How can I vectorize or otherwise speed up this function?

I have a function intended to do some initial calculations on an array of test data collected during mechanical fatigue tests run under cyclic load control. It's formatted in 3 columns of time, position and load (each row is a single measurement) The datasets are huge... sometimes 20 million rows or more. Can anyone think of good ways to vectorize these FOR loops? The 2nd loop is especially inefficient, but I'm new to coding with vectors and am having trouble wrapping my head around it.

Thanks for any help or resources you can provide!

function [test,cycleposition] = fatigueimport(test)
% Get length of test and find derivative of load to count cycles
% test data is array with time, position, load in columns

cycle = zeros(length(test),1); %initiate cycle variable
cyclecount=1;
cycle(1)=1;
changeload = diff(test(:,3));
count = length(test)-1
for i=2:count 
    if and(changeload(i-1)>0,changeload(i)<0)
        % tests for point where load changes from increasing
        % (positive changeload) to decreasing (negative changeload) and 
        % increments cyclecount.
        cyclecount=cyclecount+1
    end
        cycle(i)=cyclecount;
        i=i+1;
end
test = [test,cycle]; % adds calculated cycle to original array

% Find position of greatest deflection for each cycle
count = max(test(:,4)) % count set to total number of cycles
cycleposition = zeros(count,1); % initiate vector to record deflection values
for i=1:count
    cycleposition(i)=min(test(test(:,4)==i,2));
    i=i+1;
end
end

Answer 1

[Matt J]

    function [test,cycleposition] = fatigueimport(test)
    % Get length of test and find derivative of load to count cycles
    % test data is array with time, position, load in columns

      changeload = diff(test(:,3));

      changes=[ 1 ;  changeload(1:end-1)>0 & changeload(2:end)<0 ; 0];

     cycle=cumsum(changes);

     test=[test,cycle];

     cycleposition=accumarray(cycle, test(:,2) , [],@min);

[emm0119]

Thank you so much, this is perfection! What took hours now takes mere seconds.

[Matt J]

You're quite welcome. Welcome to Matlab.