Vectorised insertion sort algorithm slower than regular one?

Hello,

My goal is to vectorise a selection sort algorithm so that it performs better than a regular one with nested for loops. Despite extensive efforts, the vectorised version is 2 times slower, and I cannot find a way to fix it.

The regular algorithm was easy:

function sorted = insertionsort(A)   
for i = 2:length(A)
    value = A(i);
    j = i-1;
    insert = true;
    while insert == true
        if A(j) > value
            A(j+1) = A(j);
            j = j-1;
            if j < 1
                insert = false;
            end
        else
            insert = false;
        end
    end
    A(j+1) = value;
end

To sort rand(1,5000), it takes about 0.07 seconds. This is the fastest vectorised version I have written:

function sorted = insertsort(a)
% An insertion sort algorithm that orders a vector from smallest to largest
% values. Input any vector.
a                                 % display original
b = zeros(1,length(a));           % preallocate solution vector
for i = 2:length(a)
  f = find((a(1:i-1) > a(i)),1);% index of first element larger than a(i)
  if ~isempty(f)                % only shift elements if a larger one exists
      b(f) = a(i);              % insert a(i) into place
      b(f+1:i) = a(f:i-1);      % place larger elements one place to the right
  else
      b(i) = a(i);              % simply insert a(i) into b(i)
  end %(if)
  a(1:i)=b(1:i);                % update sorted section of a for next loop
end %(for)
sorted = a;                       % display sorted vector
end %(function)

To sort rand(1,5000), on average it takes about 0.12 seconds. More specifically, the find component takes 0.06 seconds while the insertion takes 0.06 seconds. My lecturer said that his vectorised version is 5 times faster, so there must be some shortcut I can make.

Any hints? Thanks.

[Stephen23]

"the vectorised version is 3 times slower"

The times that you give are "0.7 seconds" for the non-vectorized, and "about 0.17 seconds" for the vectorized version.

Surely 0.17 s is faster than 0.7 s ?

Also note that the output should be a or b, and not sort, and that using the name sort is a very very bad idea because that will shadow the very important inbuilt sort function.

[Rubidium]

Yes, I accidentally put 0.7 instead of 0.07. I have fixed that now. Thank you