Jump to content
Posted 27 June 2012 - 03:30 AM
Posted 27 June 2012 - 04:36 AM
Posted 27 June 2012 - 04:50 AM
Here streaming meant to load data on-the-fly while the game is running, this is common practise in most modern games.
what do you mean by stream
Edited by Ashaman73, 27 June 2012 - 04:54 AM.
Posted 27 June 2012 - 04:55 AM
Posted 27 June 2012 - 05:02 AM
If you want to avoid threads, you could try to stream/post process only a little portion of the city block per frame until a whole city block has been loaded.
Posted 27 June 2012 - 05:07 AM
Posted 27 June 2012 - 05:41 AM
Edited by mark ds, 27 June 2012 - 05:50 AM.
Posted 27 June 2012 - 05:56 AM
Edited by DiegoFloor, 27 June 2012 - 06:06 AM.
Posted 27 June 2012 - 06:05 AM
Posted 27 June 2012 - 12:24 PM
Posted 27 June 2012 - 06:12 PM
Posted 27 June 2012 - 10:49 PM
Posted 28 June 2012 - 04:56 AM
You chose an interesting problem to look at. Does the number of people in the crowd have any effect, or only the area? Just wondering whether it would help if you could in the distance have "trains" of people, e.g. 10 low poly people in a line moved and animated together.
As far as streaming chunks of city, definitely a good plan. How would you manage people moving between chunks? I'm wondering whether individuals would be recognisable enough that you could tell if they were destroyed if you went away then came back.
Another possibility is to move only the ones you can see at a high refresh rate. You could update the rest less often but at a higher velocity, e.g. 4 times as fast every 4 frames. Or if the crowd is a uniform density skip the vector fields for out-of-sight people altogether by just saying that each person takes the place of the person in front of him/her (and fall back to vectors if two people pick the same target).
Just a few random thoughts. Let me know if I'm off-base.
Posted 28 June 2012 - 06:06 PM
Posted 28 June 2012 - 06:34 PM
Posted 28 June 2012 - 07:13 PM
Posted 29 June 2012 - 05:00 AM
With the method you're using you could perhaps put some performance enhancements in, e.g. skip calculating cells with nobody in them (which may happen with very small cells), use something more agent based on low-density areas, etc.
Posted 04 July 2012 - 04:38 PM