Re: Wanted: Deciduous Tree Models
Posted: 22 Jan 2008, 05:15
Ok, for my purposes, it only needs to run once. So don't worry about CPU costs- it's going to be very costly, but only for one frame. After that, it's not going to run again. I run a second pass, making everything Neutral, etc., but that's straightforward, and it already works. So, two frames of big CPU usage, then it all calms down, and it's down to script speed, LOS issues, and other performance drains, but this will help a lot, because I can then play with the size and frequency of clumping to help mitigate this, instead of relying on the frequency of Trees and random factors.
The stuff for mobile actors will need to check the blocking map, if I'm going to want to place them during gameplay, and that will be rather expensive, but that's a problem for another day.
So, checking the blocking map and mapping out "no-go zones" for the Commanders should be fine. Simply block out a 96X96 zone around each Commander, or the closest matching set of grid coordinates, and that's fine.
As for sizes, in an ideal world, it'd read a config, which would tell it that it was looking for a "good spot" that was 32X32, 64X64, or 96X96- I think any larger than that is probably pointless, as it'd lead to a lot of bad solutions. Then, once it's found a given "good spot", I can put in logic placing actors with X,Z units of the spot's center, to confuse the issue of it using a grid-based system and make it feel natural (it'd be good if the x,z coordinates of that center were readily-available in an obvious-to-Argh variable, like "centerOfGoodSpotX,centerOfGoodSpotZ" lol
The stuff for mobile actors will need to check the blocking map, if I'm going to want to place them during gameplay, and that will be rather expensive, but that's a problem for another day.
So, checking the blocking map and mapping out "no-go zones" for the Commanders should be fine. Simply block out a 96X96 zone around each Commander, or the closest matching set of grid coordinates, and that's fine.
As for sizes, in an ideal world, it'd read a config, which would tell it that it was looking for a "good spot" that was 32X32, 64X64, or 96X96- I think any larger than that is probably pointless, as it'd lead to a lot of bad solutions. Then, once it's found a given "good spot", I can put in logic placing actors with X,Z units of the spot's center, to confuse the issue of it using a grid-based system and make it feel natural (it'd be good if the x,z coordinates of that center were readily-available in an obvious-to-Argh variable, like "centerOfGoodSpotX,centerOfGoodSpotZ" lol
