On Tue, Dec 10, 2013 at 8:58 AM, Lorenzo Zanon
wrote:
> Hi,
>
> After re-installing petsc with Hypre should I recompile libmesh as well?
Definitely.
> Otherwise here are the detailed outputs (JFNK, LU, and -ksp_rtol 1e-4):
>
>
> NL step 0, |residual|_2 = 3.464102e-05
> 0 KSP Residual norm 3
Note that your "JFNK" is actually preconditioned by ILU(0) (and accelerated
by GMRES(30)), which is the default.
These results look more or less correct, but the problem here is small, of
course.
Dmitry.
On Tue, Dec 10, 2013 at 9:58 AM, Lorenzo Zanon
wrote:
> Hi,
>
> After re-installing petsc w
Actually - those look really good. What preconditioner are you using now?
If you're happy with the speed I would stop there...
But - yes, after recompiling PETSc you would want to recompile libMesh...
Derek
On Tue, Dec 10, 2013 at 8:58 AM, Lorenzo Zanon
wrote:
> Hi,
>
> After re-installing p
Hi,
After re-installing petsc with Hypre should I recompile libmesh as well?
Otherwise here are the detailed outputs (JFNK, LU, and -ksp_rtol 1e-4):
> ./example-opt -ksp_monitor -snes_view -snes_linesearch_type basic
> -snes_mf_operator
> Mesh Information:
> mesh_dimension()=3
> spatial_d
Please, include -snes_view so that we can see exactly what solver options
are being used.
Dmitry.
On Mon, Dec 9, 2013 at 12:48 PM, John Peterson wrote:
> > On Mon, Dec 9, 2013 at 11:31 AM, Lorenzo Zanon
> > wrote:
> >
> >> I get faster convergence with e.g. -snes_linesearch_type basic -ksp_rtol
> On Mon, Dec 9, 2013 at 11:31 AM, Lorenzo Zanon
> wrote:
>
>> I get faster convergence with e.g. -snes_linesearch_type basic -ksp_rtol
>> 1e-4:
>>
>> NL step 0, |residual|_2 = 3.464102e-05
>> NL step 1, |residual|_2 = 2.417540e-04
>> NL step 2, |residual|_2 = 6.174706e-08
>> NL step 3,
Good to hear Lorenzo!
How many linear iterations are you taking? (use -ksp_monitor to see)
You might be able to get better performance using a better
preconditioner... I still highly recommend trying Hypre...
Derek
On Mon, Dec 9, 2013 at 11:31 AM, Lorenzo Zanon
wrote:
> Hello,
>
> I'm sorry
Hello,
I'm sorry it took so long.
There were actually a couple of mistakes in my jacobian and residual.
Now for the small problem without specifying any options I get:
> Running ./example-opt
>
> Mesh Information:
> mesh_dimension()=3
> spatial_dimension()=3
> n_nodes()=66
> n_local
On Tue, Nov 26, 2013 at 11:59 AM, Dmitry Karpeyev wrote:
> LU, naturally, should do better, and we should at least see quicker linear
> convergence.
> If not, it's an indication that your problem is singular.
>
LU should only take 1 (ish) linear iteration.
However, I still suspect that the issue
1. It looks like you are using petsc-3.x or older. I would recommend
upgrading to petsc-3.4,
since we don't really maintain earlier versions.
2. Could you run with -snes_view so we know exactly what solvers you are
using?
3. For a small problem running with -pc_type lu, could you post the
nonlinear
Hi,
If you don't like deriving the stuff by hand you might have a
look at AceGen (http://www.fgg.uni-lj.si/Symech/), it uses Mathematica
for the linarization and generates C code for the residual and jacobian.
Robert
On Mon, Nov 25, 2013 at 12:01:20PM -0700, Derek Gaston wrote:
> If it works wit
If it works with JFNK and fails without it... that means your Jacobian is
incorrect. Take another look at your derivatives that are going into your
Jacobian...
Derek
Sent from my iPhone
On Nov 25, 2013, at 11:19 AM, Lorenzo Zanon wrote:
>> Are you sure the Jacobian is correct? Is this a bu
> Are you sure the Jacobian is correct? Is this a buckling problem, or
> something nonsmooth like plasticity or contact? If not, Newton should
> converge fairly fast.
It should be, running on a 10x2x1 mesh I get convergence and a roughly correct
output (wrt FEAP). The problem is just a bar with
On Nov 25, 2013, at 6:23 AM, Lorenzo Zanon
wrote:
> Hello,
>
> Thanks for the detailed answer!
>
> I've just re-run the executable taking off the -snes_mf_operator and LU
> options, but it doesn't look any better... :
>
> Running ./example-opt -snes_type ls -snes_linesearch_type basic -ksp_
Hello,
Thanks for the detailed answer!
I've just re-run the executable taking off the -snes_mf_operator and LU
options, but it doesn't look any better... :
Running ./example-opt -snes_type ls -snes_linesearch_type basic -ksp_rtol 1e-4
NL step 0, |residual|_2 = 1.936492e-05
NL step 1, |res
-snes_mf_operator is a PETSc option specifying that you want to do
preconditioned Jacobian Free Newton Krylov (JFNK). Is the guy using FEAP
doing JFNK?
