failed to generate MEX function, generation of trial code failed

I am trying convert my matlab code in order to speed up thye computational time.

let me describe my function is, first I read a filename with some extention name and then passing into low pass filter and get the dominant freq, also calculate instatneous freq and calculate permutation entropy. I need to speed up my calculation because my input file is 24 hour with 100 sampling rate and for each month it contain 2232 files. I will do more than 3 years.

I use matlab coder to transfor to max file, but when it arrived at check for runtime issues, it said that generation of trial code is failed. I don't know why because it doesn't show anything in "Target Build Log" part.

Almost forget to mention that in command window it appear message eventough Matlab Coder window appear

Caught "std::exception" Exception message is:
Unable to load bundle binary /usr/local/MATLAB/R2020b/bin/glnxa64/builtins/shared_dastudio_builtins/mwdastudio_builtinimpl.so. Error: /lib64/libharfbuzz.so.0: undefined symbol: FT_Get_Transform: Success

here I include my function

function  [time_exec]=	calcPE_test(mfilename)
%#codegen

list=dir([mfilename '/*.DAT']);
N = length(list);    
PEval=cell(1,N);
PEvalIF=cell(1,N);
domfreq=cell(1,N);
tStart=tic;

for m=1:length(list)
    thisfile = list(m).name ;
    if(coder.target('MATLAB'))
    sig = importdata(thisfile);
    end
    
   %lowpass filter (<7Hz)
   fc=7; %frequency corner 
   fs=100; %frequency sampling
   Wn = fc/(fs/2);
   n=5; %filter order, check filter response using 
   [b,a] = butter(n, Wn, 'low'); %lowpass filter
   sigfil= filtfilt(b,a,sig); %sigfill is filtered signal
   
   % calculate fft
   n=length(sigfil);
   srate=100; %sampling rate 
   nyq=srate/2;    
   dataX=fft(sigfil-mean(sigfil))/n; %dataX is fft result from filtered signal
   hz=linspace(0,nyq,floor(n/2)+1);
   pow=2*abs(dataX(1:length(hz)));
    
   %get the maximum freq/dominant freq
   [~,idx] = max(pow) ;
   maxFreq = hz(idx);
   domfreq{m}=maxFreq; 

   %Hilbert transform to get real and imag part
   hilsig = hilbert(sigfil);
   inphase=atand(imag(hilsig)./real(hilsig)); %inphase is instantaneous phase
   
   %calculate permutation entropy from filtered signal and IF
   PEfilt = permuentropy_test(sigfil,5,3);
   PEval{m}=PEfilt;

   PEIF = permuentropy_test(inphase,5,3);
   PEvalIF{m}=PEIF;
   
end

%writting result in array and save in ascii
if(coder.target('MATLAB'))
dlmwrite ('PE_filt.dat',  PEval)
dlmwrite ('PE_IF.dat', PEvalIF)
dlmwrite ('dom_freq.dat', domfreq)
end
tEnd = toc(tStart);
time_exec=datevec(tEnd./(60*60*24)); %change time in tictoc to  

%PLOTTING PART
fig=figure();
%plot PE vs PE_IF
subplot 211
PE=load('PE_filt.dat'); IF=load('PE_IF.dat');
plot(PE,'o-')%,'markerfacecolor','b')
hold on
plot(IF,'o-')%,'markerfacecolor','r')
title('Comparation PE of filtered signal and PE of Instantaneous Phase')%,'(both after lowpassfiltered applied)')
%legend('PE from ori signal','PE from filtered IF', 'location','southeast')
ylabel('PE value')
%plot dominant freq
subplot 212
freq=load('dom_freq.dat');
plot(freq,'o-','markerfacecolor','b')
title('Dominant Frequency')
xlabel('time(*20minutes)')
ylabel('frequency (Hz)')
if(coder.target('MATLAB'))
saveas(fig,'PE_IF_DominantFreq_test.png');
end
end

[SACHIN KHANDELWAL]

Hi @nirwana,

I couldn't exactly replicate the issue, but I found this MATLAB Answers post that might help you:

https://www.mathworks.com/matlabcentral/answers/504980-how-do-i-resolve-bundle-error-messages

Thanks,

Sachin

Answer 1

[Naman Saraf]

Hi @nirwana,

Can you provide the details on the input to the code generation process? Also, a suggestion, you can only generate MEX for the processing part of the system, i.e. filtering etc. and keep the data import and export part in MATLAB.