Added bench mark groups and some more python 3 benchmarks

Created originally on Bitbucket by rjnienaber (Richard Nienaber)

Python 3 Benchmarks

The following benchmarks were copied from CPython or ported:

• CPython: chaos, django, fannkuch, float, go, hexiom2, json, meteor-contest, nobody, pidigits, raytrace, richards, spectral-norm, telco, ai (now nqueens)
• Ported: htm5lib, crypto_pyaes, pyflate, sympy, scimark, mako, chameleon

Not ported (Python 3 not supported or not released)

• spambayes
• twisted
• eparse (twisted code)
• rietveld (requires app_engine as well)
• sphinx (http://bugs.python.org/issue10224)
• trans2 (rpython specific test which will require lots of work)
• genshi
• spitfire

Benchmark groups

The benchmark group logic from unladen swallow has been reused to separate python3 benchmarks from the original benchmarks.

The following are examples for running benchmark code:

• pypy runner.py --fast -c pypy --baseline=pypy19 (all default/original)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b default (all default/original)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b math (math group)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b 2n3 (benchmarks valid in python 2 and python 3)
• pypy runner.py --fast -c pypy3 --baseline=pypy3 --baseline=pypy19 -b python3 (python3 only benchmarks)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b spitfire (individual test)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b regex,ai (group and individual test)
• pypy runner.py --fast -c pypy --baseline=pypy19 -b scimark,-scimark_fft (scimark group except scimark_fft)

Note: regex tests have been enabled in the default group