Author: Matti Picus <matti.pi...@gmail.com> Branch: extradoc Changeset: r5614:5f22d5e547b6 Date: 2016-03-23 23:44 +0200 http://bitbucket.org/pypy/extradoc/changeset/5f22d5e547b6/
Log: proposal for ScipyUS2016 diff --git a/scipyUS2016/proposal.rst b/scipyUS2016/proposal.rst new file mode 100644 --- /dev/null +++ b/scipyUS2016/proposal.rst @@ -0,0 +1,35 @@ +PyPy and the Numpy/Scipy Stack +============================== + +Abstract (500 words or less) +____________________________ + +PyPy is used in the "other" python world of web servers and text processing. +What can it do for data crunching? How can it possibly work with Numpy and the +rest of the Scientific Python data stack? In this talk I will briefly survey +what is PyPy, our two approaches to compatibility with Numpy, and what that +means for those who are looking for a drop-in solution to their processing +challenges. + +Longer Description +__________________ + +PyPy is maturing as a drop-in replacement for python 2.7. In the "other" world +of web servers and text processing, PyPy's speed on long-running processes +and compatibility with pure python packages makes it a good fit for mature +technologies looking for a quick speed increase. + +What about the world of number crunching, can PyPy possibly contribute +anything to the crowded field of ahead-of-time solutions like cython or +other just-in-time solutions like Numba? The PyPy team feels very strongly +that we can, and in this talk I will try to outline PyPy's approach to +intgrating numpy and the SciPy stack over this alternative interpreter. + +We have been working for a number of years on an alternative +implementation of the ndarray, tightly integrated to the PyPy machinary. +Reimplementing numpy completely comes with costs, like a constant need to +replicate updates when a new version of numpy is released. +We've been recently experimenting with reusing +more of C numpy through the C API and we have ideas how to merge both +approaches to have the best of both worlds. + _______________________________________________ pypy-commit mailing list pypy-commit@python.org https://mail.python.org/mailman/listinfo/pypy-commit
