PyPy 2.0 Beta 1 supports ARM processors
The first beta of version 2.0 of PyPy has been released with support for ARM CPUs and CFFI compatibility. PyPy is an alternative Python 2.x implementation with a just-in-time compiler, a stackless mode and a sandbox for untrusted code. It is described by its developers as faster and "almost a drop-in replacement for CPython 2.7.3". The new version of the "very compliant" Python interpreter is the first version to officially support the ARM processor architecture. The software will work on soft-float ARM/Linux builds on ARMv7 or later CPUs that have a floating-point unit.
Another major addition is official support for CFFI, the Foreign Function Interface for easily calling C code from Python. Users can install CFFI, which has just been updated to version 0.4, by using
pip install cffi after installing PyPy and pip. According to the announcement, the developers hope to add one optimisation before the release of 2.0 to make CFFI as fast as a native call from C. One regression which is also preventing the beta from being a full release is an issue with a slow ctypes implementation; this is being fixed by reimplementing the ctypes fast path using CFFI. Other changes include upgrading the Python standard library to version 2.7.3, various numpy updates and many performance improvements.
The developers point out that there are some features, such as greenlets support in the JIT, that did not make it into the beta but should be ready in the final release. However, some work items including speeding up the JIT warmup and support for Software Transactional Memory, will not make it into 2.0, but are being actively worked on.
Despite being a beta, its developers say that PyPy 2.0 Beta 1 is suitable for production use as its "stability is the same or better" than the current 1.9 release. Further information about the beta, including a list of known regressions, can be found in a post on the PyPy Status Blog and on the features page. The first beta for version 2.0 of PyPy is available from the project's download page. PyPy source code is hosted on Bitbucket and is made available under the MIT Licence.
- "Yard Wolf" PyPy 1.9 released into wild, a report from The H.