Re: [deal.II] Vector eigenvalue problem with SLEPc

2021-08-26 Thread Wolfgang Bangerth
On 8/25/21 9:37 PM, Anton Ermakov wrote: >>> You can do that, but I believe that it's also possible to work with the original system where one of the two matrices is only semidefinite. I have redone it with the original system, but it seems that PETSc is complaining about missing diagonal ent

Re: [deal.II] Vector eigenvalue problem with SLEPc

2021-08-25 Thread Anton Ermakov
>>> *That might not easily be possible with our PETSc interfaces.* *PETSc has the concept of MatShell, which is in essence what deal.II'sLinearOperator is: A class that implements a matrix-vector operation,without giving access to individual matrix elements. You could trywhether you can imp

Re: [deal.II] Vector eigenvalue problem with SLEPc

2021-08-25 Thread Wolfgang Bangerth
On 8/25/21 11:53 AM, Anton Ermakov wrote:  I  I¯  A  0  ¯I    *eigenvalue     +  I¯    0 B ¯ I I  * eigenvector =0  I  I_   0 0   _I                                  I_   C D   _I I  ¯                                                                     ¯ where eigenvector consists of a vector