Our software project has been using boost for years and we are very comfortable with it.  We have a new requirement to calculate eigenvalues from 6x6 matrix.  This is our first foray into linear algebra.  At first glance it appears boost ublas supports the 'vocabulary' types of matrix and vector and basic operations between them.  I also read about boost bindings between ublas and other popular linear algebra libraries.  

I took a python class on linear algebra and remember near the end of the course we had to write our own algorithm for computing eigen values so its probably in the realm of possible todo this ourselves using ublas only.  

Or we could use something non boost like eigen:

Any strong opinions folks?