Building fast VMs quickly

Slide 1

Slide 1 text

Friday, April 13, 2012

Slide 2

Slide 2 text

Hit me with questions as we go Friday, April 13, 2012

Slide 3

Slide 3 text

Building fast VMs quickly Alex Gaynor Friday, April 13, 2012

Slide 4

Slide 4 text

Building high performance VMs for dynamically typed languages is hard Friday, April 13, 2012

Slide 5

Slide 5 text

It doesn’t have to be Friday, April 13, 2012

Slide 6

Slide 6 text

PyPy A framework for implementing high performance dynamic languages A Python implementation built on this framework Friday, April 13, 2012

Slide 7

Slide 7 text

Architecture of a VM Tokenize Parse AST construction Bytecode compilation Interpretation Friday, April 13, 2012

Slide 8

Slide 8 text

Let’s build a VM Friday, April 13, 2012

Slide 9

Slide 9 text

https://bitbucket.org/ alex_gaynor/example-vm/ Friday, April 13, 2012

Slide 10

Slide 10 text

Our language Approximately a subset of Javascript Variables Floats, bools Arithmetic if, while, print Friday, April 13, 2012

Slide 11

Slide 11 text

Lexing and Parsing Write a grammar in EBNF That’s it Friday, April 13, 2012

Slide 12

Slide 12 text

AST Construction Take a tree of syntax nodes, turn them into semantic nodes Friday, April 13, 2012

Slide 13

Slide 13 text

Bytecode compilation Turn the AST into a bytecode. Friday, April 13, 2012

Slide 14

Slide 14 text

Object model How you represent your objects. Friday, April 13, 2012

Slide 15

Slide 15 text

Interpreter Simple bytecode interpreter. Friday, April 13, 2012

Slide 16

Slide 16 text

RPythonize and JIT Friday, April 13, 2012

Slide 17

Slide 17 text

That’s all folks! Thanks to Professor Moorthy Everyone in RCOS Friday, April 13, 2012