C++

A simple question...

is it better to traverse a std::vector like this:

std::vector<int> myvec;
//populate the vector
for(int i = 0; i < myvec.size(); ++i){
    myvec[i] += 1;
}

or is better like this?

std::vector<int> myvec;
typedef std::vector<int>::iterator IT_int;
for(IT_int it = myvec.begin(); it != myvec.end(); ++it){
     *it += 1;
}

>>24
>>25
Why is that wrong? On x86 gcc uses cmp to implement < and test-instruction to implement !=. What else do think causes the difference in speed if the only difference in generated loops is that one uses cmp and one uses test?