MetaPhysicL is a set of template classes and metaprogramming tools
useful for the generation of physics codes which are optimized at
compile-time.

This includes many classes which are designed to be compatible with
generic codes written for plain numeric data types like float and
double, but which may provide more featureful output:
* DualNumber, for automatic differentiation. 
* NumberArray, for vectorization.
* SparseNumberArray and SparseNumberStruct, for homogeneous or
  heterogeneous sparse vectors whose structure can be inferred at
  compile time.
* EquationSet, for the automatic assembly of physics equations into
  evaluation codes.

WWW: https://github.com/roystgnr/MetaPhysicL