Speeding up for loop with if statement

Hi all,

Got what I believe is a noob question here: I have a function and it seems that 90% of my runtime is taken up by one loop. I'm wonderin if you guys can help me bring that down/out

The code is as follows (I've replaced the actually O and gamma variables with rands)

Note I originally wrote this with a lot of for loops as I was going to translate it to C and wanted the proting to be straight forward, but now I have this huge bottlneck (I would still prefer to solve it in a way that translates easily to C --- when I get time to port the project :P).

N = 4;
T = 3000000;
M = 170;
O = randi(M,1,T);
gamma = rand(N,T);
B = zeros(N,M);

for j = 1:N
    denom = 0;
    for t = 1:T
        denom = denom + gamma(j,t);
    end %t
    for k = 1:M
        numer = 0;
        for t = 1:T
            if(O(t) == k)
                numer = numer + gamma(j,t);
            end
        end %t
        B(j,k) = numer/denom;
    end %k
end %i

[Walter Roberson]

accumarray(). You could probably do all of numer calculations at the same time. denom = sum(gamma, 2) in vectorized form

[Josh Parks]

Thanks, in case anyone was wondering about the solution:

denom = sum(gamma,2);

for j = 1:N
    numerB = accumarray(O',gamma(j,:)',[M 1],@sum)';
    B(j,:) = numerB./denom(j);
end %i