The number of unknowns in the third case is 89373 (31*31*31*3). I ran it on
my pc with a cpu of i7-7700K. For the cg case, it used the PETSc's cg and
PETSc's blockJacobi preconditioner (which is the default setting in step-17).
The timing includes the time to build the preconditioner and the time to
solve. And I re-checked the timing, I it really just about 0.4s (to timing cg,
I only added timing, and changed the meshing process, and did no other
modifications to step-17).
BlockJacobi builds an LU decomposition of that part of the matrix that is
stored locally. So it's a really expensive preconditioner to build (which I
gather you don't include in the time?) but then solves the problem in only a
few iterations. If you want a fair comparison, you need to include the time to
build the preconditioner.
I understand the situation now. I am interested in the "block" variants of
CG you mentioned, does TrilinosWrappers or PETScWrappers has that kind of CG
variant?
There are no wrappers. I believe that Trilinos has them, and PETSc does not.
So you would have to interface with Trilinos directly.
Best
W.
--
------------------------------------------------------------------------
Wolfgang Bangerth email: bange...@colostate.edu
www: http://www.math.colostate.edu/~bangerth/
--
The deal.II project is located at http://www.dealii.org/
For mailing list/forum options, see
https://groups.google.com/d/forum/dealii?hl=en
---
You received this message because you are subscribed to the Google Groups "deal.II User Group" group.
To unsubscribe from this group and stop receiving emails from it, send an email
to dealii+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
