Introducing the Speed Center
Until now, Python developers interested in PyPy's performance had to rely on a particular dev or tester to manually post some benchmark data. To remedy that, as well as to aid PyPy developers in analyzing trunk performance, speed.pypy.org has been launched, just in time for the next PyPy release!
Right now, it is mostly a tool for monitoring trunk performance, but in time it should become more than that. Let us describe what it currently offers:
The Overview. Test results for last trunk revision are shown in a table, detailing the performance changes the current revision has produced, the trend over a number of last revisions to spot subtler regressions, and a comparison to Python 2.6.2.
A note about the “Average” column: when looking at it for http://speed.pypy.org/overview/?revision=71603&interpreter=2 for example, does it mean that pypy is overall 2.85 times faster than stock Python?. Not at all. It is just the average of the collection of benchmarks we currently happen to be using, and which is not representative of average usage in the real world. Is is actually pretty meaningless right now. Why was it included then? To give devs an indication of how a particular commit influences performance across the board, in addition of a particular benchmark.
If you want to examine the evolution of a particular benchmark over different revisions, you may just click on that row and go to...
The Timeline view.  A plot showing the improvements and regressions over a number of revisions, for the different interpreters.

There are many improvements planned to those views, as well as new functionality altogether that could be implemented if enough people deem it interesting:
A Comparison view: it will allow to compare different Python interpreters side by side in a classic bar chart modus. This will be the most interesting view for non-developers, specially when we manage to run other interpreters like jython, ironpython and unladen swallow
Reports: summary of the last benchmark run with regression alerts. A user will be able to configure alert thresholds and be sent an email as soon as the last run finishes.

I would also like to note, that if other performance-oriented opensource projects are interested, I would be willing to abstract more the code and set-up such a Speed Center configured for them. And of course, any helping hands are welcome and would allow it to become a project in its own right. A possible line of development would be integration with buildbot and with software forges. Speed is being developed with Django and jquery .

That's all for now!