GDC Report: Threading for Performance tutorial

Yesterday, day 2 of GDC, I sat in on the day long "Threading for Performance" tutorial put on by Paul Lindberg and Brad Werth of Intel.

They went over the details needed of taking a completely "serialized" application, where everything has to wait for the thing before it to complete before it can happen, and move by steps to a threaded, parallel, multicore-ready application. There were a couple dozen quad core "Core 2 Extreme" lab systems around the room, but the place was packed all day, so no one really got to have a system to themselves.

The labs used the "Destroy the Castle" demo from last year's GDC, which is supposed to be available for download on the Intel Software Network site, but I couldn't find it after a cursory browse, and it wasn't showing up in the search (oops!). We need to fix that, to make the download more accessible and easy to find, because from the feeling in the room, lots of the developers there are going to want to get their hands on it, and look at how it went from about 40 frames per second to almost 300 frames per second by optimizing the app for multicore systems.

I'll try to track down a link to the demo files and the tutorial materials, and get it placed somewhere more prominent. If you go looking for it, and can't find it, let me know (also, if you DO find it, please tell me, so I can at least post a link).
Einzelheiten zur Compiler-Optimierung finden Sie in unserem Optimierungshinweis.