symbolic units: sym.saveobj bottleneck

1 view (last 30 days)
Ben Mercer
Ben Mercer on 2 Aug 2021
Hi all,
I'm experiencing a quite significant bottleneck when saving data which includes symbolic measurement units. I'm using symbolic units (e.g. symunit('s')) to represent all of my measurement units - it seemed like a pretty neat feature at face value! Here is some code to demonstrate the issue:
%% Create some dummy data
clear
N = 30;
for n1 = 1:N
myData(n1).x = (0:0.01:5).';
myData(n1).xUnit = symunit('s');
% myData(n1).xUnit = sym('a');
myData(n1).y = sin(2*pi*(1+rand) * myData(n1).x );
myData(n1).yUnit = symunit('V');
% myData(n1).yUnit = sym('b');
end
%% Save to a file - takes ~20s
profile on
save('myData.mat', 'myData')
profile report
%% Save again to another file - takes ~0.2s
profile on
save('myData2.mat', 'myData')
profile report
So for 30 channels of data with a time and voltage unit expressed as symbolic, saving to a file will take ~20s. If we save the data again to another file, it is much faster at ~0.2s. If we run again from the start (clearing the data in the process), it will take ~20s again. If we replace the symunit() calls with sym() (comment out the symunit calls and uncomment the sym calls), it takes ~0.2s again. So I gather that A) the issue is specific to symbolic units, not symbolics in general, and B) the result of the expensive computation is held persistently and is not re-done on a second call.
I'm running MATLAB R2019B, but can update to a later release if necessary.
I'd be very grateful if anyone can help understand:
  • Is this bottleneck addressed in a later version of Matlab?
  • Are there any possible workarounds?
Many thanks,
Ben

Answers (0)

Products


Release

R2019b

Community Treasure Hunt

Find the treasures in MATLAB Central and discover how the community can help you!

Start Hunting!

Translated by