Improved solution of nonsymmetric eigenvalue problems
I am posting this question here because I do not know how to contact Research & Development at ANSYS directly.
I am researcher in the field of high performance scientific computing. I work at Umeaa University in Sweden. My colleagues and I have produced a new library (StarNEig) for solving dense standard and generalized nonsymmetric eigenvalue problems. It can be substantially faster than either LAPACK with parallel BLAS or ScaLAPACK. Moreover, it realizes new parallel and block algorithms for computing eigenvectors without suffering from floating point overflow.
I am trying to determine the need for solving nonsymmetric eigenvalue problems in industry. I gather that your products are solving the quadratic eigenvalue problem (lambda^2*M + lambda*C+K)x = 0 and that you must support general rather than symmetric matrices M, C, K, because you accept rotating coordinate systems.
I would dearly like exchange emails with someone in ANSYS R & D. My immediate objective is to learn if you could use better software for solving nonsymmetric eigenvalue problems. I am also looking for information that can help guide the future development of the StarNEig library. What features would be desirable and what performance is wanted in the future?
StarNEig is freely available to anyone. The initial development of StarNEig was funded by the Horizon 2020 project NLAFET. I need to document that our project is needed by industry in order to improve the chances of obtaining additional funding from the Swedish government.