So... I just committed the initial round of changes to allow attaching a
Preconditioner object to either a linear or nonlinear solver and it
works!
I'm already using this in my code with a special preconditioner object to do
fairly complicated solver library (petsc/trilinos) agnostic preconditi
Dear Derek,
On Mon, 9 Feb 2009, Derek Gaston wrote:
> On Feb 9, 2009, at 9:18 AM, John Peterson wrote:
>
>> Short answer: check out aclocal.m4 around line 869. We just use grep
>> and sed to figure out what the petsc version is during configure. You
>> could probably do the same for HYPRE_foo..
On Feb 9, 2009, at 9:18 AM, John Peterson wrote:
> Short answer: check out aclocal.m4 around line 869. We just use grep
> and sed to figure out what the petsc version is during configure. You
> could probably do the same for HYPRE_foo...
I just committed a configure test for Hypre that will def
On Mon, Feb 9, 2009 at 9:55 AM, Derek Gaston wrote:
> So it looks like we only snoop petscconf in our Makefile(.in). Should I
> just snoop for HYPRE_LIB there and add a -DPETSC_HAS_HYPRE to our compile
> line? Or should I go snoop petscconf manually during configure so a #define
> can get set in
On Mon, 9 Feb 2009, Derek Gaston wrote:
> Does anyone know how to snoop to see if Petsc is configured with Hypre
> support? I don't see a simple #define HAS_HYPRE or some such anywhere
> in the Petsc source.
Does PETSc end up including any HYPRE headers from the PETSc headers?
If so then #ifdef
So it looks like we only snoop petscconf in our Makefile(.in). Should
I just snoop for HYPRE_LIB there and add a -DPETSC_HAS_HYPRE to our
compile line? Or should I go snoop petscconf manually during
configure so a #define can get set in libmesh_configure.h?
Derek
-
On Feb 9, 2009, at 8:46 AM, John Peterson wrote:
> Looks like HYPRE_INCLUDE and HYPRE_LIB will be set in
>
> $PETSC_DIR/bmake/${PETSC_ARCH}/petscconf
>
> if it's there. That was as of 2.3.1.
Thanks John... I'll work up a configure test for this.
Derek
-
On Mon, Feb 9, 2009 at 9:35 AM, Derek Gaston wrote:
> Does anyone know how to snoop to see if Petsc is configured with Hypre
> support? I don't see a simple #define HAS_HYPRE or some such anywhere in
> the Petsc source.
Looks like HYPRE_INCLUDE and HYPRE_LIB will be set in
$PETSC_DIR/bmake/${PE
Does anyone know how to snoop to see if Petsc is configured with Hypre
support? I don't see a simple #define HAS_HYPRE or some such anywhere
in the Petsc source.
Thanks,
Derek
On Feb 9, 2009, at 2:42 AM, Tim Kroeger wrote:
> Dear Derek,
>
> On Fri, 6 Feb 2009, Derek Gaston wrote:
>
>> Ok -
On Feb 9, 2009, at 2:46 AM, Tim Kroeger wrote:
> Just to briefly join in the preconditioner discussion: Have you also
> thought about something like a ShellPreconditioner class
> (analogously to the ShellMatrix)? It's not that I would need it
> now, but someone (perhaps me) might need it in
I'll take care of this immediately... thanks for pointing it out. I
looked at the code and thought that that function would be present
regardless of whether or not you configured Petsc with Hypre... guess
I was wrong.
Thanks!
Derek
On Feb 9, 2009, at 2:42 AM, Tim Kroeger wrote:
> Dear De
Dear Derek/Roy,
Just to briefly join in the preconditioner discussion: Have you also
thought about something like a ShellPreconditioner class (analogously
to the ShellMatrix)? It's not that I would need it now, but someone
(perhaps me) might need it in the future, and it would certainly be
ni
Dear Derek,
On Fri, 6 Feb 2009, Derek Gaston wrote:
> Ok - I went ahead and committed that so that I can work on the next
> step...
Unfortunately, it doesn't link for me. I get:
/home/tim/fem-libs/libMesh/libmesh-svn/libmesh/lib/i686-pc-linux-gnu_opt/libmesh.so:
undefined reference to `PCHYPR
On Fri, Feb 6, 2009 at 6:04 PM, Kirk, Benjamin (JSC-EG)
wrote:
> This is definitely one of those times I am glad this list is archived...
>
> The various combinations of potential values for
>
> -pc_hypre_boomeramg_strong_threshold
> -pc_hypre_boomeramg_grid_sweeps_down
> -pc_hypre_boomeramg_grid_
This is definitely one of those times I am glad this list is archived...
The various combinations of potential values for
-pc_hypre_boomeramg_strong_threshold
-pc_hypre_boomeramg_grid_sweeps_down
-pc_hypre_boomeramg_grid_sweeps_up
-pc_hypre_boomeramg_relax_type_all
-pc_hypre_boomeramg_relax_type
On Feb 6, 2009, at 4:29 PM, Roy Stogner wrote:
> Not necessarily separate classes, just the ability to have separate
> classes.
Agreed.
> AMG is still pretty black-box (that being the whole point) and can
> probably just be a type parameter to whatever numerics packages have
> that interface -
On Fri, 6 Feb 2009, Derek Gaston wrote:
> On Feb 6, 2009, at 3:47 PM, Roy Stogner wrote:
>> Supporting old APIs is always good. If you want to stick a deprecated()
>> warning in there now and plan to get rid of it in the future, though,
>> that's fine.
>>
>> In the near term, I'd like to hav
On Feb 6, 2009, at 3:47 PM, Roy Stogner wrote:
> Supporting old APIs is always good. If you want to stick a
> deprecated() warning in there now and plan to get rid of it in the
> future, though, that's fine.
>
> In the near term, I'd like to have a factory method that takes a
> Preconditione
Derek Gaston wrote:
> Ok - I went ahead and committed that so that I can work on the next
> step... which is allowing you to attach a Preconditioner to the linear
> solver. But now we need some discussion on that.
>
> First, do we want to support the old style of specifying a
> precondition
Ok - I went ahead and committed that so that I can work on the next
step... which is allowing you to attach a Preconditioner to the linear
solver. But now we need some discussion on that.
First, do we want to support the old style of specifying a
preconditioner type for a LinearSolver? Or
On Fri, Feb 6, 2009 at 2:45 PM, Derek Gaston wrote:
> So... I've got the initial Preconditioner class and a PetscPreconditioner
> class based on it all coded up and working with my current code. I've
> attached the diff with all of the changes. If no one sees anything big I'll
> go ahead and com
21 matches
Mail list logo