JFNK means that every linear iteration inside a Newton step you must
recompute your residual. That means a full sweep over the mesh,
re-evaluati
Hello,
I have an example based on miscellaneous_ex3.C, where I implemented a nonlinear
elastic problem with StVenant stress-strain law on a 3D domain 5x1x1, blocked
at x=0 and with an applied load at x=5. The problem is, the computation of the
displacement (along x y and z) is really slow (hou
Thanks for the many answers!
Still, according to the PETSc manual "matrix-free methods cannot be used if a
direct solver is to be employed", I think this means I cannot use
-snes_mf_operator and "-ksp_type preonly -pc_type lu
-pc_factor_nonzeros_along_diagonal" at the same time. Do I need some
es diffuse to uniformity.
> Systems tend to subside to very simple states,independent of their specific
> history."
> > Herbert Callen . Thermodynamics and an Introduction to Thermostatics.
> >
> > - Original Message -----
> > From: "Lorenzo Zano
A couple of things:
1. -pre is just an option for that particular example. It was an attempt
to provide a solver agnostic way of turning on preconditioning. If you
don't specify -pre or a petsc option to turn on preconditioning then you
are doing zero preconditioning... which is bound to fail.
For nonlinear problems the linesearch may very well fail if the linear
update is a bad descent direction
(usually indicating a nonconvexity of the corresponding energy functional,
if it exists; is this a p-Laplacian problem? To the best of my knowledge
it is convex for p > 1, albeit the PDE is not
On Thu, Mar 7, 2013 at 9:31 AM, Subramanya Gautam Sadasiva <
ssada...@purdue.edu> wrote:
> Yes, that is right... my bad..
>
Nevertheless, I think you are correct that there is *something* wrong with
his Jacobian. Maybe it's "good enough" to work with non-direct solvers but
fails on LU?
--
Joh
t;
Herbert Callen . Thermodynamics and an Introduction to Thermostatics.
- Original Message -
From: "John Peterson"
To: "Subramanya Gautam Sadasiva"
Cc: "Lorenzo Zanon" , "libmesh-users"
, "Derek Gaston"
Sent: Thursday, March 7, 2013 11:22:4
On Thu, Mar 7, 2013 at 9:20 AM, Subramanya Gautam Sadasiva <
ssada...@purdue.edu> wrote:
> Hi Lorenzo ,
> Again, this just suggests to me that lu is not appropriate to your system.
> You might have zeros on your diagonals and the petsc lu does not pivot.
>
If this was the case, wouldn't he see an
t;Lorenzo Zanon"
To: "Subramanya Gautam Sadasiva"
Cc: "libmesh-users" , "Derek Gaston"
Sent: Thursday, March 7, 2013 10:06:12 AM
Subject: Re: [Libmesh-users] Newton iteration
Hi,
My problem converges using:
-pre
-snes_type ls
-snes_type basic
but I stil
uot;Lorenzo Zanon"
> To: "Derek Gaston"
> Cc: "Subramanya Gautam Sadasiva" , "libmesh-users"
>
> Sent: Wednesday, March 6, 2013 1:26:21 PM
> Subject: Re: [Libmesh-users] Newton iteration
>
> I get convergence only using the precondit
history."
Herbert Callen . Thermodynamics and an Introduction to Thermostatics.
- Original Message -
From: "Lorenzo Zanon"
To: "Derek Gaston"
Cc: "Subramanya Gautam Sadasiva" , "libmesh-users"
Sent: Wednesday, March 6, 2013 1:26:21 PM
Subject:
Message -
> From: "Lorenzo Zanon"
> To: libmesh-users@lists.sourceforge.net
> Sent: Wednesday, March 6, 2013 6:53:02 AM
> Subject: [Libmesh-users] Newton iteration
>
> Hello,
>
> I'm modifying miscellaneous_ex3.C (2D Laplace-Young Problem Using Nonli
and an Introduction to Thermostatics.
>
> - Original Message -
> From: "Lorenzo Zanon"
> To: libmesh-users@lists.sourceforge.net
> Sent: Wednesday, March 6, 2013 6:53:02 AM
> Subject: [Libmesh-users] Newton iteration
>
> Hello,
>
> I'm modifying
o Thermostatics.
- Original Message -
From: "Lorenzo Zanon"
To: libmesh-users@lists.sourceforge.net
Sent: Wednesday, March 6, 2013 6:53:02 AM
Subject: [Libmesh-users] Newton iteration
Hello,
I'm modifying miscellaneous_ex3.C (2D Laplace-Young Problem Using Nonlinear
Solvers,
Hello,
I'm modifying miscellaneous_ex3.C (2D Laplace-Young Problem Using Nonlinear
Solvers, in my current libmesh version). After adapting residual and jacobian
to my own problem, although the tolerance is not met, the Newton solver stops
after one iteration with an all-zero (incorrect) solutio
31 matches
Mail list logo
