# Is the QR algorithm for computing eigenvalues efficient for today's standards?

I was looking at the QR factorization algorithm of a matrix to approach eigenvalues. At the Wikipedia page they state that it was developed in the 50's and took over the LR algorithm. They also state a few variants. However I couldn't find what the state of this algorithm is today.

Is it still efficient for today's standards? Is it used in modern mathematical software to compute eigenvalues? If not, are there any variants that are considered more efficient?

-

I'm no expert, but I believe QR is alive and kicking. To expand on lalala's answer, last I checked, MATLAB's general-purpose eig() function invokes LAPACK's QR routines to do its work for general matrices. For matrices that have special structure, e.g. symmetric matrices, there are alternatives that may be better, but for arbitrary matrices, QR is still the standard as far as I know.

As a side note, I wouldn't let the fact that the QR algorithm was developed in the 1950s put you off. In fact, there are lots of algorithms that were developed in that era that were shelved at the time because the computers back then were not fast enough to make them practical. This is especially true for a lot of iterative methods. For instance, the conjugate gradient algorithm was discovered around that time and set aside for years only to become utterly ubiquitous in modern-day scientific computing!

-
Thanks for the answer.. The additional information is very interesting too. –  geo909 Nov 24 '12 at 18:52