How to make up the diagonal summation without for loop and with the fastest way!

A =

     1     7    13   
     2     8    14   
     3     9    15   
     4    10    16   
     5    11    17   
     6    12    18

B =

    1, (2+7), (3+8+13), (4+9+14), (5+10+15)  ... (12+17), 18

How to make the B vector?

Vector-wise would be good to perform the enhancement of computational time?

Answer 1

[Andrei Bobrov]

A =[ 1     7    13   
     2     8    14   
     3     9    15   
     4    10    16   
     5    11    17   
     6    12    18];

B = sum(spdiags(rot90(A)));

[Sean de Wolski]

Nice!

[Andrei Bobrov]

Thank you Sean.

Answer 2

[Azzi Abdelmalek]

 A =[ 1     7    13
      2     8    14
      3     9    15
      4    10    16
      5    11    17
      6    12    18]
[n,m]=size(A);
B=fliplr(A);
out=zeros(1,m+n-1);
for k=-n+1:m-1
    out(k+n)=sum(diag(B,k));
end
out=fliplr(out)

Answer 3

[Mischa Kim]

Hyunchul, something like

 fA = fliplr(A);
 B  = fliplr(arrayfun(@(i) sum(diag(fA,i)),-length(A(:,1))+1:length(A(1,:))-1));

Answer 4

[Sean de Wolski]

And just for fun (requires Image Processing):

B = zeros(size(A));
B(1) = 1;
B2 = bwdist(B,'cityblock')+1;
v = accumarray(B2(:),A(:))

Use Azzi's solution.