Numba 0.8 adds support for autojit classes and methods, building on the perfect hash table ideas from SEP 201 [1]. This allows methods to automatically specialize to argument types and instance attribute types. Many thanks to Dag Sverre Seljebotn for his work and ideas on the SEPs and his ideas and implementation of the hash table and surrounding facilities, and many thanks to Robert Bradshaw for his contributions and ideas.
Further functionality includes Python 3 support for pycc, and retrieving pointers from numba jit functions. A memory leak that plagued users has also been fixed. Many thanks for all the bug reports and feedback! Over the coming months we hope to stabilize numba and allow distribution of portable numba code and a well-defined and stable runtime. We also hope to address some of the math portability and numerical stability issues. We have a roadmap [2] which details what we want to do, and also provides entry points for new potential developers to contribute. Stay tuned as we try to make numba more robust, portable, interoperable and usable :) Download ======== http://numba.pydata.org/download.html Website ======= http://numba.pydata.org/ Documentation ============ http://numba.pydata.org/numba-doc/0.8/index.html Numba ====== Numba is an just-in-time specializing compiler which compiles annotated Python and NumPy code to LLVM (through decorators). Its goal is to seamlessly integrate with the Python scientific software stack and produce optimized native code, as well as integrate with native foreign languages. References ========= [1]: https://github.com/numfocus/sep/blob/master/sep201.rst [2]: http://numba.pydata.org/numba-doc/dev/roadmap.html -- http://mail.python.org/mailman/listinfo/python-announce-list Support the Python Software Foundation: http://www.python.org/psf/donations/
