How can I make a circshift function, without using circshift nor shift matlab functions?

I'm trying to make a circshift so I can get an efficient convolution program, and I have already obtained this by using the code below, however 'my circshift' is not efficient enough given that I'm trying to get the convolution between a pair of (1x70000) matrices.

I'd like to know if you can give me an idea of how can I improve my circshift. Thanks.

   %CODE
%Here we introduce the two discrete signals 
a=[1,2,3,4,5,6];
b=[1,2,3];

%We flip the second signal
bc=fliplr(b);
%We are going to make a circular shifting and a dot product, so we
%need nxn matrices. Here we assure matrix A gets as many leading zeros as elements in
%B, and matrix B gets as many leading zeros as elements in A to get a pair
%of nxn matrices.

 A = [a,zeros(1,length(b)-1)];
 B = ([bc, zeros(1, length(a)-1)]);
 convol = [1,zeros(1,length(A)-1)];

    %HERE IS MY PROBLEM
    %This is 'my circhsift'

    for i=1:length(a)-1; i; fb=fliplr(B); f=fb(1);
    nb=wkeep(fb,length(B)-1,'r');fnb=fliplr(nb); B=[f,fnb]; 
    end
    BB=B;

    for i=1:length(A)
    A;
    i;
    fb=fliplr(BB);
    f=fb(1);
    nb=wkeep(fb,length(B)-1,'r');
    fnb=fliplr(nb);
    BB=[f,fnb];
    convol(i)=dot(A,BB);
    end
    convol;

[Walter Roberson]

wkeep is going to have overhead, so it would be more efficient to write the code more directly without using wkeep.

Answer 1

[Andrei Bobrov]

a = randi(100,1,12)
n = 5;

out1 = circshift(a,n)

m = numel(a);
out2 = a(mod((1:m) - n - 1,m) + 1)

[Heber Alvarez]

Thank you Andrei Bobrov!