PyPy Technologies

Author:	Samuele Pedroni, Open End AB
Date:	January '10

Generated Tracing JIT compiler

Template: use the streamlined approach introduced by tracing JITs

• identify hot loops
• produce linear traces with stay-on-trace checks (guards), follow calls!
• compile the traces to assembler which run the loop
• exploit when heap allocation can be forgone, mirror Python frames on the CPU stack

Example

Status

• Intel x86-32
• Small and some medium examples run
• Good results for algorithmic code, see benchmarks
• Reasonable heuristic when traces get too long

Benchmarks

Stackless Python features

• C stack frame chains can be migrated back and forth to the heap
• coroutines
• tasklets and channels (inspired by Limbo and predecessors)

Memory Management and GC

Pluggable GCs

• generational GC with semi-space 2nd generation
• experimented with mark-and-compact GC

Footprint

• larger basic footprint (code generation overhead, more extensions compiled in by default)
• smaller layout for many objects though (e.g. shared dicts for object instances)

QA

• writing tests first for new code
• 10 thousands of tests
• nightly run of tests, both own plus running tests with translated PyPy
• running against CPython own test suite

Links

Status blog http://morepypy.blogspot.com

Website http://codespeak.net/pypy