I actually like both this and the "1000 fold performance improvement" articles. There is a problem, and then there is a solution, which works, and is faster in a spectacular way.
However, both articles mention that a C++/STL solution would bring in a vtable runtime penalty. This is just not true - quite the opposite: using C functions a la qsort(3) require you to explicitely pass in a pointer to the comparison function. Now: this is the vtable approach, but in plain sight. A C++ version would quite likely not call any functions via pointers.
However, both articles mention that a C++/STL solution would bring in a vtable runtime penalty. This is just not true - quite the opposite: using C functions a la qsort(3) require you to explicitely pass in a pointer to the comparison function. Now: this is the vtable approach, but in plain sight. A C++ version would quite likely not call any functions via pointers.
I prepared a small test which shows that in simple cases C++ can actually be faster than C by a factor of up to 3: http://radiospiel.org/sorting-in-c-3-times-faster-than-c