Wednesday, November 04, 2015

 

croissant

I've advanced enough in the past two weeks to reach a conclusion: the mmo option is not the right way. The reason is simple and has lots to do with sampling. Since the game has no tiles it's in continuous terrain, but this means that if a unit has to advance from a to b I must partition this portion of land into very small pieces, larger chunks of terrains would mean that the unit would jump between points, possibily flying over portions of land, and this is obviously not correct. For example if one has to travel 10 kilometers, and in the middle there's a small forest, the program would need to read that forest as well, to account for the smaller speed of march.
At this point it is clear that the optimal sampling is equal to the number of pixels of the textures, otherwise I could miss some terrain feature. Good, but this means to perform thousands of calculations for every unit every turn. I made some rough calculations and it turns out that I'd need one big amazon instance for every match...
So in the end I came back to the old idea of rather standard wego turns, but still I could not afford to perform the calculations, so I found a possible way to have the calculations performed by the end users' clients.
A new job is keeping me busy at the moment, but who cares, this game will never see the light.

This page is powered by Blogger. Isn't yours?