(1) Type Definitions

For the sake of brevity I created a simple naming convention (similar to Hungarian notation) to represent the most common collection of used modified types. This allows for a more succinct representation when writing code. For example [const float * __restrict const] becomes [PCU_TgFLOAT32]. The brevity makes the interface and code easier to understand once the nomenclature is understood.

Similarly the 3.1 files follow a similar pattern for the types used in templates and are as follows:

(2) File Conventions:

See Information for more details on the choice of language. Since I am using straight C I wanted a method to define the same interface and in some cases implementation for different combinations of variable types. In C++ this was done using template and template specialization. The method I used was to create a file convention for header, inline and sources files that would be included multiple times in the compilation with different pre-processor macro definitions. The macros are used for name mangling so that the functions can be generated for each of the types required. It is important that code itself is never written through macros as it is too troublesome when debugging. Since the code is only name mangled and included into the regular compilation this problem does not manifest.

(3) Geometry Nomenclature:

Each primitive type is assigned a simple two letter short form. The exception are points/vectors which sometimes has more definitions based on usage (point vs directions).

A point triangle is one described using only the three vertices and a normal to the plane. An edge triangle adds a clockwise edge definition. A collision triangle augments an edge triangle with feature reduction information (for instance if a particular edge should be considered during collision). Finally, a space triangle adds edge plane definition to a collision triangle.

(4) Function naming system:

tgDistSq: Returns the distance square between two primitives. Will return negative type max is they are interesting/penetrating.

tgClosestSq: Returns the distance square between two primitives. Will return negative type max is they are interesting/penetrating. There are two version of this function. The outputs are always listed first, and they will be either floats or vectors. If they're vectors they represent the points of closest proximity on each of the two primitives. If they're floats they represent the parametric values required to generate to point given the definition of the primitive. For instance for a line, it will be the scalar multiple along the direction vector from the origin; in the case of a triangle it will be the barycentric values.

tgClip: Like tgClosestSq this function often has two possible definitions. One takes a clip structure in which it returns the legal points, and the second returns back the two float values for a line. In the case of an open primitive its assumed that the object is scaled to infinity along its primary axis (for instance along the long axis of a tube).

tgContact_Penetrate: This is the first function where parameter order has an impact on the resolution. The values returned will be points on the second primitive describing the penetration of the first primitive.

tgContact_Sweep: Its assumed that the first primitive is static and the second primitive is moving along the sweep path.