Saturday is the last day of the Python Need for Speed sprint. We had to vacate the room by 5pm, so it was quite a short day. Most other days we were hacking until 8 or 9pm before taking a dinner break (remember, it only starts getting dark at around 11pm, and then is only dim for 4 hours). The change in daylight definitely didn't prompt us to stop hacking. Read on for more.

Neil Norwitz on python-dev was hoping the sprint was going to be over soon, because he couldn't keep up. I imagine he was particularly overwhelmed on Saturday afternoon, because people were franticly working to finish things up.

Nothing was specifically listed on the pages we were maintaining for tracking daily changes for Saturday though. We also didn't do a wrap meeting, largely because we ran out of time, so it was hard to tell what was going on Saturday.

Earlier in the sprint I had sent out a mid-sprint review to Linux Weekly News. It got posted on Saturday, largely because my initial e-mail to the folks at LWN didn't include the article as an attachment, so it got delayed a day.

Steve Holden posted a wrap message to his blog. It's a good overview of the event.

I spent the whole day, starting from 5am, trying to figure out how to make a better benchmark. The main problem is that pybench is much too inconsistent. Pybench, particularly on Linux, seems to vary a lot from run to run. Individual tests can vary by as much as 10% from one run to the next. I came up with a test that varied much less from run to run, but it also takes a very long time to run. For the current set of tests, around 15 minutes. I mostly got my new version working, but there are a few changes I need to do before it's usable yet.

So, sadly, not much specific to report for Saturday.
(Post Reply)