Matlab engine for python with shared session

Hello,

I want to run multiple matlab engines for python at the beginning of my application, because the library I developed embeds many large functions that have to be loaded at the beginning of the app. This is well described in the documentation:

# initialization
engines = []
# load engines
for i in range(3):
    engines.append(matlab.engine.start_matlab())
# use one engine
result = engines[1].MyFunction(arg1,arg2)

I also found in the documentation the possibility to use shared matlab sessions:

 matlab.engine.shareEngine('MATLABEngineName')
 engine = matlab.engine.connect_matlab('MATLABEngineName')

What is this?! There is no information about these shared sessions.

Does it allow to simultaneously run multiple functions in the same engine? What about the execution times (for comparable configurations between shared and classic sessions)?

Thanks

Answer 1

[Bo Li]

You can send multiple requests to the same engine, they will be queued up and executed one by one. To run multiple functions simultaneously, you may start multiple MATLAB through matlab.engine.start_matlab in Python, or launch multiple shared MATLAB session and connect to them using matlab.engine.connect_matlab. There is no difference between a shared and classic session regarding the performance of running a MATLAB function.

Following MATLAB command can be used to share a MATLAB session:

>>matlab.engine.shareEngine

Or you can launch a MATLAB session as shared from system prompt using "-r" argument:

%matlab -r matlab.engine.shareEngine

Reference:

http://www.mathworks.com/help/matlab/ref/matlab.engine.shareengine.html

[Alessandro]

Hi @Bo Li, I was investigating the use of the matlab engine with python. According to this answer it is possible to connect from Python to multiple shared Matlab sessions. This means that I can simulate models in parallel?