Current Version 2.1  subversion revision #2227+
- new directory structure - all genFem-depending stuff will go below genFem
- but still possible to use genfem as standalone library from "outside" for external projects with a lot of dependencies

cadxfem
 |- genFem
 |   |
 |   |- applications
 |   |   |- piezo
 |   |   |- ...
 |   |
 |   |- tests
 |   |- genTensors
 |   |- autodiff
 |   |- ...
 |
 |- ...
 
Version 2.0beta - subversion revision #1027
- use genTensorX<class T,int N> instead of the STensor family, T can be any scalar type with the usual operators, and N>=1 is the dimension.
  There is no container class for scalars yet ( X>=1 )
  The genTensorX classes are in a separate folder.
- added a fullScalar<class T> container (replacing the bare use of T), in line with fullVector and fullMatrix classes
  rationale : automatic type inference mechanisms expects this, this allows writing code that is more reusable ane easier to debug
- now one can use any scalar type as base type (defaults to double)
- also , the dimension of space where the objects live can be adjusted (e.g. dimension 2 for shells) 
- Matricial operators (Compose for instance) are now completely independent to tensorial operators (full contracted product for instance)
  see e.g. genTermCompose.h , genTensorsFullContrProd.h and genTermApply.h 
- introduction of class dfloat<T> : "scalar" type for automatic differentiation. Again, it lives in a separate folder for now.
- Big cleanup in genOperators. Now almost all operators are written using Apply / or Compose.
- Added partial support to FTensor


Version 1.0 (subversion revision #988)
- Use of STensorXX family classes from gmsh
- Fixed dimension = 3
- Fixed scalar type = double
- Huge set of operators (basically, any combination between matricial and tensorial operators had to be hardcoded)
- Use of smart pointers instead of references
- introduction of arithmetic operators on forms