I want to get eigenvalue of (Symmetric) Matrix.
Is there any source for Matrix Diagonalize?
(QR method etc..)
I want to see the source, please help me.
Thanks.

C source must be out there; it might be fairly
straightforward to port it to ASM.
APL is an excellent language if you have a lot
of work to do with vectors, matrices, and
polynomials.

Download the Intel Math library at www.intel.com

I don't suppose anyone can send my that library, or post alternate link for it. I just don't want to have to register.

I looked for it with Google. Bad news: MKL50.EXE
(Math Kernel Library) from Intel is 47.1 megs,
and that's the library and documentation, with
(I assume) no source.

You can find C++ sources at www.simtel.net. Do a search for "matrix library". Good luck!
This message was edited by karim, on 6/11/2001 7:05:32 AM

Hey!! I'm working in a code to calculete eigenvalues and eigenvectors!! Maybe two weeks to finish. :(
I'll post it here later.

Just curious... Are you working with finite elements?
That matrix is a tensor?
forget this if it doesn't matter...

Thank you for your reply.
I use matrix, eigenvalue for solid state physics
simulation.(Called D matrix)
It is symmetric, real. And it represents physical value,
so it is a tensor, of course.
Thank you for your interest.

Go to
http://www.ire.pw.edu.pl/numrcp/
and use your mouse to paste the code you want to a text file. You want part (or all) of Chap. 11 of "Numerical Recipes in C".
I have the book, but didn't want to spend $50 for the code in computer-readable format. Fortunately, the above site (with the publisher's blessing) lets you download everything. I downloaded the entire pdf file (section by section), edited the file quite extensively, and built a Windows dll file. This was not quite trivial, took me a few days. The numerical algorithms are excellent, the C code less so (requires editing). Great Monte Carlo, FFT, multivariate minimization code, etc.
If you *must* use assembler, just translate the algorithms from Numerical Recipes into assembler. There's little to gain from this, though -- good modern compilers generate quite efficient code.