I have to do a parallel simulation without a parallel simulation tool box? The main idea is to Remove a for loop that does one iteration after another to speed up computation.

I have long lines of code where iam trying to evaluate 36 rows of data having 7 columns. The iterations happen one after other via a for loop. To reduce computational time i have to use do parallel simulation but there is no parallel simulation toolbox available

[Jeffrey Clark]

If this is a Simulink question I have no idea. However, if you are coding a simulation in MATLAB you can have two parallel computations using 'parfeval' even without the Parallel Processing Toolbox.

[Riccardo Scorretti]

It depends on what you have to do exactly. Basically, without the parallel computational toolbox you have two options:

1. exploit as much as possible vectorization: can you make use of .* operator instead of *, and so on,
2. to really parallelize your algorithm, you could write a MEX function and use OpenMP.

Share an exemple of what you want to do; perhaps writing a MEX function is not that complicate.

[arjun ramesh]

It said while using a parfeval i need a toolbox else it runs in serial. I have to try the next part now

[Walter Roberson]

In sufficiently new versions you can use "background pools" without the toolbox.

[arjun ramesh]

Iam using a 2019 b version and it's still says 'parpool' require a parallel computing toolbox

Answer 1

[Walter Roberson]

You have a small number of possibilities:

• upgrade to a version that supports background pools
• obtain the parallel computing toolbox
• use Java threads to do your processing in Java code
• use system() to create additional matlab processes to execute some of the work
• In Windows, use .NET to create additional processes to do some of the work
• Use the file exchange contribution popen() to create additional processes to do some of the work
• use mex C or C++ code to invoke parallel processing using OpenMP or pthreads or popen or whatever is available