perf.rst

Python performance

Codespeed websites

• speed.python.org: CPython benchmarks, run performance and (tobami's) codespeed
• speed.pypy.org: PyPy benchmarks, run PyPy benchmarks
• speed.pyston.org: Pyston benchmarks, run pyston-perf

Projects

Python benchmarks:

• performance (GitHub project): Python benchmark suite
• perf (GitHub project): Python module to run, analyze and compare benchmarks
• performance_results: Results of the of performance benchmark suite
• pymicrobench: Collection of Python microbenchmarks written for CPython.
• PyPy benchmarks: PyPy benchmark suite
• Codespeed (tobami's flavor): Django application to explose benchmark results. Pages: Homepage, Changes, Timeline, Compare.
• pyston-perf: Pyston benchmark suite

Other Python Benchmarks:

• Numba benchmarks
• Cython Demos/benchmarks
• pythran numpy-benchmarks

Old deprecated projects:

• benchmarks: the old and deprecated Python benchmark suite

Mailing list

• Python speed mailing list

Old Python benchmarks

The following benchmarks were removed from pyperformance:

• bm_rietveld.py: use Google AppEngine and Rietveld application which are not available on PyPI
• bm_spitfire.py: Spitfire project is not available on PyPI
• bm_threading.py (threading_iterative_count, threading_threaded_count)
• gcbench.py
• pystone.py
• tuple_gc_hell.py

Zero copy

Python3:

offset = 0
view = memoryview(large_data)
while True:
    chunk = view[offset:offset + 4096]
    offset += file.write(chunk)

This copy creates views on large_data without copying bytes, no bytes is copied in memory.