Reference for unit 'matrix'

Two, three, four dimensional matrix unit

Overview

The unit matrix is a unit that provides objects for the common two, three and four dimensional vectors matrices. These vectors and matrices are very common in computer graphics and are often implemented from scratch by programmers while every implementation provides exactly the same functionality.

It makes therefore sense to provide this functionality in the runtime library. This eliminates the need for programmers to reinvent the wheel and also allows libraries that use matrix operations to become more compatible.

The matrix unit does not provide n-dimensional matrices. The functionality needs of a general matrix unit varies from application to application; one can think of reduced memory usage tricks for matrices that only have data around the diagonal etc., desire for parallelization etc. etc. It is believed that programmers that do use n-dimensional matrices would not necessarily benefit from such a unit in the runtime library.

Design goals:

• Provide common dimensions, two three and four.
• Provide multiple floating point precisions, single, double, extended.
• Simple trivial binary representation; it is possible to typecast vectors into other implementations that use the same trivial representation.
• No dynamic memory management in the background. It must be possible to write expressions like matrix A * B * C without worrying about memory management.

Design decisions:

• Class object model is ruled out. The objects object model, without virtual methods, is suitable.
• Operator overloading is a good way to allow programmers to write matrix expressions.
• 3 dimensions * 3 precision means 9 vector and 9 matrix objects. Macro's have been used in the source to take care of this.