Should we go for our Vector2Int/Vector3Int implementation or just use built-in float based vectors?

Question

So in one of our projects we're using Vector2/Vector3's a lot but we do only use integer/byte values mostly. So up to the point we've implemented our Vector2Int, Vector3Int, Vector2Byte and so structures;

public struct Vector2Int
    {
        public int X;
        public int Z;

        public Vector2Int(int x, int z)
        {
            this.X = x;
            this.Z = z;
        }

        public override bool Equals(object obj)
        {
            if (obj is Vector2Int) return this.Equals((Vector2Int)obj);
            else return false;
        }

        public bool Equals(Vector2Int other)
        {
            return ((this.X == other.X) && (this.Z == other.Z));
        }

        public static bool operator ==(Vector2Int value1, Vector2Int value2)
        {
            return ((value1.X == value2.X) && (value1.Z == value2.Z));
        }

        public static bool operator !=(Vector2Int value1, Vector2Int value2)
        {
            if (value1.X == value2.X) return value1.Z != value2.Z;
            return true;
        }

        public override int GetHashCode()
        {
            return (this.X.GetHashCode() + this.Z.GetHashCode());
        }

        public override string ToString()
        {
            return string.Format("{{X:{0} Z:{1}}}", this.X, this.Z);
        }
    }

But i don't know if we're on right track? Should we just better use XNA's float based vectors because of performance-wise concerns?

On the other hand a Vector2Byte or Vector3Bytes consumes less space which is also quite important when you use so much vectors.

I've searched for tests / information on XNA's built-in vector performance but couldn't find a bit even on it.

I'm happy to hear your experiences.

Answer 1

This comes down to use the appropriate data structure for the task.

If you are using this for internal functionality key to your game and such then your own data types that hold the values in a format you are using them would make sense.

If however this is dealing with rendering and graphics hardware then you will want to use floats as that is what the hardware is generally built to use and accelerated for.

Hope this helps.

Answer 2

I would be more concerned that there are operations (divisions, square roots, et cetera) done on or to components of vectors that would cause integer-based representations to become inaccurate. Beyond that I'm not sure it's worth switching implementations mid-project.

There are minor potential performance nitpicks here and there, but none of them are likely to be serious. I would suggest using the XNA versions simply for uniformity and scalability, but probably only if you're about to start a new project. I don't see any compelling reasons to change everything up mid-stream.

Answer 3

I use integer vectors when my concern is in "which bucket" something occupies or "from which bucket onward" something occupies, and float vectors when my concern is "where in space" something "is."

Integers are much better than floats at bucketing things, and floats are much better at saying "where" something "is" in space.