This helps to speed up the calls to count the utf8 code points (e.g. len(unicodestr)). On the benchmark on python2 this is pretty insignificant. Doing a lot of string length calls in a loop on unicode objects shows speedup. The effect would be much more interesting on Python 3.X