Vectorizing interdependent nested 4-for loop

How to speed up this piece of code? How to vectorise this code? Is no other way possible?

 for i1=1:num_agents % for all i
    for rows=1:num_tasks*num_parts-1 % for all m_i,n_i in E_i
        for ell=1:gL-G(rows,rows+1,i1) % for all ell
            for i2=1:num_agents
                xtemp = clearer;
                % half-sum 1st part
                for s = ell:gL%-G(rows,rows+1,i1)
                    if mod(rows,num_tasks)~=0
                        xtemp(i1,ceil(rows/num_tasks),mod(rows,num_tasks),s) = 1;
                    elseif mod(rows,num_tasks)==0
                        xtemp(i1,ceil(rows/num_tasks),num_tasks,s) = 1;
                    end
                end

                % half-sum 2nd part
                for s = 1:ell+G(rows,rows+1,i1)-1
                    if s <= gL
                        if mod(rows,num_tasks)~=0
                            xtemp(i2,ceil(rows/num_tasks),mod(rows,num_tasks)+1,s) = 1;
                        elseif mod(rows,num_tasks)==0
                            xtemp(i2,ceil(rows/num_tasks)+1,1,s) = 1;
                        end
                    end
                end
                xtemp = xtemp(:);
                Aineq(counter,:) = xtemp';
                bineq(counter) = 1;
                counter = counter + 1;
            end
        end
    end
end

Answer 1

[Rick Rosson]

1. Did you pre-allocate Aineq prior to the loop?
2. Did you pre-allocate bineq prior to the loop?

[ingstalam]

Yup.

Aineq = spalloc(J*K+J*K+I*ELL+I*I*gL*(J*K-1),matwid,3*I*J*K*ELL+I*I*gL*(J*K-1)*(G_max)); % Allocate  Aineq
bineq = zeros(J*K+J*K+I*ELL+I*I*gL*(J*K-1),1); % Allocate bineq

[ingstalam]

In fact i used sparse matrix allocation to speed up the results.