Print

Please login or register.
1 Hour 1 Day 1 Week 1 Month Forever
Login with username, password and session length

[Kholint]

Did a quick search and couldn't find anything that mentioned this, so here you go (hope it's not already been linked to):

http://developer.amd.com/documentation/videos/pages/froblins.aspx

Really worth reading:
http://ati.amd.com/developer/SIGGRAPH08/Chapter03-SBOT-March_of_The_Froblins.pdf

What's especially exciting (and uncanny) about this is that the GPU-accelerated froblins world is tile/cell-based and populated with hundreds of little easily frightened, bumbling creatures that potter about en masse. Food for thought!

[Idiom]

Eventually.

Parallel processing and DF won't happen until someone else who can ends up with the DF project or Adams stocks up on enough aspirin and learning drugs to take a crack at it.

[macdonellba]

It's not really pathfinding as such: as explained in great detail in the article, the demo sacrifices global planning update frequency and restricts itself to a small set of goal types in order to make the gradient-based system work efficiently. As far as algorithms go, it's next to useless for interactive instructions, seeking out other units, or even just dealing with a significant number of goal types (due to the fact that the global planner must traverse the entire map periodically à la Dijkstra's algorithm.) In fact, the only real parallelism implemented in their model is mandated in order to make it perform as well as another running serially on a CPU. It's a nice tech demo, but not at all applicable to DF (or any game with actual dynamic goals.)