Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

PyPy London Demo Evening 2013

77,512 views

Published on

The slides we used at the PyPy London Demo Evening:

http://morepypy.blogspot.co.uk/2013/08/preliminary-london-demo-evening-agenda.html

Published in: Technology
  • Be the first to comment

PyPy London Demo Evening 2013

  1. 1. Looking for Python developers! Speak to us at PyPy demo night or visit www.pitchup.com/jobs Python / Django Postgres Celery Redis nginx memcache Jquery Solr S3 Leading booking site for campsites and caravan parks, founded in 2009 by lastminute.com alumni ● 65k visits / day, £6m bookings / year ● 650 bookable sites ● Huge market ○ 26k campsites and 300m bednights in Europe ○ 600m bednights in US ○ 47% more bednights than hotels (GB) ○ More trips to campsites than holidays to France + Spain combined (GB) ● Team of 15, based in west London
  2. 2. Welcome to the PyPy Demo Evening Laurence Tratt 2013-08-27 1 / 5 http://soft-dev.org/
  3. 3. Why are we here? 2 / 5 http://soft-dev.org/
  4. 4. What do we do? (l-r) Vasudevan, Bolz, Tratt, Barrett, Diekmann 3 / 5 http://soft-dev.org/
  5. 5. What do we do? • Aim: identify important challenges in software development. 3 / 5 http://soft-dev.org/
  6. 6. What do we do? • Aim: identify important challenges in software development. • Strengths: language design and implementation. 3 / 5 http://soft-dev.org/
  7. 7. What do we do? • Aim: identify important challenges in software development. • Strengths: language design and implementation. • Immediate benefits: faster VMs. 3 / 5 http://soft-dev.org/
  8. 8. What do we do? • Aim: identify important challenges in software development. • Strengths: language design and implementation. • Immediate benefits: faster VMs. • Long-term benefits: language composition. 3 / 5 http://soft-dev.org/
  9. 9. This evening 1 Carl Friedrich Bolz PyPy overview. 2 Lukas Diekmann Storage strategies. 3 Maciej Fijalkowski NumPy. 4 Armin Rigo Software Transactional Memory (STM). 5 Edd Barrett Language composition using meta-tracing. 4 / 5 http://soft-dev.org/
  10. 10. Can you help? • Contributors. 5 / 5 http://soft-dev.org/
  11. 11. Can you help? • Contributors. • Resources. • Software Freedom Conservancy 5 / 5 http://soft-dev.org/
  12. 12. A Very Brief Introduction to PyPy Carl Friedrich Bolz PyPy Demo Evening, King’s College London, August 27, 2013 Carl Friedrich Bolz A Very Brief Introduction to PyPy
  13. 13. CPython is slow CPython 1-3 orders of magnitude slower than C BinaryTrees Dhrystone FannkuchRedux Fasta Knucleotide Mandelbrot Nbody RegexDNA RevComp Richards SpectralNorm 0.1 1 10 100 1000 SlowerthanC,lowerisbetter C Java Cpython Carl Friedrich Bolz A Very Brief Introduction to PyPy
  14. 14. Reasons for Bad Performance interpretation overhead late binding dispatching boxing Carl Friedrich Bolz A Very Brief Introduction to PyPy
  15. 15. Enter PyPy a modern efficient implementation of Python Carl Friedrich Bolz A Very Brief Introduction to PyPy
  16. 16. Enter PyPy a modern efficient implementation of Python open source, MIT license written in Python itself, then bootstrapped to C uses a tracing JIT compiler to produce machine code at runtime Carl Friedrich Bolz A Very Brief Introduction to PyPy
  17. 17. Performance of PyPy significantly faster than CPython, typically in the same order of magnitude than C BinaryTrees Dhrystone FannkuchRedux Fasta Knucleotide Mandelbrot Nbody RegexDNA RevComp Richards SpectralNorm 0.1 1 10 100 1000 SlowerthanC,lowerisbetter C Java PyPy CPython Carl Friedrich Bolz A Very Brief Introduction to PyPy
