As graphics gets pushed more and more onto specialized hardware, programmers are finally getting the spare CPU cycles they need to bring game physics to realistic levels. Unfortunately, as realism increases, so does the complexity of the code. Starting off with the right data structures can greatly simplify the tedious, error-prone task of programming and maintaining a physics engine, and this article offers suggestions for helping to structure your C++ code more efficiently .
Why Use C++ instead of C?
It seems like I spend about 10% of my time writing code and 90% of my time debugging it, so clean code makes my job a lot easier. The C++ programming language offers many nice extensions to C while keeping its clean, flexible syntax. With it I can write clean, optimized code quickly so I can get to the serious business of finding out why it doesn’t work. Here are the main reasons why I prefer C++ over C.
1. Function inlining. It is a common misconception that C++ produces slow, bulky code and is therefore unsuitable for game programming. With function inlining, C++ can produce functions that are as fast as macros. But unlike macros, inline functions have strict type checking and are viewable in a debugger. And let’s face it: if you really need blistering speed, you write it in assembly. For a more detailed comparison of the performance of C and C++, see  in the bibliography at the end of this article.
2. Operator overloading. If my house was burning down and I could take only one feature of C++ with me, it would have to be the ability to overload operators. Operator overloading allows you to use the language’s built in operators (+, -, /, etc.) with user defined types as arguments. As an example, let’s look at linear interpolating between two vectors:
const VECTOR c = (1-u)*a + u*b;
The above C++ statement reads as if it was copied directly from a textbook. The fastest way to do the same operation in C is with macros:
VECTOR s, t, c;
vmul( s, a, 1-u );
vmul( t, b, u );
vadd( c, s, t );
What took one line to write in C++, took four lines to write in C, and it’s is not very easy to read. It can get worse. I didn’t fully appreciate C++ until I wrote the physics engine for a racing game in C. The code took ten times as long to write, and when I came back a week later to fix a bug, I could barely follow what I had written.
3. Inheritance. In C, building complex data types can be done by including structures within structures. This makes accessing deep down members a real pain:
VECTOR R1 = body.PreviousState.Frame.Basis;
In C++, data types can be built with inheritance, so any member variable or function a base class had the derived class has as well. By using inheritance, you can make the above C statement look like:
VECTOR R1 = body.Basis;
Now that I’ve rambled on about how great C++ is, let me start describing some useful data types.
The SCALAR type
In any meaningful physics program, you’ll need to represent decimal numbers. You could explicitly make everything a float or a double, but what would you do if your program requirements change mid-project and you suddenly need double’s instead of float’s, or vice-versa? You’d have to globally replace every declaration in your program. Or what if you’re randomly experiencing floating-point overflows and you need some range checking in debug mode? Then you might have to go through and wrap every arithmetic operation with a macro. In any case, it’s probably not the best idea to hardcode a standard type for your scalar quantities. The SCALAR type (see Listing 1) is there to abstract your scalar quantities so you can easily change its internal representation.