ABSTRACT
The rendering of virtual globes requires constant reading from disk and network connections to stream in new terrain, imagery, and vector data. Before these new data are rendered, they usually go through CPU-intensive processing and are made into a renderer resource like a vertex buffer or texture. We call these tasks as a whole resource preparation. Doing resource preparation in the rendering thread dramatically reduces performance and responsiveness because it frequently blocks the thread on I/O or keeps it busy number crunching instead of efficiently feeding draw commands to the GPU.
