Intel® oneAPI Math Kernel Library
Ask questions and share information with other developers who use Intel® Math Kernel Library.

QZ algorithm and some advice

Dear all,

I need to solve a generalized non-hermetian eigenvalue problem with the QZ algorithm, however I have a special structure in my matrices, namely, given the problem as,

A\\phi = \\lambdaB\\phi

Matrix A is structured as

[ A11 A12 ]
[ 0 A22]

so there is a large zero block. I was wondering if I can make use of this large zero block in the computations with the routines in the MKL related to QZ algorithm? Or is there a way to make use of this block structure in the algorithm, since I need to solve a rather dense eigenvalue problem at each iteration of an algorithm that I am working on, this sometimes increases the cost. So, I was wondering if some optimization is possible or not?

0 Kudos
0 Replies