Looking for an elegant and efficient C++ matrix library
Greetings, googling for that subject brings, e.g., MTL, exmat, LAPACK and also here. I also seem to remember that Microsoft Research released one, but can't put my hands on it. I look for advice from someone who actually used (or developed...) one of those, hoping to achieve a Matlab experience inside C++ (as much as possible). Thanks in advance, Robi
I use both Eigen and Matlab and like both of them a lot. Eigen supports SIMD and lazy evaluations for extra performance. But users don't need to bother with internals. Eigen's interface is very simple and intuitive. Going from Matlab to Eigen should be relatively straightforward, which I can't say about uBLAS or LAPACK.
EDIT: Here is Eigen Quick Reference for Matlab Users
boost has a math library capable of matrix algebra - uBLAS.
NIST has the TNT (template numerical toolkit). Very lightweight and simple.
A semi-serious answer (straight from Stroustrup): http://www.stroustrup.com/Programming/Matrix/index.html
Read Jack Crenshaw's articles over at www.embedded.com. He's been working up a matrix class for a number of years, while developing the numerical methods code that goes with it and uses it.
I just started using the Gmm++ library. It's header-files only, supports sparse matrices, has a wide array of solvers, and interfaces for LAPACK and BLAS. Its interface doesn't seem as nice as Eigen, but it is more complete.