Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
> > Absolutely. Also, with Krylov-Schur you can adjust the restart parameter > (which is hidden in ARPACK); it may help improve convergence in some cases. Awesome. I'll just use Krylov-Schur then. Thanks. On Wed, Mar 25, 2015 at 5:18 PM, Jose E. Roman wrote: > > El 25/03/2015, a las 22:06, Harshad Sahasrabudhe escribió: > > > > > I thought ARPACK was faster when the system size is large and number of > eigenvalues required is small. I will be working with sparse matrices of > size ~60,000. Does SLEPC's Krylov-Schur have about the same performance as > ARPACK for calculating ~50 eigenvalues for such matrices? > > > > Absolutely. Also, with Krylov-Schur you can adjust the restart parameter > (which is hidden in ARPACK); it may help improve convergence in some cases. > > http://slepc.upv.es/documentation/current/docs/manualpages/EPS/EPSKrylovSchurSetRestart.html > > Jose > >
Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
El 25/03/2015, a las 22:06, Harshad Sahasrabudhe escribió: > > I thought ARPACK was faster when the system size is large and number of > eigenvalues required is small. I will be working with sparse matrices of size > ~60,000. Does SLEPC's Krylov-Schur have about the same performance as ARPACK > for calculating ~50 eigenvalues for such matrices? > Absolutely. Also, with Krylov-Schur you can adjust the restart parameter (which is hidden in ARPACK); it may help improve convergence in some cases. http://slepc.upv.es/documentation/current/docs/manualpages/EPS/EPSKrylovSchurSetRestart.html Jose
Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
> > Anyway, why do you insist in using ARPACK when SLEPc's Krylov-Schur work? > ARPACK will not give you any further advantage. I thought ARPACK was faster when the system size is large and number of eigenvalues required is small. I will be working with sparse matrices of size ~60,000. Does SLEPC's Krylov-Schur have about the same performance as ARPACK for calculating ~50 eigenvalues for such matrices? On Wed, Mar 25, 2015 at 5:00 PM, Jose E. Roman wrote: > > El 25/03/2015, a las 21:47, Harshad Sahasrabudhe escribió: > > > With MUMPS you should not get spurious eigenvalues. > > I get only a few spurious eigenvalues when using MUMPS with ARPACK, but > the eigenvectors are definitely wrong. > > > > Did you try the krylovschur solver? > > Yes, Krylov-Schur gives me correct results. > > > > How do you know the eigenvalues are wrong? > > I'm testing my implementation of Schrodinger equation solver with a 3D > harmonic oscillator potential. I'm getting correct results using FEAST and > krylovschur solvers. > > > > Are you setting problem type to GHEP? > > Yes > > > > Did you try the arpack-ng version? Seems that some people have taken over > maintainance. > > Anyway, why do you insist in using ARPACK when SLEPc's Krylov-Schur work? > ARPACK will not give you any further advantage. > > Jose > >
Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
El 25/03/2015, a las 21:47, Harshad Sahasrabudhe escribió: > With MUMPS you should not get spurious eigenvalues. > I get only a few spurious eigenvalues when using MUMPS with ARPACK, but the > eigenvectors are definitely wrong. > > Did you try the krylovschur solver? > Yes, Krylov-Schur gives me correct results. > > How do you know the eigenvalues are wrong? > I'm testing my implementation of Schrodinger equation solver with a 3D > harmonic oscillator potential. I'm getting correct results using FEAST and > krylovschur solvers. > > Are you setting problem type to GHEP? > Yes > Did you try the arpack-ng version? Seems that some people have taken over maintainance. Anyway, why do you insist in using ARPACK when SLEPc's Krylov-Schur work? ARPACK will not give you any further advantage. Jose
Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
> > With MUMPS you should not get spurious eigenvalues. I get only a few spurious eigenvalues when using MUMPS with ARPACK, but the eigenvectors are definitely wrong. Did you try the krylovschur solver? Yes, Krylov-Schur gives me correct results. How do you know the eigenvalues are wrong? I'm testing my implementation of Schrodinger equation solver with a 3D harmonic oscillator potential. I'm getting correct results using FEAST and krylovschur solvers. Are you setting problem type to GHEP? Yes On Wed, Mar 25, 2015 at 4:37 PM, Jose E. Roman wrote: > > El 25/03/2015, a las 21:29, Harshad Sahasrabudhe escribió: > > > Hi, > > > > I'm trying to use the ARPACK interface in SLEPc for calculating smallest > eigenvalues with eigenvectors of a generalized eigenproblem. The matrices > are symmetric. > > > > What are the suggested linear solvers/preconditioners for this type of a > system when using ARPACK? I am using shift and invert with a shift of 0. > > > > I get a lot of spurious eigenvalues when I use GMRES linear solver or > MUMPS for LU factorization. Chebyshev doesn't seem to converge (I don't > have a good guess for the higher end of eigenvalues). > > > > Thanks, > > Harshad > > With MUMPS you should not get spurious eigenvalues. Did you try the > krylovschur solver? How do you know the eigenvalues are wrong? Are you > setting problem type to GHEP? > > Jose > >
Re: [petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
El 25/03/2015, a las 21:29, Harshad Sahasrabudhe escribió: > Hi, > > I'm trying to use the ARPACK interface in SLEPc for calculating smallest > eigenvalues with eigenvectors of a generalized eigenproblem. The matrices are > symmetric. > > What are the suggested linear solvers/preconditioners for this type of a > system when using ARPACK? I am using shift and invert with a shift of 0. > > I get a lot of spurious eigenvalues when I use GMRES linear solver or MUMPS > for LU factorization. Chebyshev doesn't seem to converge (I don't have a good > guess for the higher end of eigenvalues). > > Thanks, > Harshad With MUMPS you should not get spurious eigenvalues. Did you try the krylovschur solver? How do you know the eigenvalues are wrong? Are you setting problem type to GHEP? Jose
[petsc-users] SLEPc - Generalized symmetric eigenproblem using ARPACK
Hi, I'm trying to use the ARPACK interface in SLEPc for calculating smallest eigenvalues with eigenvectors of a generalized eigenproblem. The matrices are symmetric. What are the suggested linear solvers/preconditioners for this type of a system when using ARPACK? I am using shift and invert with a shift of 0. I get a lot of spurious eigenvalues when I use GMRES linear solver or MUMPS for LU factorization. Chebyshev doesn't seem to converge (I don't have a good guess for the higher end of eigenvalues). Thanks, Harshad