Debugging MATLAB-related problems

The .m program is using a software extension to MATLAB called “Dynare”

See Using Dynare in MATLAB.

Mention of MEX programs, what do I do?

(or: MEX files provided for Linux/MacOS, but you need to run on Windows)

• There should be .cpp files provided - if not, they should be requested from the author

• From within Matlab, run mex name_of_file.cpp (though this should also be provided by the author).

MATLAB complains about a missing Toolbox

Check that the Path includes said Toolbox, see https://it.mathworks.com/help/matlab/ref/pathtool.html. It may also be that we are not licensed to use that Toolbox. Make a note of the error in the replication report.

MATLAB uses too much memory or too many CPU cores (parpool)

When running MATLAB on BioHPC (or other clusters), you will sometimes find that authors have hard-coded the number of CPU cores to use, or might ignore the amount of memory needed.

Code might look like this:

%% Parallelizing code

c = parcluster;
c.NumWorkers =30;
parpool(30)

The following code might be helpful in making this somewhat easier to debug:

%% Parallelizing code dynamically

% 1. Get the CPU count from SLURM environment
% SLURM_CPUS_PER_TASK is set when you use --cpus-per-task
num_cpus = str2double(getenv('SLURM_CPUS_PER_TASK'));

% 2. Fallback: If running locally or variable isn't set, default to 1 or a safe number
if isnan(num_cpus)
    num_cpus = 1; 
    fprintf('SLURM variable not found. Defaulting to 1 worker.\n');
else
    fprintf('Detected %d CPUs assigned by SLURM.\n', num_cpus);
end

% 3. Initialize the pool
c = parcluster;
c.NumWorkers = num_cpus;
parpool(c, num_cpus);