Simple for loop taking way longer than it should

I am running a for loop on 10 million lines of data and it has taken over a day and is still running. I have preallocated the array I am modifying according to suggestions provided by Matlab for optimizing runtime. This post says 100 million lines of data ran in a simple for loop in 0.2 seconds. I am testing my code both on my personal computer (Mac) and a remote server. Why is my code so slow??

%% read in file
cwaFileToRead = '6011549_0000000003.cwa';
ptID = '6011549_0000000003';
sampling_rate = 100;
Fs = sampling_rate;
output_dir = '';

rawData = CWA_readFile(cwaFileToRead, 'verbose', 0);

%% calculate vector magnitude
xRaw = rawData.AXES(:,2);
yRaw = rawData.AXES(:,3);
zRaw = rawData.AXES(:,4);
vmRaw = sqrt((xRaw.^2)+(yRaw.^2)+(zRaw.^2));
vmRaw = vmRaw.*(1000);

numrows=size(vmRaw); 
peak=zeros(numrows);%add empty rows for peak identification

time = datetime(rawData.AXES(:,1), 'ConvertFrom', 'datenum'); % convert mtime to datetime
timedata = timetable(time, xRaw, yRaw, zRaw, vmRaw, peak); % convert data to timetable

 %% CHECK FOR UNIFORM SAMPLING, RESAMPLE IF NOT UNIFORM
    % Check if sampling rate is uniform
    isUniform = isregular(timedata);
    % If sampling rate is non-uniform, resample at regular sampling intervals
    if isUniform == 0
        timeStep = seconds(1/Fs); % how long between each regular sample
        regularizedData = retime(timedata,'regular','linear','TimeStep',timeStep);
        fprintf('Sampling rate is non-uniform, resampling using linear interpolation\n')
    else
        regularizedData = timedata;
    end
    
%FOR LOOP THAT IS SO SLOW!! HELP!
tic
for row = 2:numrows-1 
    if regularizedData.vmRaw(row)>regularizedData.vmRaw(row-1)&& regularizedData.vmRaw(row)>regularizedData.vmRaw(row+1)
        regularizedData.peak(row)=1;
    else
        regularizedData.peak(row)=0;
    end
end
toc

[Ive J]

can you share regularizedData struct?

[Samantha Huff]

head(regularizedData)

ans =

  8×5 timetable

            time             xRaw        yRaw        zRaw      vmRaw     peak
    ____________________    _______    ________    ________    ______    ____

    13-Nov-2021 10:00:05     1.5724    -0.93819       1.077      2111     0  
    13-Nov-2021 10:00:05     1.1085    -0.80026     0.60829    1503.3     0  
    13-Nov-2021 10:00:05     0.6409    -0.71118     0.14545    1000.7     0  
    13-Nov-2021 10:00:05    0.32321    -0.50765    0.080661    625.47     0  
    13-Nov-2021 10:00:05    0.27804    -0.06346     0.20332     439.7     0  
    13-Nov-2021 10:00:06     0.4884     0.48265      0.1555    723.28     0  
    13-Nov-2021 10:00:06     0.8808     0.77716     0.14205    1188.9     0  
    13-Nov-2021 10:00:06     1.2542     0.73893     0.11437    1467.7     0  

Answer 1

[Matt J]

I don't see why it would be slow, but there's no need for a for loop at all,

regularizedData.peak(2:end-1) = regularizedData.vmRaw(2:end-1)>regularizedData.vmRaw(1:end-2)&...
                                regularizedData.vmRaw(2:end-1)>regularizedData.vmRaw(3:end);      

[Steven Lord]

Or you could just use islocalmax.

y = randi(50, 1, 10)

y = 1×10

33 47 24 14 33 16 47 28 32 16

[   y; ...
    islocalmax(y); ...
    false (y(2:end-1) > y(1:end-2)) & (y(2:end-1) > y(3:end)) false]

ans = 3×10

33 47 24 14 33 16 47 28 32 16 0 1 0 0 1 0 1 0 1 0 0 1 0 0 1 0 1 0 1 0