Skip to content

Latest commit

 

History

History
48 lines (36 loc) · 3.67 KB

README.md

File metadata and controls

48 lines (36 loc) · 3.67 KB
Raw

agow

About

An experiment in fragment shading, open world simulation, and style generation.

Cityscape with grass and water

View of buildings that overloaded Bullet physics at 60 fps

Lower-level fiew of trees, grass, and rocks

The Map Editor for the game

Status

Agow is on indefinite hold, as I've learned enough to move onto more feasible projects.

I started developing agow with a compelling concept, that unfortunately didn't pan out once I started writing the story. Like [TemperFine] (https://github.com/GuMiner/TemperFine) I learned a significant amount, which for this project included game design, OpenGL performance, Bullet physics and GIS data systems.

What did work:

  • Decoding King County into a series of 1000x1000 pixel heightmaps, breaking those heightmaps into smaller 100x100 heightmaps, and dynamically-loading the sub-heightmaps in real-time and no visible seams.
  • Generating grass, trees, buildings, etc. from the type assigned to each heightmap grid pixel.
  • Using external dependencies (GLM/STB/etc) to enable me to focus on the unique aspects of this project.
  • Segmenting buildings for random height generation and interesting building demolitions.
  • Event-based dialog, with an auto-paging dialog pane.
  • Basic integration of real-time Bullet physics with OpenGL rendering.

What didn't work:

  • Dynamically changing a subset of items from being drawn dynamically to statically every frame. Each building segment is initially drawn statically, however if it is hit with a collision that segment is added to a list of items whose matrices are sent to OpenGL every frame. While drawing all the building segments statically is efficient, and some of the building segments dynamically is also efficient, moving segments between these two lists is very inefficient.
  • Insufficient design. The premise was compelling, but not feasible for a single-person prototype, much less a full game. While I leveraged common code between TemperFine and agow, I didn't create [GuCommon] (https://github.com/GuMiner/GuCommon) until much later. The AI wasn't fully thought out, as I was assuming behavioral imputs that were not easily achievable for the AI to use. My attempts at creating TODO lists were not comprehensive nor cohesive enough, which served as a good warning indicator that the premise would not lead to a feasible nor compelling story.
  • Bullet physics doesn't perform well with an immense number of physical objects (3D buildings) in realtime. I attempted to work around this problem by assigning a ghost object to each building, that would spawn into building pieces when hit by another object. When that wasn't sufficient, I split the physics computation onto a separate thread. Unfortunately, that brought back the lag and interpolation problems that I had encountered previously writing this game, but in an unavoidable manner that became significantly worse when many physical collisions were happening simultaneously.

Dependencies

Licenses for dependencies are within the licenses folder.