||[Mar. 22nd, 2009|03:02 am]
Spectre Gruedorf Challenge Blog
Dredmor: now has an automatic updater which is, well, basically rsync + a wxWindows based wrapper knocked up by the Mighty Icculus. So we can now download updates automatically, distribute them to our users, watch them play the demo, and scrape their new log files to see what the heck they are, and are not, doing correctly. That's how you run a beta, folks.|
Hooked shadow maps back into the pipeline for spotlights only. This would have been significantly less of a pain in the ass if I was passing the correct world->eye transform matrix to the spotlight rendering pass. As a result, I spent much of my time today chasing after swimming texture bugs. Now, however, everything actually stays where you draw it. Which is good.
In the process, I tried about three different sets of precision values for various things, thinking that my problems were due to some combination of poor precision reconstructing the depth value, bad sampling offsets, or storing post-projection (and hence with varying precision emphasis) z coordinates as opposed to linear ones. No, it was just a stupid bad matrix.
Shots: here and here. Mmm, jagged chunkies. I also have no idea why Darin isn't casting shadows; probably he isn't writing to the depth buffer. Odd.
There's lots of cleanup that needs to happen here. Point lights don't cast shadows at all, and everything needs variance. I also need to build the various systems responsible for doing first-order radiosity stuff to see if it will work. However, the emphasis right now is on Ruby scripting and cleaning up the level editor now that I'm sort of back where I started with the renderer.
EDIT: Incidentally, the competition seems to be moving. The GG guys - or whatever's left of them, it's all very weird - have announced their new version of Torque complete with various exciting new bangs and whistles. The terrain system looks nice. The lighting system is a light-index deferred renderer... okay, same sort of idea as what I'm doing. I think they still have some kind of a lightmapping solution. The kicker is the price point hitch - you now don't get source code (without which I suspect Torque is pretty useless) or the advanced lighting for less than $1000 USD. This suits me fine in the long term. Torque has fundamentally never been a good product, and this now gives me more room to maneuver vis-a-vis price range. It also indicates that at least somebody else thinks I'm targeting the right things...