reducing number of operations in matrix multiplication

I want to perform the follow operation: synch = spect1(1:num_pts ,2)*spect2(1:num_pts,2)'; The problem is that num_pts > 300,000. Would take too long. Fortunately, I don't need the entire output. I only need the diagonal elements plus elements from the say 10 adjacent diagonals on either side. Anybody know how to do this or want to take a guess? Thanks

Answer 1

[Douglas Brenner]

    testm = [1,2,3,4,5;2,4,5,7,10]
    num_pts = 5;
    testm = transpose(testm)
    corr = testm(:,2)*testm(:,2)'
    diag(corr)
    main_diag = testm(:,2).*testm(:,2)
    for i=1:3
      diags1{i} = testm(i+1:num_pts ,2).*testm(1:num_pts-i-1,2);
      diags2{i} = testm(i+1:num_pts ,2).*testm(1:num_pts-i-1,2);
    end

failed at diags1{i} = testm(i+1:num_pts ,2).*testm(1:num_pts-i-1,2); Error using .* Matrix dimensions must agree. Error in test (line 8) diags1{i} = testm(i+1:num_pts ,2).*testm(1:num_pts-i-1,2);

Got to run.

[Torsten]

The number of elements of testm(i+1:num_pts,2) and testm(1:num_pts-i-1,2) are not the same ...

Answer 2

[Douglas Brenner]

Clearly but that was the suggested solution. Trying for i=1:2 diags1{i} = testm(i+1:num_pts-i-1 ,2).*testm(1:num_pts-i-1,2); diags2{i} = testm(i+1:num_pts-i-1 ,2).*testm(1:num_pts-i-1,2); end

gives the same error as does other attempts to fix it