  18. 18. Performance of PyPy significantly faster than CPython, typically in the same order of magnitude than C BinaryTrees Dhrystone FannkuchRedux Fasta Knucleotide Mandelbrot Nbody RegexDNA RevComp Richards SpectralNorm 0.1 1 10 100 1000 SlowerthanC,lowerisbetter C Java PyPy CPython on average about 6.3 faster than CPython Carl Friedrich Bolz A Very Brief Introduction to PyPy
  19. 19. Demo Carl Friedrich Bolz A Very Brief Introduction to PyPy
  20. 20. Architecture Python interpreter written in RPython JIT-compiler automatically added via meta-tracing Carl Friedrich Bolz A Very Brief Introduction to PyPy
  21. 21. Status Python 2.7.3 support, 2.7.4 coming soon beta-level support for Python 3.2, more coming eventually pure Python code fully supported, please report as bug if not Carl Friedrich Bolz A Very Brief Introduction to PyPy
  22. 22. Status Python 2.7.3 support, 2.7.4 coming soon beta-level support for Python 3.2, more coming eventually pure Python code fully supported, please report as bug if not C extension modules partially supported, if they are well-behaved they are slow use cffi (a ctypes replacement) instead Carl Friedrich Bolz A Very Brief Introduction to PyPy
  23. 23. Questions? PyPy is a fast JITted Python implementation (if something is not fast, please report it as a bug) open source under MIT license http://pypy.org Carl Friedrich Bolz A Very Brief Introduction to PyPy
  24. 24. Storage Strategies for Fast Containers Lukas Diekmann August, 27 2013 1 / 6 http://soft-dev.org/
  25. 25. Collection strategies introduced in PyPy 1.9 optimisation of collections for certain data types improving speed reducing memory 2 / 6 http://soft-dev.org/
  26. 26. Idea typical programs have homogeneously types collections create optimised versions of collections for certain types so far: lists: ints, floats, strings/unicode sets: ints, floats, strings/unicode dicts: ints, strings/unicode 3 / 6 http://soft-dev.org/
  27. 27. Optimisations collections automatically change to most efficient strategy store elements more memory efficiently fast elements access 4 / 6 http://soft-dev.org/
  28. 28. Further optimisations collection creation and initalisation split(d), set([1,2,3]) type based operations: contains, difference, issubset special strategies RangeListStrategy: calculates elements on the fly Tracing JIT interaction: faster (low-level) comparisons, remove type checks 5 / 6 http://soft-dev.org/
  29. 29. Results paper at OOPSLA on average ∼18% speedup ∼6% less memory usage more info at http://soft-dev.org/pubs/ 6 / 6 http://soft-dev.org/
  30. 30. Numpy on PyPy Maciej Fijałkowski King’s College London August 27 2013 fijal Numpy on PyPy
  31. 31. Goals fully compliant numpy replacement for PyPy fast looped operations fast vectorized operations fijal Numpy on PyPy
  32. 32. Why? fast looping single language fijal Numpy on PyPy
  33. 33. Model some programs have numerical kernels that can be written in C some don’t http://arxiv.org/abs/1301.1334 image manipulation demo abstraction unfriendly fijal Numpy on PyPy
  34. 34. Status fast looped operations ok vectorized operations fijal Numpy on PyPy
  35. 35. Future goals finish numpy make it fast make it compatible with more software (matplotlib, scipy) fijal Numpy on PyPy
  36. 36. Funding about $20k left we likely need more behind schedule, but not behind budget fijal Numpy on PyPy
  37. 37. Q&A Thank you! fijal Numpy on PyPy
  38. 38. Software Transactional Memory on PyPy
  39. 39. Pseudo-Goal “Kill the GIL” GIL = Global Interpreter Lock
  40. 40. Real Goals Multi-core programming But reasonable multi-core programming Using the recent model of Transactional Memory
