[petsc-users] Accelerate convergence of some specified unknowns of the matrix equation.

[Jed Brown]

This usually isn't possible for iterative solvers and saves very little
for direct solvers (you still have to pay the factorization cost, but
can save a little on solve cost which is already vastly cheaper).

While it's easy to make the residual zero for some subset of unknowns,
that says nothing about the solution, which can still be completely
wrong until the solution converges everywhere.

Ji Zhang <gotofd at gmail.com> writes:

> Dear all,
>
> I'm a PETSc user, solving some matrix equations using GMRES method. The
> matrix is dense and huge, i.e. 10,000~100,000 unknowns.  But I'm only
> interested in the last six unknowns. Is it possible to accelerate the
> convergence rate of them, even others are less accuracy.
>
> Thanks.
>
> Best,
> Regards,
> Zhang Ji, PhD student
> Beijing Computational Science Research Center
> Zhongguancun Software Park II, No. 10 Dongbeiwang West Road, Haidian
> District, Beijing 100193, China
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 832 bytes
Desc: not available
URL: <http://lists.mcs.anl.gov/pipermail/petsc-users/attachments/20170802/839cad78/attachment.sig>