![Replace list (build at runtime) with tuple
(constant):
for x in [1, 2, 3]: ...
→ for x in (1, 2, 3): ...
Replace list with set (Python 3 only):
if x in [1, 2, 3]: ...
→ if x in {1, 2, 3}: ...
In Python 3, {1,2,3} is converted to a
constant frozenset (if used in a test)
AST optimizations (3)](https://image.slidesharecdn.com/fastercpython-140318044150-phpapp01/85/Faster-Python-FOSDEM-13-320.jpg)
![Evaluate operators:
"abcdef"[:3] → "abc"
def f(): return 2 if 4 < 5 else 3
→ def f(): return 2
Remove dead code:
if 0: ...
→ pass
AST optimizations (4)](https://image.slidesharecdn.com/fastercpython-140318044150-phpapp01/85/Faster-Python-FOSDEM-14-320.jpg)
![def func():
x = 33
return x + 1
LOAD_CONST 1 (33) # stack: [33]
STORE_FAST 0 (x) # stack: []
LOAD_FAST 0 (x) # stack: [33]
LOAD_CONST 2 (1) # stack: [33, 1]
BINARY_ADD # stack: [34]
RETURN_VALUE # stack: []
(6 instructions)
Stack-based bytecode](https://image.slidesharecdn.com/fastercpython-140318044150-phpapp01/85/Faster-Python-FOSDEM-19-320.jpg)
![Using registers allows more optimizations
Move constants loads and globals loads (slow)
out of loops:
return [str(item) for item in data]
Constant folding:
x=1; y=x; return y
→ y=1; return y
Remove duplicate load/store instructions:
constants, names, globals, etc.
registervm optim (1)](https://image.slidesharecdn.com/fastercpython-140318044150-phpapp01/85/Faster-Python-FOSDEM-21-320.jpg)
Faster Python, FOSDEM
- 1. FOSDEM 2013, Bruxelles Victor Stinner <[email protected]> Distributed under CC BY-SA license: http://creativecommons.org/licenses/by-sa/3.0/ Two projects to optimize Python
- 2. CPython bytecode is inefficient AST optimizer Register-based bytecode Agenda
- 3. Part I CPython bytecode is inefficient
- 4. Python is very dynamic, cannot be easily optimized CPython peephole optimizer only supports basic optimizations like replacing 1+1 with 2 CPython bytecode is inefficient CPython is inefficient
- 5. def func(): x = 33 return x Inefficient bytecode Given a simple function:
- 6. LOAD_CONST 1 (33) STORE_FAST 0 (x) LOAD_FAST 0 (x) RETURN_VALUE LOAD_CONST 1 (33) RETURN_VALUE RETURN_CONST 1 (33) Inefficient bytecode I get: (4 instructions) I expected: (2 instructions) Or even: (1 instruction)
- 7. Parse the source code Build an Abstract Syntax Tree (AST) Emit Bytecode Peephole optimizer Evaluate bytecode How Python works
- 8. Parse the source code Build an Abstract Syntax Tree (AST) → astoptimizer Emit Bytecode Peephole optimizer Evaluate bytecode → registervm Let's optimize!
