Being that you guys already have the gmp mp libraries as part of the mkl, would it not be useful to consider adding support for gmp mp data types in addition to that of float and double ??
This would basically extend calculations to pure numerical formats whereby you never want to lose accuracy, as well as arbitrary length floating point data types where you simply want more accuracy than a double.
( Im referring here to the idea of adding these types to vector/matrix types & functions et al )
Just a thought :)
