Add support for double precision floats

[aaronfranke]

Double-precision floats are necessary for games that take place on extremely large scales, such as in the vastness of space. There are workarounds, but they all come with issues. If Flax Engine wishes to support extremely large scales, it would make sense to add support for double precision floats.

For further justification on this feature, please see this proposal I made for Godot: godotengine/godot-proposals#892

As for the implementation, it is possible to support compiling with both single and double precision floats at the same time, if done correctly. This can be handled in C# via preprocessor directives by defining an alias, like this:

#if REAL_T_IS_DOUBLE
using real_t = System.Double;
#else
using real_t = System.Single;
#endif

The downside is that this alias definition must be repeated in every C# file. For a more detailed example, see this file in Godot: https://github.com/godotengine/godot/blob/master/modules/mono/glue/GodotSharp/GodotSharp/Core/Vector3.cs

[mafiesto4]

Added to the master! Can be enabled in a custom engine build by setting this to true:

FlaxEngine/Flax.flaxproj

Line 13 in 18cd7b6

"UseLargeWorlds": false

Soon there will be proper support for the rendering relative to the camera (work in progress) and implementation of 64-bit coordinates in other systems (audio, physics,..). Then, of course we will add documentation.

By setting UseLargeWorlds=true the Vector2/3/4 types are using double instead of float:

• C++
  

  FlaxEngine/Source/Engine/Core/Types/BaseTypes.h

  Lines 93 to 124 in 18cd7b6

  	#if USE_LARGE_WORLDS
  	// 64-bit precision for world coordinates
  	API_TYPEDEF(Alias) typedef double Real;
  	#define MIN_Real MIN_double
  	#define MAX_Real MAX_double
  	#else
  	// 32-bit precision for world coordinates
  	API_TYPEDEF(Alias) typedef float Real;
  	#define MIN_Real MIN_float
  	#define MAX_Real MAX_float
  	#endif
  	// Vector2
  	template<typename T> struct Vector2Base;
  	API_TYPEDEF() typedef Vector2Base<float> Float2;
  	API_TYPEDEF() typedef Vector2Base<double> Double2;
  	API_TYPEDEF() typedef Vector2Base<int32> Int2;
  	API_TYPEDEF(Alias) typedef Vector2Base<Real> Vector2;
  	// Vector3
  	template<typename T> struct Vector3Base;
  	API_TYPEDEF() typedef Vector3Base<float> Float3;
  	API_TYPEDEF() typedef Vector3Base<double> Double3;
  	API_TYPEDEF() typedef Vector3Base<int32> Int3;
  	API_TYPEDEF(Alias) typedef Vector3Base<Real> Vector3;
  	// Vector4
  	template<typename T> struct Vector4Base;
  	API_TYPEDEF() typedef Vector4Base<float> Float4;
  	API_TYPEDEF() typedef Vector4Base<double> Double4;
  	API_TYPEDEF() typedef Vector4Base<int32> Int4;
  	API_TYPEDEF(Alias) typedef Vector4Base<Real> Vector4;

• C#
  

  FlaxEngine/Source/Engine/Core/Math/Vector3.cs

  Lines 1 to 5 in 18cd7b6

  	#if USE_LARGE_WORLDS
  	using Real = System.Double;
  	#else
  	using Real = System.Single;
  	#endif

We do have vector types Float2/3/4, Double2/3/4 and Int2/3/4 to explicitly use certain precision and data size.