Why is Python slow? Python Nordeste 2013
0 likes•5,453 views
The document discusses why Python can be slow compared to other languages and provides tips for optimizing Python code. It explains that Python is an interpreted language and lacks type safety, which contributes to slower performance than compiled languages. However, Python includes tools like NumPy that optimize certain operations. The document recommends profiling code to identify bottlenecks, using appropriate data structures and algorithms, minimizing interpreter overhead through concurrency or C extensions, and as a last resort considering other languages. Overall it provides guidance on architectural choices, algorithms, memory usage, and language options to improve Python performance.
![Benchmark
>>> import timeit
>>> timeit.timeit("[i * i for i in xrange(100)]")](https://image.slidesharecdn.com/pyne-whypythonisslow-130530142230-phpapp02/85/Why-is-Python-slow-Python-Nordeste-2013-18-320.jpg)
![[i * i for i in range(100)]]
[i * i for i in xrange(100)]]
[i * i for i in np.arange(100)]]](https://image.slidesharecdn.com/pyne-whypythonisslow-130530142230-phpapp02/85/Why-is-Python-slow-Python-Nordeste-2013-22-320.jpg)
Why is Python slow? Python Nordeste 2013
- 1. Why is Python slow? Daker Pinheiro
- 2. Why is CPython slow? Daker Pinheiro
- 3. Why is CPython 2.x slow? Daker Pinheiro
- 4. $ whois dakerfp Daker Fernandes Pinheiro UFPE INDT Recife WebKit (Nix) Qt, KDE, ... C++, C, Python, Javascript, Prolog, ... Pythonist since 2009
- 13. Virtual Stack Machine >>> (z * y) + x + z
- 15. Bytecode Inspection >>> import dis >>> dis.dis(lambda x, y, z: (z * y) + x + z) 2 0 LOAD_FAST 2 (z) 3 LOAD_FAST 1 (y) 6 BINARY_MULTIPLY 7 LOAD_FAST 0 (x) 10 BINARY_ADD 11 LOAD_FAST 2 (z) 14 BINARY_ADD 15 RETURN_VALUE
- 16. 100 * 100 * 100 * 100 vs 100 ** 4
- 17. dict() vs {}
- 18. Benchmark >>> import timeit >>> timeit.timeit("[i * i for i in xrange(100)]")
- 19. Concurrency Global Interpreter Locker Avoid Threads Try Event Loops Try Multiprocess
- 20. Know your Data Structures Time Complexity
- 21. Use C/C++ Bindings numpy PyQt, PySide ... standard library
- 22. [i * i for i in range(100)]] [i * i for i in xrange(100)]] [i * i for i in np.arange(100)]]
- 23. ar = np.arange(100); ar * ar
- 24. Memory
- 25. Python 3 Similar to Python 2.7 performance Python 3 - Mailing list
- 26. Cython cdef average(int a, int b): return (a + b) / 2.0
- 30. Optimization Checklist 1. Legibillity 2. Architecture 3. Algorithm 4. Memory 5. Buffering 6. Caching 7. IO 8. Consider other languages :-(
- 31. Q & A Daker Fernandes Pinheiro http://codevereal.blogspot.com