What’s new in PyPy xxx

Fixed the performance of gc.get_referrers()

Provides cppyy module (disabled by default) for access to C++ through Reflex. See doc/cppyy.rst for full details and functionality.

Check that axis arg is valid in _numpypy

Remove numpy lazy evaluation and simplify everything

Support for array[array-of-ints] in numpy

Implement better JIT hooks

Improve handling of **kwds greatly, making them virtual sometimes.

Introduce __int128 on systems where it’s supported and improve the speed of rlib/rbigint.py greatly.

Start to clean up a bit the flow object space.

Support CFFI. http://morepypy.blogspot.ch/2012/08/cffi-release-03.html

The stdlib was updated to version 2.7.3

Complex dtype support for numpy

Improve dtypes intp, uintp, void, string and record

Add float16 numpy dtype

major cleanups including killing some object support

implement threadstate-related functions in cpyext

add dict/list/set strategies optimized for unicode items

Support for utf-8 encoding in RPython

Support ARM in the JIT.