I met up with the Gorgeous Ladies Of Prima (G.L.O.P.) today, and I actually bought a copy of my own book (I also signed all of the copies at the Prima booth and at the GDC bookstore). So, I was flying high today, and my ego swelled to enormous and dangerous size. I think the other guys are going to kill me. No, seriously.

Anyway, this whole "being an author" thing finally sank in.

Now that Mason and Dave are here, my contributions will be a bit less frequent, the drive to just get something up in the way of coverage is gone for me.

Today, I had interviews with Vicon motion systems (you know, the guys with the hot chick in the skin tight black body suit). I found out that for the low, low price of $100,000.00 (US), you too can have a small motion capture system which is really great for skeletal animation. Mere pocket change!

My other interview for the day was with NxN Software, and I talked with them a bit about AlienBrain. For about the first ten minutes, even after it was being demonstrated for me, I still did not know what it did. The fact that the dude presenting it to me had a rather thick accent didn't help. Apparently, AlienBrain is like Visual SourceSafe (i.e. a resource management tool for team based development), but looks cooler. They were a little evasive on the price, but apparently you can get this tool starting around $1500-$2000.

The thing I really don't like about interviews: they always have the marketing or public relation people there to tell me how spiffy their product is. I'm not trying to take away the spiffiness of the product, but I'm a programmer, and I prefer talking to other programmers, who can actually answer the questions I'm likely to ask. Sigh.

My second lecture of the day was by a couple of ATI guys (Jason Mitchell and Chris Bennett), about why the new 2.0 Pixel Shader spec rules.

I have absolutely no doubt, after seeing all of these really sweet looking ATI demos, that pixel and vertex shaders are going to be THE way to achieve killer graphics in the future. You don't truly grasp the range of effects you can create by just looking at the spec, but when you see the technology in expert hands, you start to realize just how much of a revolution this whole shader thing really is.

It was a good lecture; a lot of stuff that was really math-intensive, so it's definately something to download later and spend time analyzing, but it had lots of cool demos to keep everyone entertained, and it was packed full of technical info.

Not that this has anything to do with the conferences, but...

I flew in last night, touched down in San Jose at about 7:45pm. Today at precisely 10:47am I spotted the first reference to the infamous "All Your Base Are Belong To Us" meme, in the form of a Zero Wing t-shirt. It was a fleeting sighting; the subject disappeared into a mass of people before I had a chance to talk to him.

14 hours, 2 mins - frankly much slower than I had anticipated, what with this being a game development conference and all. I still haven't seen it on the Expo floor yet, though I imagine it's only a matter of time... Maybe the AYBABTU thing is finally dying?

Well, considering this is my first ever update as a member of the GameDev.net press at the 2001 GDC (I just flew in last night), I should have some snappy "this is your faithful reporter in the field" line to start with, but I couldn't come up with anything clever so I thought I'd just launch into the meat and potatoes of the story instead.

I just finished listening to a presentation by Ronen Zohar on optimizing Vertex Shaders for the Intel Pentium III and IV CPUs. At first, I was a little apprehensive - after all, the words "fastest performance" and "software vertex processing" don't exactly go hand-in-hand. Imagine my surprise then when I learned that properly optimized vertex shaders running in software can hold their own against the silicon.

It's all about using some subtle tricks. Here's the short list:

• Use the m?x? macro multiply instructions. If you use these, the vertex shader compiler will know what you're up to, and will take steps to optimize your code as much as possible. Note that this run against advice given by Nvidia, which had said that the macro instructions should be avoided.
• Try to avoid using the address register. If you must use it, align your vertices so that they all use the same address register. This will help out the compiler, because it uses SIMD instructions behind the scenes.
• Write only what you need. Use the dot modifiers (r0.xyz, for example) to ensure that the compiler only spends time calculating useful data.

Those are the big things. Ronen had said that the slides for his talk would be available on the Intel Developer Services web site (I don't know the URL offhand but it, when posted, will probably end up somewhere under http://developer.intel.com).

All in all, it was a good talk. Very little "sales pitch," very strong on practical techniques. The presentation had lots of little trick information you can't get anywhere except at the GDC. More news on other talks as they develop - stay tuned!

Dude! They had a Lego game! Well, it's not REALLY a lego game, its just a version of nine-man-morris, but it had a lego skin! It rules.