according to unladen swallows' bm_unpack_sequence benchmark:
- ~25% faster for tuples (60% faster than CPython 2.7)
- ~3x faster for lists (40% faster than CPython 2.7)
also, apparently using less memory during gcc compilation

--HG--
extra : rebase_source : 1069d7a0c9c338c16e96cc289114e0d0052a7588

--HG--
extra : rebase_source : 18e87455b812d7b9a1f2f2b28593455ca5521e89

OpenMP support, see https://github.com/cython/cython/pull/28 and https://github.com/markflorisson88/cython/tree/openmp

slightly faster for-in iteration for arbitrary iterables and generators by caching 'tp_iternext' function pointer