Question

In the help for XNA Vector3, two of the public methods that it lists are Subtract and op_Subtraction. What’s the difference between them and when should one be used instead of the other?

Answer 1

op_Subtraction (the - operator implementation) and Vector3.Subtract are essentially the same thing. They create a new vector with values from a subtraction operation. Vector3 is a value type, so there is no memory-management cost, however there is a small construction cost.

The other overload of Subtract is slightly different:

void Vector3.Subtract(ref Vector3 value1, ref Vector3 value2, out Vector3 result)

This one takes references for its input and for its output. It is better for data-oriented programming (for example: a tight loop in, say, a physics engine, which takes data from one array and outputs it into another array). This removes the construction cost, and the cost of passing 6 floats as arguments, however there is still a tiny cost for the function call and dereferencing.

You can get even faster by doing the subtraction directly:

result.X = value1.X - value2.X;
result.Y = value1.Y - value2.Y;
result.Z = value1.Z - value2.Z;

Of course, as you select faster options, you also lose readability! This is much clearer:

result = value1 - value2;

So to answer your question – you should absolutly use the vector - operator when doing vector subtraction, unless you’re writing some kind of very high-performance data-oriented thing and you have profiled your code and found it to be a worthwhile performance improvement.

If you’re interested in more low-level performance information like this, take a look at Understanding XNA Framework Performance by Shawn Hargreaves. (In fact it includes a benchmark comparing the three methods I have listed here, when used in a particle system.)