Can other computers that install my compiled appdesigner app make use of its parfor statements?

I am having a very hard time finding the answer to this question because there seems to be a lot of conflicting information out there. Here are the facts:

1. I have MATLAB Compiler and the Parallel Computing Toolbox on my computer
2. I have coded an app using appdesigner. One of the buttons on the UI enacts a certain algorithm that makes use of parfor statements.
3. I compiled the app and sent the installer to someone who does not have matlab or the parallel computing toolbox on their computer.
4. Although the app works, the parfor statements behave like regular for statements.

Is it possible to compile an app for use by a different user that has working parfor statements?

• Certain answers have said no. To this, I wonder, will matlab ever add this feature or is it not a feature on purpose?
• Other answers have said yes and detail that you need to send the user an exported local cluster file and add a setmcruserdata('ParallelProfile', 'profile') command before the parfor statements. This didn't work. To this, I wonder, am I doing something wrong?

Long story short, does a user have to MATLAB and a Parallel Computing Toolbox license in order to properly run an app that I have created that uses parfor statements, even though I own the matlab compiler which is supposed to make standalone apps that users can use without having a MATLAB license?

This implies you can. However, this implies you cannot. Which one is true? Am I missing something?

Thanks!

[Adam]

I have done this many times with apps built in GUIDE. I've never done it with an AppDesigner application, but I don't see why the GUI wrapper should make a difference.

Are you sure the target computer actually has multiple cores?

I have never exported any cluster files or anything like that. My understanding from when I looked into it quite a while ago was that it takes the parallel settings from the Matlab from which it is built, i.e. in my case it took my settings which tell it to use up to a 28-core pool.

[Tahoe Schrader]

The target computer is identical to mine in every way except for the fact that it does not have the parallel computing toolbox installed.

My GUI has a button that enacts the following algorithm:

parpool local;
parfor some range
    some algorithm
end

This statement ran as slowly as it does on my computer when I change parfor to for. Since our computers are completely identical, this tells me the target computer is not properly parallel processing. So, I added another button that does this:

setmcruserdata('ParallelProfile', 'profile')

and then I sent the target computer a profile.settings file that I exported from my own computer that tells the target computer to use its local cores. This didn't work either.

I then compiled the following function:

function run_parallel_application
    setmcruserdata('ParallelProfile', 'profile')
    system('the GUI application .exe')
end

The target computer installed the original gui application. Then installed the new compiled run_parallel_application function. Then, opened up run_parallel_application.exe. This didn't work either.

[Adam]

They definitely don't need the Parallel Computing Toolbox, that is absolute 100% certain. Executables are to run on machines that do not require base Matlab.

I don't use

parpool local

so that may be a difference. I'm not familiar with exactly what it does, I simply call

parpool( )

to open a parallel pool so it will use a default cluster and this appears to work fine on target machines with no extra work to be done. It may be that 'local' does need to you include something else in the executable that is not there, but that would be beyond my knowledge of the subject!

[Rik]

Comment posted as answer by Akanksha Mishra:

@Tahoe Schrader

Hi ,

i am facing the same issue with my GUI build executable file.

Did you find a solution for your problem then ?

Please care to share.

Thanks.