Inefficient C code generated from Real Time Workshop Embedded Coder

I am using Matlab R2008a on Mac.

Here is some example code to illustrate how inefficient the generated C code is:

function [dataStruct,dataStruct2,csum,csum2] = testFunc() %#eml
dataStruct = struct('array',ones(10,3,5),'csum',zeros(10,3,5));
dataStruct2 = struct('array',ones(10,3,5),'csum',zeros(10,3,5));
csum = zeros(10,3,5);
csum2 = zeros(10,3,5);
vars = [1 2 3 4 5];
divisor = [1 2 3];
for i = 1:5
    for k = 1:3
        dataStruct2.array(vars(i),k,i) = dataStruct2.array(vars(i),k,i)+3;
        dataStruct2.array(:,k,i) = dataStruct2.array(:,k,i)/divisor(k); *bold*
    end
end

Here is the generated C code corresponding to the division step:

for(eml_i0 = 0; eml_i0 < 10; eml_i0++) {
        eml_b_dataStruct2[eml_i0] = eml_dataStruct2->array[(eml_i0 + 10 * eml_k)
          + 30 * eml_i] / (real_T)eml_B;
}
for(eml_B = 0; eml_B < 10; eml_B++) {
        eml_dataStruct2->array[(eml_B + 10 * eml_k) + 30 * eml_i] =
          eml_b_dataStruct2[eml_B];
}

This code is creating a copy of the output in eml_b_dataStruct2, and then copies that into the output variable eml_dataStruct2->array. I would like for the generated code to look like this:

for(eml_i0 = 0; eml_i0 < 10; eml_i0++) {
        eml_dataStruct2->array[(eml_i0 + 10 * eml_k)
          + 30 * eml_i] = eml_dataStruct2->array[(eml_i0 + 10 * eml_k)
          + 30 * eml_i] / (real_T)eml_B;
}

Is this something that has been fixed in newer versions of Matlab?

Regards