Raspberry-Python-Pi for Hardware Hacking Fun!

Slides from my talk at Vancouver Python Day 2013 (http://www.vanpyday.com)

Github: http://github.com/dbader/piradio
My blog: http://dbader.org

Transcript

1. Raspberry-Python-Pi for Hardware Hacking Fun! Daniel Bader – @dbader_org
2. Hardware-hacking for software people.
3. +
4. piradio
5. Hi!
6. • My project and the lessons learned • Python on the Raspberry Pi • Inspire you to try this too
7. Why?
9. How?
10. Raspberry Pi
13. Python on the Pi • Python is a first-class citizen • CPython & PyPy (ARM) • A great platform for Python on an embedded system
14. RaspiLCD kit by Martin Stepphun (emsystech.de)
15. Soldering
16. The case
17. Custom CNCed front plate schaeffer-ag.de/en
18. “Front-plate Designer” schaeffer-ag.de/en
19. Margins
20. Aluminum case Proma 1030 - conrad.de
21. Parts list • Raspberry Pi Model B ($35) • Display + buttons (RaspiLCD) ($55) • Aluminum case + custom CNCed front-plate ($70) • 32 GB SD Card ($20) • Wi-Fi USB stick ($20) • USB power supply ($10) • Total: ~ $200 (in theory)
22. Software
24. Architecture
25. Philosophy: Do as much as possible in pure Python.
26. Architecture Linux piradio CPython mpd freetype raspilcd
27. piradio services audio podcast clock weather … panels radio clock alarm weather … podcast Architecture graphics lcd fonts commons config ui
29. Things I learned
30. Simulate where possible
31. • Implemented as a “fake LCD” driver • pygame + libSDL • Saved me so much time
32. Take advantage of the familiar programming environment
33. Fast Python (on the Raspberry Pi)
34. 1. Avoid the dots ! ! def bad(): a = A() ! for i in xrange(10000): a.b.c.d.e.foo() ! ! def good(): a = A() foo = a.b.c.d.e.foo for i in xrange(10000): foo() ! ! # 3.3x faster!
35. 1. Avoid the dots (...) ! FOR_ITER STORE_FAST ! LOAD_FAST LOAD_ATTR LOAD_ATTR LOAD_ATTR LOAD_ATTR LOAD_ATTR CALL_FUNCTION ! POP_TOP JUMP_ABSOLUTE POP_BLOCK ! (...) (...) ! FOR_ITER STORE_FAST ! LOAD_FAST CALL_FUNCTION ! POP_TOP JUMP_ABSOLUTE POP_BLOCK ! (...)
36. 2. range() vs xrange() range(): • returns a list • needs more memory • support slices • usually slower xrange(): • returns a generator • needs less memory • no slices • usually faster
37. 3. Access bitmaps row-wise instead of column-wise … = reality “memory layout” abstraction “2d bitmap”
38. 3. Access bitmaps row-wise instead of column-wise … = nice, sequential memory access
39. 3. Access bitmaps row-wise instead of column-wise … = jumping all over the place
40. 3. Access bitmaps row-wise instead of column-wise ! ! def bad(): bitmap = bytearray(w * h) v = 0 for x in xrange(w): for y in xrange(h): v += bitmap[y * w + x] ! ! def good(): bitmap = bytearray(w * h) v = 0 for y in xrange(h): for x in xrange(w): v += bitmap[y * w + x] ! ! # 1.1x faster (Pi)
41. Fast Python (on the RPi) • Profile and benchmark everything • Trade memory for speed: cache things • Use the right data structures
42. Fun parts • Building a physical object • Building something useful and seeing people use it • UI design and programming in a restricted environment • Programming nostalgia – Pixels!
43. Not-so-fun parts • Soldering • Front plate margins • Raspberry Pi audio crackling • Things get expensive quickly
45. The Future • I'll release this as open-source on GitHub • Finish the blog post on it • Users and contributors welcome
46. github.com/dbader/piradio dbader.org/blog/tags/python ! @dbader_org Thanks! Questions? Ideas?