Wednesday, April 20, 2016

PyPy 5.1 released

We have released PyPy 5.1, about a month after PyPy 5.0.

This release includes more improvement to warmup time and memory requirements, extending the work done on PyPy 5.0. We have seen an additional reduction of about 20% in memory requirements, and up to 30% warmup time improvement, more detail in the blog post.

We also now have full support for the IBM s390x. Since this support is in RPython, any dynamic language written using RPython, like PyPy, will automagically be supported on that architecture.

We updated cffi to 1.6 (cffi 1.6 itself will be released shortly), and continue to improve support for the wider python ecosystem using the PyPy interpreter.

You can download the PyPy 5.1 release here:
We would like to thank our donors for the continued support of the PyPy project.
We would also like to thank our contributors and encourage new people to join the project. PyPy has many layers and we need help with all of them: PyPy and RPython documentation improvements, tweaking popular modules to run on pypy, or general help with making RPython’s JIT even better.

What is PyPy?

PyPy is a very compliant Python interpreter, almost a drop-in replacement for CPython 2.7. It’s fast (PyPy and CPython 2.7.x performance comparison) due to its integrated tracing JIT compiler.

We also welcome developers of other dynamic languages to see what RPython can do for them.

This release supports:
  • x86 machines on most common operating systems (Linux 32/64, Mac OS X 64, Windows 32, OpenBSD, FreeBSD),
  • newer ARM hardware (ARMv6 or ARMv7, with VFPv3) running Linux,
  • big- and little-endian variants of PPC64 running Linux,
  • s390x running Linux

Other Highlights

(since the release of PyPy 5.0 in March, 2016


  • New features:

    • A new jit backend for the IBM s390x, which was a large effort over the past few months.
    • Add better support for PyUnicodeObject in the C-API compatibility layer
    • Support GNU/kFreeBSD Debian ports in vmprof
    • Add __pypy__._promote
    • Make attrgetter a single type for CPython compatibility

  • Bug Fixes

    • Catch exceptions raised in an exit function
    • Fix a corner case in the JIT
    • Fix edge cases in the cpyext refcounting-compatible semantics (more work on cpyext compatibility is coming in the cpyext-ext branch, but isn’t ready yet)
    • Try harder to not emit NEON instructions on ARM processors without NEON support
    • Improve the rpython posix module system interaction function calls
    • Detect a missing class function implementation instead of calling a random function
    • Check that PyTupleObjects do not contain any NULLs at the point of conversion to W_TupleObjects
    • In ctypes, fix _anonymous_ fields of instances
    • Fix JIT issue with unpack() on a Trace which contains half-written operations
    • Fix sandbox startup (a regression in 5.0)
    • Fix possible segfault for classes with mangled mro or __metaclass__
    • Fix isinstance(deque(), Hashable) on the pure python deque
    • Fix an issue with forkpty()
    • Issues reported with our previous release were resolved after reports from users on our issue tracker at https://bitbucket.org/pypy/pypy/issues or on IRC at #pypy

  • Numpy:

    • Implemented numpy.where for a single argument
    • Indexing by a numpy scalar now returns a scalar
    • Fix transpose(arg) when arg is a sequence
    • Refactor include file handling, now all numpy ndarray, ufunc, and umath functions exported from libpypy.so are declared in pypy_numpy.h, which is included only when building our fork of numpy
    • Add broadcast

  • Performance improvements:

    • Improve str.endswith([tuple]) and str.startswith([tuple]) to allow JITting
    • Merge another round of improvements to the warmup performance
    • Cleanup history rewriting in pyjitpl
    • Remove the forced minor collection that occurs when rewriting the assembler at the start of the JIT backend
    • Port the resource module to cffi
     
    • Internal refactorings:

      • Use a simpler logger to speed up translation
      • Drop vestiges of Python 2.5 support in testing
      • Update rpython functions with ones needed for py3k
    Please update, and continue to help us make PyPy better.
    Cheers
    The PyPy Team







    No comments: