Hi,
In the example (FEVector- Uncoupled Poisson Problem). I see that the relevant
part is as follows,
for (unsigned int i=0; i<phi.size(); i++)
Fe(i) += JxW[qp]*f*phi[i][qp];
Now, phi[i][qp] and f here are vectors and the multiplication seems to be a
term by term multiplication. What worries me is the addition. Does doing this
automatically slot in the other components in the right places in Fe?
Thanks,
Subramanya Sadasiva
PS. The reason I don't want to manually average things is that I can foresee
things getting messy in parallel where I'd have to explicitly move data. On the
other hand, using implicit system will do the stuff automatically for me, and a
mass matrix should not be that hard to solve- especially if I am able to lump
it.
"But memory eventually fades. Turbulences damp out, internal strains yield to
plastic flow, concentration inhomogeneities 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: "Benjamin Kirk (JSC-EG311)" <[email protected]>
To: "Subramanya Gautam Sadasiva" <[email protected]>
Cc: "John Peterson" <[email protected]>, "libmesh-users"
<[email protected]>
Sent: Tuesday, March 12, 2013 9:59:14 AM
Subject: Re: [Libmesh-users] FEVectorBase element base.
The most robust way to do this, such that it doesn't depend on libMesh
implementation details, is to declare a system with Ndim components stored at
the nodes, and then access the rhs/solution components through their dof
indices as you usually would.
If you want to manually compute the average rather than form a linear system
you can simply create an ExplicitSystem.
-Ben
On Mar 11, 2013, at 9:52 PM, "Subramanya Gautam Sadasiva" <[email protected]>
wrote:
> I am trying to create a method to average gradient values at the quadrature
> points to the nodes.
> To do this, I create a FEVector System, and you essentially have something
> like
> [N^T N][gradient_at_nodes] = [gradient_at_qp]
> While constructing the gradient_at_qp vector i am unable to figure out how to
> add them to the element rhs.
> For the default version, if n_dofs is the number of nodes, my loop should be
> doing
> for (i=0; i<n_dofs; i++)
> {
> for (j=0; j<_dim; j++)
> Fe(n_dofs*i+j) = terms(j)
> }
>
> Does the above sound okay?
>
> Thanks,
>
>
>
> Subramanya Sadasiva
>
> "But memory eventually fades. Turbulences damp out, internal strains yield to
> plastic flow, concentration inhomogeneities 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: "John Peterson" <[email protected]>
> To: "Subramanya Gautam Sadasiva" <[email protected]>
> Cc: "libmesh-users" <[email protected]>
> Sent: Monday, March 11, 2013 11:35:04 PM
> Subject: Re: [Libmesh-users] FEVectorBase element base.
>
>
>
>
>
>
>
> On Mon, Mar 11, 2013 at 9:30 PM, Subramanya Gautam Sadasiva <
> [email protected] > wrote:
>
>
> Hi,
> How are the components arranged in the solution vector for each element? are
> the degrees of freedom arranged as (for a linear quadrilateral element)
> [1 1 1 1 2 2 2 2 3 3 3 3] or is it [1 2 3 1 2 3 1 2 3 1 2 3]. I am unable to
> figure out the code in fe.C .
>
>
>
> I think by the first you mean "var major" ordering (?) which is the default
> in libmesh. The second can be enabled by passing -- node - major -dofs on the
> command line.
>
>
> --
> John
>
> ------------------------------------------------------------------------------
> Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester
> Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the
> endpoint security space. For insight on selecting the right partner to
> tackle endpoint security challenges, access the full report.
> http://p.sf.net/sfu/symantec-dev2dev
> _______________________________________________
> Libmesh-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
------------------------------------------------------------------------------
Symantec Endpoint Protection 12 positioned as A LEADER in The Forrester
Wave(TM): Endpoint Security, Q1 2013 and "remains a good choice" in the
endpoint security space. For insight on selecting the right partner to
tackle endpoint security challenges, access the full report.
http://p.sf.net/sfu/symantec-dev2dev
_______________________________________________
Libmesh-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/libmesh-users