  41. 41. PyPy-STM An executable pypy-stm which uses internally Software Transactional Memory Optimistically run multiple threads in parallel The only new feature is atomic: with atomic: piece of code...
  42. 42. Example of higher-level API def work(...): ... several more calls to: transaction.add(work, ...) ... Starts N threads, scheduling work() calls to them Each work() is done in an atomic block Multi-core, but as if all the work() are done sequentially
  43. 43. Status Kind of working without the JIT Roughly three times slower (you need four cores to see benefits) Working on the JIT support
  44. 44. Q&A Thank you! Budget of $10k left, likely more needed too
  45. 45. Unipycation Combining Prolog and Python Edd Barrett August 27, 2013 1 / 9 http://soft-dev.org/
  46. 46. Our Goal The softdev team is exploring language compositions. Ideally our compositions should be: Easy to implement. Transparent (as possible) to the user-programmer. High performance. Can meta-tracing help? 2 / 9 http://soft-dev.org/
  47. 47. Unipycation: A Language Composition Experiment PyPy + Pyrolog = Unipycation Unipycation Both interpreters implemented in RPython. About 600 LoC of integration code. A few months to develop. Languages communicate via an API. No syntax integration yet. 3 / 9 http://soft-dev.org/
  48. 48. Why? Explore composition of opposing paradigms. Evaluate RPython as a language composition framework. Performance/ease of development. Composition with many realistic applications. Example Scenario Data acquisition by JSON/XML/Sqlite. Easy in Python, not easy in Prolog. Some kind of knowledge inference based upon data. Perhaps not so easy in Python, trivial in Prolog. Visualisation of Results. Easy in Python, lack of library support in Prolog. 4 / 9 http://soft-dev.org/
  49. 49. Example Suppose we have a directed graph (London Underground?): a c b d e d g And we need to ask questions like: Where can I get to from ’b’ via at most 4 nodes and how? 5 / 9 http://soft-dev.org/
  50. 50. This is Easy with Prolog path.pl: edge(a, c). edge(c, b). edge(c, d). edge(d, e). edge(b, e). edge(c, f). edge(f, g). edge(e, g). edge(g, b). path(From , To , MaxLen , Nodes) :- path(From , To , MaxLen , Nodes , 1). path(Node , Node , _, [Node], _). path(From , To , MaxLen , [From | Ahead ], Len) :- Len < MaxLen , edge(From , Next), Len1 is Len + 1, path(Next , To , MaxLen , Ahead , Len1). query: path(b, To, 4, Path). 6 / 9 http://soft-dev.org/
  51. 51. Example: Python → Prolog from uni import Engine engine = Engine.from_file (" path.pl") paths = engine.db.path.iter for (to , nodes) in paths ("b", None , 4, None): print ("To %s via %s" % (to , nodes)) To b via [’b’] To e via [’b’, ’e’] To g via [’b’, ’e’, ’g’] To b via [’b’, ’e’, ’g’, ’b’] Calling from Prolog to Python also possible. E.g. python:somefunc(blah) 7 / 9 http://soft-dev.org/
  52. 52. Performance 8 / 9 http://soft-dev.org/
  53. 53. In Summary Compositions are relatively easy to implement with RPython. We were able to implement a fairly transparent API-like interface. Performance promising. Further Reading https://bitbucket.org/vext01/pypy http://soft-dev.org/ 9 / 9 http://soft-dev.org/
  54. 54. Looking for Python developers! Speak to us at PyPy demo night or visit www.pitchup.com/jobs Python / Django Postgres Celery Redis nginx memcache Jquery Solr S3 Leading booking site for campsites and caravan parks, founded in 2009 by lastminute.com alumni ● 65k visits / day, £6m bookings / year ● 650 bookable sites ● Huge market ○ 26k campsites and 300m bednights in Europe ○ 600m bednights in US ○ 47% more bednights than hotels (GB) ○ More trips to campsites than holidays to France + Spain combined (GB) ● Team of 15, based in west London

×