From: P Zoltan <zoltan.padrah@gm...>  20091025 19:44:58

On Sun, 25 Oct 2009 20:27:08 +0100, Alan Grimes <agrimes@...> wrote: > 'ey Zoltan! 'wuz doing some maintanence on my PC (a perpetual > occupation) and came across a set of libraries that I hadn't previously > known about called gsl. > > http://www.gnu.org/software/gsl/ > > It appears to be a blas type system without the Fortran. =P > > I know you gave up with Eigen, perhaps gsl will prove to be useful. > Either that or I should be far less modest about my quickvector and > quickmatrix classes. ;) > > Look at the examples: http://www.gnu.org/software/gsl/manual/html_node/LinearAlgebraExamples.html That code looks like some OOP forced in C; also quite error prone. Here are the LU methods: http://www.gnu.org/software/gsl/manual/html_node/LUDecomposition.html It's not very sympathic to me... Actually I haven't gave up entirely on Eigen... Just the part with LU needs to be reworked: it we use LU, then we have to allocate a new matrix to store it  this is how the interface of Eigen is defined. However, for nonlinear equations, we need some other method to solve the equations, because it's inefficient to allocate a new matrix for LU in each attempt to solve the system. Another question is the overhead introduced by Eigen  I didn't manage to test it extensively (but it should be done, with and without debug info vs optimised code). Here are a few linear algebra software listed: http://en.wikipedia.org/wiki/List_of_numerical_analysis_software 