Column vector multiplied to a row vector, memory v.s. speed issue

Hello everyone. I have a column vector (A) that is left multiplied to a row vector (B). The result is a matrix C=A*B. What I need is the sum of C in each row, that is sum(C,2). However, both A and B are very long vector that makes C rather huge that exceeds memory. Is it possible to bypass the matrix C and get the results sum(C,2) without using a loop, which is a bit slow.

Many thanks

[Birdman]

 sum(C,2)

calculates the sum of each row in resultant C matrix.

[Xin]

Thanks for pointing it out. I have corrected my question.

[Birdman]

Is it possible to bypass the matrix C

What do you mean by bypass?

[Xin]

Don't calculate C, which is memory consuming. Is it possible to find a fast way to get the row sum of A*B?

[Birdman]

What are the sizes of A and B?

[Xin]

Depending on the data, A can be more than 1e4, and B can be more than 1e6~1e7.

Answer 1

[Stephen23]

Loop over A, it won't take very long as long as long as the output is preallocated. Replace this:

D1 = sum(A*B,2)

with

N = numel(A);
D2 = nan(N,1);
for k = 1:N
    D2(k) = sum(A(k)*B);
end