Author: Maciej Fijalkowski <[email protected]> Branch: extradoc Changeset: r5506:ff4a28b22fc0 Date: 2015-02-20 12:00 +0200 http://bitbucket.org/pypy/extradoc/changeset/ff4a28b22fc0/
Log: work on talk diff --git a/talk/divio-zurich-2015/talk.rst b/talk/divio-zurich-2015/talk.rst --- a/talk/divio-zurich-2015/talk.rst +++ b/talk/divio-zurich-2015/talk.rst @@ -45,12 +45,28 @@ cffi ---- -XXX slide about cffi and demo +xxxx image + +cffi (2) +-------- + +* demo + +numpy and numerics +------------------ + +* big subset of numpy reimplemented + +* same speed as numpy, much faster with python code + +* aim at compatibility + +* pymetabiosis (scipy, matplotlib) Speed status ------------ -xxx graph +* speed.pypy.org Speed status (2) ---------------- @@ -68,11 +84,65 @@ * objects are smaller than CPython +* interpreter size is large (~60M) + +* additional assembler consumes memory + +* plus bookkeeping info + +* measure! + Can I run PyPy? --------------- -Xxxx +* mostly, yes! -Case study - magnetic +* look at C extensions + +* try, measure, benchmark + +Case study - magnetic.com +------------------------- + +* magnetic.com - online search retargetting + +* 2 weeks of work, including benchmarking and analysis + +* swapped ujson for json + +* rewrote bindings for protobuf + +* ~2x speedup after warmup, no additional noticable memory + +Future +------ + +* work on warmup + +* STM work + +* more numpy compatibility + +Crowdfunding +------------ + +* moderate success in crowdfunding + +* numpy, STM, py3k + +* donate + +Commercial consulting --------------------- +* baroquesoftware.com + +* measure, analyze, port to pypy + +Q&A +--- + +* http://pypy.org + +* http://baroquesoftware.com + _______________________________________________ pypy-commit mailing list [email protected] https://mail.python.org/mailman/listinfo/pypy-commit
