpool(2)` with MATLAB R2021b and R2022a on Mac and Ubuntu as follows.

Error using parpool

Undefined function 'pCurrentTaskIsPartOfAPool' for input arguments of type

'double'.

Error in solveAllProblems (line 68)

parpool(numJobs);

The code is fine with the latest verion of MATLAB.

Any comment helps.

[Walter Roberson]

Do you have the Parallel Computing Toolbox installed?

[Cunxin Huang]

Hi @Walter Roberson

Sure, following is part of the code for the workflow of GitHub Action.

name: test parallel computing toolbox

on:
  push:
    branches: [matlab]

jobs:

  test:

    name: MATLAB ${{ matrix.matlab }} on ${{ matrix.os }}
    runs-on: ${{ matrix.os }}

    strategy:
      fail-fast: false
      matrix:
        os: [ubuntu-latest, windows-latest, macOS-latest]
        matlab: [latest, R2021b, R2022a, R2022b, R2023a, R2023b]

    steps:

      - name: Check out repository
        uses: actions/checkout@v4

      - name: Set up MATLAB with optimization toolbox and parallel computing toolbox
        uses: matlab-actions/setup-matlab@v2
        with:
          release: ${{ matrix.matlab }}
          cache: true
          products: Optimization_Toolbox Parallel_Computing_Toolbox

      - name: Test parallel computing
        uses: matlab-actions/run-command@v2.1.1
        with:
          command: |
            parpool(2)

You can directly try the above code.

Answer 1

[Sandeep Mishra]

Hi Cunxin,

I tried to execute the GitHub Action and observed the same error message on all OS (Windows, Ubuntu and Mac) for MATLAB R2022b.

The issue arises because parpools with batch licensing are not supported in MATLAB versions prior to R2023a.

To resolve this, I recommend upgrading to MATLAB R2023a or later, which supports the use of parpools with batch licensing.

I hope this helps you!

[Cunxin Huang]

Hi @Sandeep Mishra

Thanks for your kind reply.

If this is true, then MATLAB documents provide wrong information.

In the help center for parpool, version history part states that:

""""""""

R2022b: Specify number of thread workers

Starting in R2022b, you can now specify the pool size of a thread-based parallel pool using the parpool(poolsize) syntax.

""""""""

Beside this, I guess earlier version should support process-based parallel pool using the parpool(poolsize) syntax?