Benchmarks result. Some are slower (code placement issue?), the faster benchmark are not really much faster.

haypo@speed-python$ python3 -m perf compare_to /home/haypo/benchmarks/2017-02-06_07-15-default-e06af4027546.json pyobject_fastcall-3_ref_e06af4027546.json -G  --min-speed=4
Slower (4):
- scimark_lu: 352 ms +- 11 ms -> 374 ms +- 15 ms: 1.06x slower (+6%)
- float: 238 ms +- 3 ms -> 251 ms +- 4 ms: 1.05x slower (+5%)
- logging_silent: 558 ns +- 15 ns -> 585 ns +- 16 ns: 1.05x slower (+5%)
- raytrace: 1.04 sec +- 0.01 sec -> 1.09 sec +- 0.01 sec: 1.05x slower (+5%)

Faster (5):
- nbody: 233 ms +- 2 ms -> 216 ms +- 2 ms: 1.08x faster (-7%)
- pickle_dict: 56.6 us +- 4.3 us -> 52.9 us +- 2.2 us: 1.07x faster (-6%)
- nqueens: 217 ms +- 2 ms -> 205 ms +- 2 ms: 1.06x faster (-6%)
- unpickle_pure_python: 670 us +- 11 us -> 642 us +- 8 us: 1.04x faster (-4%)
- scimark_sor: 405 ms +- 11 ms -> 389 ms +- 10 ms: 1.04x faster (-4%)

Benchmark hidden because not significant (55): (...)