Today, I ran some tests with pypy and my few Project Euler solutions:
| Solution | Python 2.6.6 | PyPy 1.5 | Factor |
|---|---|---|---|
| 7 | 9.57s | 1.92s | 4.98 |
| 7_b | 0.52s | 0.14s | 3.71 |
| 9 | 0.51s | 0.10s | 5.1 |
| 14 | 6.70s | 6.97s | 0.96 |
| 25 | 1.25s | 0.68s | 1.83 |
I guess the main message to take home is (see especially 7 and 7_b): a good algorithm is way more important then the interpreter, but if you can choose (e.g. you don't need numpy), then why not use PyPy and gain some speed?