1.2.1

• Added [MethodImpl(MethodImplOptions.AggressiveInlining)] to many static functions to improve IL2CPP performance.
• Added compress() that accepts a float4 and uint4.
• Added math.project() and math.projectsafe() for vector projection.
• Added math.EPSILON, math.INFINITY, math.NAN and their double counterparts.
• Added [Serializable] to RigidTransform.
• Added math.ceillog2().
• Added math.floorlog2().
• Added math.down(), math.forward(), etc for Cartesian coordinate axes that match UnityEngine Vector3 equivalents.
• Added math.ispow2().
• Added half.MinValueAsHalf and half.MaxValueAsHalf to avoid having to explicitly convert from float.
• Added a float3x3 constructor which takes a float4x4 as input.
• Added [Serializable] to half types.
• Added some performance tests which can be run from the Unity test project.
• Added Random.CreateFromIndex() to assist in creating Random instances from loop indices.
• Fixed documentation bug where quaternion.RotateX/Y/Z referred to a float4x4 instead of quaternion.
• Fixed code generation bugs which could cause Windows and Mac to generate different test code.
• Fixed some test asserts which used NaNs and signed zeros which failed in IL2CPP builds.
• Updated documentation for math.countbits() to include equivalent names on Intel and ARM architectures to aid in discoverability.
• Fixed warnings for meta files existing even though the files they represent did not exist.