Save variable to a matrix after a for loop

Hey guys this is probably very easy, but Im having troubles doing it.

I have this code, that cross-correlates two signals, from 2 matrixes. I would like to save all the outputs of the function in 3 matrixes(afmag3,delay3,doppler3). Also the code, takes too long to run , I dont know if this can be better. Thank you

[n,m]=size(surv_matrix);
[n2,m2]=size(ref_matrix);

for i=1:m
    for ii=1:m2
        Reference_SignalCut=ref_matrix(:,ii);
        Surveillance_SignalCut=surv_matrix(:,i);
        [afmag3,delay3,doppler3] = ambgfun(Reference_SignalCut,Surveillance_SignalCut,fs,[1e6 1e6]);
        afmag3 = afmag3*1; % Select plot gain *1
        afmag3(afmag3>1 )= 1;
    end
end

Answer 1

[Jan]

[n, m]   = size(surv_matrix);
[n2, m2] = size(ref_matrix);

afmag3   = cell(m, m2);
delay3   = cell(m, m2);
doppler3 = cell(m, m2);
for i=1:m
    S = surv_matrix(:,i);      % Surveillance of cut signal
    for ii=1:m2
        R = ref_matrix(:,ii);  % Reference of cut signal
        [tmp, delay3{i, ii}, doppler3{i, ii}] = ambgfun(R, S, fs, [1e6 1e6]);
        afmag3{i, ii} = max(1, tmp);
    end
end

Concerning the run time, use the profile 'r to finde the bottleneck. If ambgfun could be vectoprized maybe a speedup is possible.

I've moved the definition of Surveillance_SignalCut out of the inner loop. This does not save a lot of time, but it is a good programming practice not to do any work repeatedly. afmag3 = afmag3*1 was a waste of time only.

Avoid complicated names for variables, because this impedes the reading. Explain the meaning of a variable in a comment instead.

[Miguel Albuquerque]

thanks a lot :D

[Miguel Albuquerque]

I have here another code, do you think I could improove this part also?

[n, m]   = size(surv_matrix);
[n2, m2] = size(ref_matrix);

[time_SS,Surveillance_Signal]=deal(nan(n,m));
for i=1:m
    [time_SS,Surveillance_Signal]=freq2time(m,doppler_freqSurv);
end

[time_RS,Reference_Signal]=deal(nan(n,m1));
for i=1:m1
    [time_RS,Reference_Signal]=freq2time(m1,doppler_freqRef);
end

[Miguel Albuquerque]

I have here another code, do you think I could improove this part also?

[n, m]   = size(surv_matrix);

Unrecognized function or variable 'surv_matrix'.

[n2, m2] = size(ref_matrix);

[time_SS,Surveillance_Signal]=deal(nan(n,m));
for i=1:m
    [time_SS,Surveillance_Signal]=freq2time(m,doppler_freqSurv);
end

[time_RS,Reference_Signal]=deal(nan(n,m1));
for i=1:m1
    [time_RS,Reference_Signal]=freq2time(m1,doppler_freqRef);
end

[Jan]

[time_SS,Surveillance_Signal]=deal(nan(n,m));

% Looks more complicated and takes longer than:
time_SS             = nan(n, m);
Surveillance_Signal = nan(n, m);

I'd omit the "_Signal". Could it be anything else than a signal? Short names reduce the clutter.

Why do you use NaNs to pre-allocate the array? All values are overwritten, so zeros are cheaper: zeros(1e6, 1) is 50 times faster than NaN(1e6, 1);

What is freq2time()? Could it be vectorized? If so, you could omit the loops. But the loops do not anything at all, because the loop index i does not occur inside the loop anywhere.

If yourt code produces the correct output, simplify:

[time_SS,Surveillance_Signal]=deal(nan(n,m));
for i=1:m
    [time_SS,Surveillance_Signal]=freq2time(m,doppler_freqSurv);
end

to

[time_SS,Surveillance_Signal]=freq2time(m,doppler_freqSurv);
% No loop, no pre-allocation!

[Miguel Albuquerque]

Thanks a lot for the help :D