Optimal performance and minimal code footprint are essential to a responsive 3D experience. 3D worlds are punishing to deliver, as they have to perform millions of rendering and motion calculations reliably every 16 milliseconds. Internet bandwidth constraints magnify these challenges; if we cannot keep world assets and the underlying engine's code compact, it will be impossible to stream them from the server to the client on a timely basis.


Although Cone's initial focus is on web-delivered 3D, it is designed to also support native platform 3D experiences. It is also intended to be usable for building 3D world servers, keeping multi-player clients synchronized to the most recent state of the world. To satisfy these focus areas, Cone has been designed to deliver blazingly-fast runtime performance (and to compile quickly too!).

A large part of Cone's performance story comes from obvious architectural choices. Cone's statically-typed programs are compiled to native executables supporting all popular platforms: Windows, Linux, Macintosh, Android, iOS, and WebAssembly. Cone accomplishes this using LLVM's award-winning backend technology and its powerful suite of code optimizers.

Inspired by the Rust and Pony languages, Cone's static type system aggressively reduces runtime overhead further:

In addition, Cone's easy-to-use custom regions, such as pools and arenas, go a long way towards lowering the runtime cost of frequent memory allocation and free events, commonly required by 3D worlds.

Minimal footprint

To keep 3D program assets small, Cone's 3D libraries are guided by a modular, opt-in philosophy. Rather than making users wait for a monolithic, do-everything 3D engine to finish loading, 3D worlds can pick-and-choose the 3D features they require and schedule when to downloaded them in the background so as to be available when required.

The Cone runtime for 3D web is quite small, as it leverages code logic already resident in the client browser.