n2k wrote:

bjacob wrote:If you're using linux or unix you can always make it work on a huge matrix so that it takes like 10 seconds to complete, and during the computation you run 'top' in a terminal. If it says "200%" this means it's using 2 cores of your CPU. Of course this experiments only works if you have a dual core CPU. If you have 1 core, disregard all I said!

I do have 2 cores. I observed that armadillo (blas/lapack) reached 191% while matlab exceeded 200%. Eigen spiked at 101%. So it actually seems that armadillo and matlab are using two cores but this happens only after a few seconds (I'm not sure if this is due to the refresh rate of "top").

I also did other tests and Eigen turned out to be the fastest at smaller matrix sizes. These are the results for the solution of a 50x50 system:
Eigen (partial lu) -> 0.216 ms.
Armadillo -> 0.238 ms
Eigen (full lu) -> 0.332 ms.
Matlab -> 0.350ms

Then increasing the matix size Eigen seems to slow down more rapidly than the other two.