May 08, 2004

seeking e3

As a last minute hack I've set up a routine that looks at all the assets that are loaded at game startup (except for geometry data & textures) and dumps that data out to a file. The file can then be used to load all the assets in a single pass, which cuts about 6 seconds off our load time[0]. The implementation is ugly, but it's compartmentalized and it'll do nicely until after E3.

I think I've fixed the last major bug in the movie player and it played a looped movie off DVD for about 10 hours without a crash[1]. The performance could use a little tweaking, but overall it's in a good shape.

So, now I'm waiting for a solid build, and then I'm on call for the rest of the weekend. Our publisher has decided that they'd rather have a couple extra days to test the build than a couple extra days to add new features. So, with a whole lot of luck I'll actually get tomorrow off.

--
[0] Reduced load time is almost entirely due to reduced seeking. And yes, it's a bad sign when you have load times that can be reduced by 6 seconds.

[1] The last crash seems to have been caused by a race condition where a callback triggered a read request as the movie player was either shutting down or starting up.

Posted by matt at 03:58 PM | Comments (1)

May 06, 2004

e3e3e3e3

With E3 just a few days off I am busy reconfirming that PS2 movie players are the spawn of the devil. I think I've managed to retune the buffer sizes so that it won't stutter anymore.

Fun fact of the day: DVD seek times can exceed 100ms. 100ms x 150 seeks = unhappy gamers. (Seek optimization is the next thing on my todo list)


In other news, I've volunteered to work two shifts during E3, so at some point I'm going to be given some training on what we're supposed to tell the press (no, I won't be wearing some sort of skimpy costume).

Posted by matt at 11:14 AM | Comments (1)