- 10. AST is high-level and contains a lot of information Rewrite AST to get faster code Disable dynamic features of Python to allow more optimizations Unpythonic optimizations are disabled by default AST optimizer
- 11. Call builtin functions and methods: len("abc") → 3 (32).bit_length() → 6 math.log(32) / math.log(2) → 5.0 Evaluate str % args and print(arg1, arg2, ...) "x=%s" % 5 → "x=5" print(2.3) → print("2.3") AST optimizations (1)
- 12. Simplify expressions (2 instructions => 1): not(x in y) → x not in y Optimize loops (Python 2 only): while True: ... → while 1: ... for x in range(10): ... → for x in xrange(10): ... In Python 2, True requires a (slow) global lookup, the number 1 is a constant AST optimizations (2)
- 13. Replace list (build at runtime) with tuple (constant): for x in [1, 2, 3]: ... → for x in (1, 2, 3): ... Replace list with set (Python 3 only): if x in [1, 2, 3]: ... → if x in {1, 2, 3}: ... In Python 3, {1,2,3} is converted to a constant frozenset (if used in a test) AST optimizations (3)
- 14. Evaluate operators: "abcdef"[:3] → "abc" def f(): return 2 if 4 < 5 else 3 → def f(): return 2 Remove dead code: if 0: ... → pass AST optimizations (4)
- 15. "if DEBUG" and "if os.name == 'nt'" have a cost at runtime Tests can be removed at compile time: cfg.add_constant('DEBUG', False) cfg.add_constant('os.name', os.name) Pythonic preprocessor: no need to modify your code, code works without the preprocessor Used as a preprocessor
- 16. Constant folding: experimental support (buggy) Unroll (short) loops Function inlining (is it possible?) astoptimizer TODO list
- 18. Rewrite instructions to use registers instead of the stack Use single assignment form (SSA) Build the control flow graph Apply different optimizations Register allocator Emit bytecode registervm
- 19. def func(): x = 33 return x + 1 LOAD_CONST 1 (33) # stack: [33] STORE_FAST 0 (x) # stack: [] LOAD_FAST 0 (x) # stack: [33] LOAD_CONST 2 (1) # stack: [33, 1] BINARY_ADD # stack: [34] RETURN_VALUE # stack: [] (6 instructions) Stack-based bytecode
- 20. def func(): x = 33 return x + 1 LOAD_CONST_REG 'x', 33 (const#1) LOAD_CONST_REG R0, 1 (const#2) BINARY_ADD_REG R0, 'x', R0 RETURN_VALUE_REG R0 (4 instructions) Register bytecode
- 21. Using registers allows more optimizations Move constants loads and globals loads (slow) out of loops: return [str(item) for item in data] Constant folding: x=1; y=x; return y → y=1; return y Remove duplicate load/store instructions: constants, names, globals, etc. registervm optim (1)
- 22. Stack-based bytecode : return (len("a"), len("a")) LOAD_GLOBAL 'len' (name#0) LOAD_CONST 'a' (const#1) CALL_FUNCTION (1 positional) LOAD_GLOBAL 'len' (name#0) LOAD_CONST 'a' (const#1) CALL_FUNCTION (1 positional) BUILD_TUPLE 2 RETURN_VALUE Merge duplicate loads
- 23. Register-based bytecode : return (len("a"), len("a")) LOAD_GLOBAL_REG R0, 'len' (name#0) LOAD_CONST_REG R1, 'a' (const#1) CALL_FUNCTION_REG R2, R0, 1, R1 CALL_FUNCTION_REG R0, R0, 1, R1 CLEAR_REG R1 BUILD_TUPLE_REG R2, 2, R2, R0 RETURN_VALUE_REG R2 Merge duplicate loads
- 24. Remove unreachable instructions (dead code) Remove useless jumps (relative jump + 0) registervm optim (2)
- 25. BuiltinMethodLookup: fewer instructions: 390 => 22 24 ms => 1 ms (24x faster) NormalInstanceAttribute: fewer instructions: 381 => 81 40 ms => 21 ms (1.9x faster) StringPredicates: fewer instructions: 303 => 92 42 ms => 24 ms (1.8x faster) Pybench results
- 26. Pybench is a microbenchmark Don't expect such speedup on your applications registervm is still experimental and emits invalid code Pybench results
- 27. PyPy and its amazing JIT Pymothoa, Numba: JIT (LLVM) WPython: "Wordcode-based" bytecode Hotpy 2 Shedskin, Pythran, Nuitka: compile to C++ Other projects
- 28. Questions? https://bitbucket.org/haypo/astoptimizer http://hg.python.org/sandbox/registervm Distributed under CC BY-SA license: http://creativecommons.org/licenses/by-sa/3.0/ Contact: [email protected]
- 29. Thanks to David Malcom for the LibreOffice template http://dmalcolm.livejournal.com/