Wed Feb 3 17:22:40 CET 2010

Never invert a matrix!

Actually, there is a reason why one would want to invert it anyway: if
the same matrix inverse is applied to a large number of vectors
inversion might pay off because it can be parallellized.  

Solving a set of equations still has a longer data dependency chain.

[1] http://www.johndcook.com/blog/2010/01/19/dont-invert-that-matrix/