MATLAB On High Performance Computing (HPC)

Hello all,

I have a python code that calles a MATLAB compiled file many times during a single run to get information from a model built in MATLAB. When I run this code on Windows, it doesn't take time since I have MATLAB installed on my machine. To use the speed of HPC, when I run the model there, it takes a very long time to get my result (triple the speed on Windows). I wanted to ask if there is anyway to speed the process. I tried using MCR but still, it is quite slow

Answer 1

[Ayush Kashyap]

Hi Mohamed,

To reslove the issue of speed of performance, you make try these two ways mainly:

• You may try out parallelization and other techniques of optimizing your code and see if your performance improves.
• If the problem still persists, then it must be due to delays in launching Matlab Runtime Environment multiple times. In this case you can try caching by using a Python Dictionary to store the results and try to avoid multiple access to the original Matlab file.

I am attaching a simple example of implementing caching using python dictionary:

def cache_function(*inputs):
    if inputs in cache_dict:
        return cache_dict[inputs]
    else:
        result = call_matlab_compiled_file(*inputs)
        cache_dict[inputs] = result
        return result

Don't forget to replace your direct calls to Matlab files as follows:

   cache_dict = {}
   
   result = cache_function(input1, input2, ...)

Hope this helps resolve your problem.

Alternatively, you may also try out other libraries like 'functools.lru_cache' or 'cachetools'.