In the morning, Paul McKenney, in "Why is parallel Programming Hard?", discussed some of the background of parallel programming research, then gave an entertaining demonstration of instruction overhead using a roll of toilet paper. Each square represented one clock cycle -- he estimated there were a few hundred clock cycles in the full roll -- and he had audience members unroll the roll carefully, passing it from one person to the next. It took a long time.
Over at the LinuxChix miniconf, Jacinta Richardson gave a wonderfully entertaining (and useful) talk "On Speaking". She explained how to hack audience members' brains, particularly the corpus callosum and the hippcampus, by using emotion, visual images and suspenseful stories to give your audience whole-brain entertainment.
After Jacinta's talk we spent some time going around the room introducing ourselves, and speakers got a chance to plug their upcoming talks.
I skipped the panel on Geek Parenting (not being a parent) to go back to the kernel miniconf's "Problem Solving Hour". Questions involved network performance, solid state disk performance, how to debug crashes, tracing (the moderator commented that if you're thinking of getting involved in the kernel effort but aren't quite sure what to do, there's a huge need for better tracing and performance analysis tools), solid-state disks (someone plugged the talk on that subject on Friday) and similar interesting topics.
I asked about an overheating problem I've been having with my laptop. I mentioned that even in single-user mode, the CPU temperature keeps going up, so I was pretty sure it was a kernel and not userspace issue. Matthew Garrett said that a lot of drivers are optimized for a normal use case -- meaning X -- and may work very poorly in text mode. You can have something that's overheating in single-user mode, then you start X and a bunch of power management systems kick in and the temperature actually goes down. So how do you figure out what's causing a temperature problem? Open up the laptop when it's hot, poke around then figure out what's hot. Then debug that component.
Lunch was a lovely BBQ provided by Google.
After lunch, Matthew Garrett, in "How I learned to stop worrying and love ACPI", was entertaining, as all his talks are. I'm not sure I actually learned much in the way of practical advice for helping ACPI work better on my machines, but at least I learned lots of new ways in which ACPI sucks more than I ever realized.
Then it was back to LinuxChix for a workshop on getting schoolgirls more interested in IT. We saw short presentations from the four workshop leaders, then split into groups -- our group went outside and sat in the hazy sunshine and talked about how to get girls, teachers, parents and school IT staff on board.
After tea, all the LinuxChix groups reported back on the discussions and there was a full-room discussion on how to get involved with educational programs like that. Then we ended with lightning talks; I got roped into giving one, so I didn't take notes on the rest, but they were all fun and interesting.
Then in the evening, after dinner, we found a spot somewhat sheltered from the lights of the hotel for some quick astronomy before bed. The sky was hazy and picking up lots of sky glow from a light beam shining from the hotel, but fortunately the sky around the Southern Cross was clear. We found both the Large and Small Magellanic clouds, as well as Eta Carina and some other clusters around the Southern Cross. A lovely view, unmatched by anything I saw from around Sydney or Melbourne. Tasmania definitely wins for stargazing!
[ 04:17 Jan 19, 2009 More conferences/lca2009 | permalink to this entry ]