ID:137126
Feb 11 2002, 5:05 am
|
|
Does anyone have a rough estimate as to how well BYOND could handle a few tens of thousands of objects on-screen?
|
Leftley wrote:
Does anyone have a rough estimate as to how well BYOND could handle a few tens of thousands of objects on-screen? As I understand this, the place where you really get in trouble is with client downloads. Icons aside (which were a huge problem for early versions of Incursion), every client needs to be able to see what's on their part of the map, and it could get hairy. Of course workload can be intensified greatly by other factors: Any kind of for(n in m) loop covering too much area could kill you. If you have lots of objs running around or performing AI routines, that can be big trouble. For object numbers in the tens of thousands, most of the ones in question should be "usually inert" or at least not become active until a player is within a certain range. You also need to watch out for BYOND's object limit--currently 65536. You'd be pushing against that. Lummox JR |
In response to Dan
|
|
Dan wrote:
Leftley wrote: Didn't think so. Is this for further development of that nifty dungeon system you were working on? I do have vague dreams of greatly expanding that, but to get that many objects on screen I'd have to boost the range up to some truly insane levels, far, far past the range where visible objects would be large enough to be meaningful. I was wondering because I'm working on a game that does thing on a pixel scale (well, 2x2 blocks of pixels--it's designed for small icon mode, though people with very high resolutions could display it full size) and was trying to figure out if it would be more practical to draw up new icons (my current system) or track each individual pixel object separately. I didn't think the latter would really work, but it seemed like an attractive possibility due to the amount of lazy shortcuts it would seem to offer. |
In response to Leftley
|
|
Leftley wrote:
I was wondering because I'm working on a game that does thing on a pixel scale (well, 2x2 blocks of pixels--it's designed for small icon mode, though people with very high resolutions could display it full size) and was trying to figure out if it would be more practical to draw up new icons (my current system) or track each individual pixel object separately. I didn't think the latter would really work, but it seemed like an attractive possibility due to the amount of lazy shortcuts it would seem to offer. The satellite map in Tanks uses up to 2401 8x8 overlays on 196 client screen atoms. On slower machines, it gets pretty sluggish even when the icons themselves do not change. I've yet to try it with single icons instead overlays, because I was using my own lazy shortcuts. ;) [edit] Your post prompted me to finally work on the merged icon display. It's amazing what can happen when you divide the number of images diplayed by a factor of 12! :) |
You mean, all on-screen at the same time?! We recently optimized the algorithms to make them more scalable, but I am quite sure that tens of thousands of objects on a single screen would still be too sluggish. And in network mode, you can be sure it will be impractical.
Is this for further development of that nifty dungeon system you were working on?
--Dan