Re: [deal.II] Re: How to get the coordinates of a given degree of freedom on an edge?

2019-02-19 Thread Phạm Ngọc Kiên 
Yes.
Let me go from my basic idea when setting up my source term.
First, I create a computational domain such that a source can be placed
exactly on an edge. It is always possible to do with unstructured meshes.
Then I go through each element and check if its edge overlap the source. If
it does then I will assign a source moment to the edge.
My problem now is to find out if any edge of an element is overlapped by
the my source.

I thought about finding the position of the degree of freedom in a cell (in
real co-ordinate) and check if its real position is my source position.
Thus, I sent a question about this.

I have checked my codes and saw that \varphi_i(x) were a vector at
quadrature point x for the i_th degree of freedom.
It determines the direction of my vector field on the i_th degree of
freedom in reference cell.
So, I set Js(x) is a vector of the same size to compute the right hand side.

I think there is no problem for the FE_NedelecSZ element. My code still run
but it do not give me an exact solution compared with analytic one because
it is not true if I set the same Js(x) for every dof in a cell containing
the source point.


I also tried to use FE_NedelecSZ(order) with different order.
And I saw that no matter how I change the order value, the functions
fe.n_dofs_per_quad() and fe.n_dofs_per_hex() returned 0. The number of dofs
was unchanged and was equal to fe.n_dofs_per_line().
Does this mean that there is no face and cell dof?
Do I need to use hp::DoFHandler to get the higher order dofs for
calculation?

Thank you very much.
Best,
Kien

Vào Th 4, 20 thg 2, 2019 vào lúc 07:44 Wolfgang Bangerth <
bange...@colostate.edu> đã viết:

> On 2/19/19 12:03 AM, Phạm Ngọc Kiên wrote:
> > I currently try to solve the electromagnetics problem, followed by the
> > research of Ross Kynch.
> > My goal is to compute the right hand side for the problem:
> > curl curl E + i*omega*mu*sigma*E = - Js
> > I only want to set the Js term here to be none zero, for example to be
> > 1, at a given degree of freedom (i.e. at source position), and  0
> > everywhere else.
> > Thus, I need to write the codes to query if a degree of freedom is at
> > source position.
> > As you mentioned, now I think that I need to check if the source point
> > is on an edge when assembling the right hand side vector.
> > Could you please tell me help me about this?
>
> The term you are trying to add (Js) is a right hand side term to the
> equation. So you need to compute the vector
>
>F_i  =  - \int_Omega  \varphi_i(x)  Js(x)  dx
>
> for which you only need to be able to evaluate Js(x) and \varphi_i(x) at
> quadrature points x, but not a specific support points. This is no
> problem for the FE_NedelecSZ element. Can you explain why this does not
> work for you?
>
> 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.
>

-- 
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.

Re: [deal.II] Re: How to get the coordinates of a given degree of freedom on an edge?

2019-02-19 Thread Wolfgang Bangerth 
On 2/19/19 12:03 AM, Phạm Ngọc Kiên wrote:
> I currently try to solve the electromagnetics problem, followed by the 
> research of Ross Kynch.
> My goal is to compute the right hand side for the problem:
> curl curl E + i*omega*mu*sigma*E = - Js
> I only want to set the Js term here to be none zero, for example to be 
> 1, at a given degree of freedom (i.e. at source position), and  0 
> everywhere else.
> Thus, I need to write the codes to query if a degree of freedom is at 
> source position.
> As you mentioned, now I think that I need to check if the source point 
> is on an edge when assembling the right hand side vector.
> Could you please tell me help me about this?

The term you are trying to add (Js) is a right hand side term to the 
equation. So you need to compute the vector

   F_i  =  - \int_Omega  \varphi_i(x)  Js(x)  dx

for which you only need to be able to evaluate Js(x) and \varphi_i(x) at 
quadrature points x, but not a specific support points. This is no 
problem for the FE_NedelecSZ element. Can you explain why this does not 
work for you?

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.

Re: [deal.II] Re: intel 19 compilation failure

2019-02-19 Thread Sean McGovern 
Hi, 

I returned to this today and using the installed PETSc module (which links 
against mkl_blacs_intelmpi_lp64 [not ilp]), setting the paths for the 
BLAS_DIR and LAPACK_DIR CMake variables in dealii.package, allowed for a 
complete installation and passing the quick_tests.
Thanks Bruno ( and also at FZJ to Lucas and Marc for their advice), 
Best,
Sean

On Thursday, February 14, 2019 at 5:04:17 PM UTC+1, Sean McGovern wrote:
>
> OK. I will look into it. 
> Thanks, Bruno!
>
> On Thursday, February 14, 2019 at 4:49:35 PM UTC+1, Bruno Turcksin wrote:
>>
>> Sean,
>>
>> It looks like you need to link against blacs (mkl_blacs_intelmpi_ilp64) 
>> see here 
>> 
>>  
>> but I don't know where you need to add the flag in CANDI :(
>>
>> Best,
>>
>> bruno
>>
>> Le jeu. 14 févr. 2019 à 10:35, Sean McGovern  a écrit :
>>
>>> Hi Bruno!
>>> Attached is the configuration log. 
>>>
>>> I use MKL for petsc and p4est. 
>>>
>>> Thanks,
>>> Sean
>>>
>>> On Thursday, February 14, 2019 at 4:22:39 PM UTC+1, Bruno Turcksin wrote:

 Sean,

 On Thursday, February 14, 2019 at 9:48:56 AM UTC-5, Sean McGovern wrote:
>
> I am trying to install dealii with petsc and p4est on the cluster in 
> Juelich with intel 19. It has been running with intel 18. As others have 
> suggested, I have turned DEAL_II_WITH_CXX17=OFF, I have used the flag 
> "-xhost", but I run into errors relating to vectorization (ALTIVEC).  I 
> have used the bundled tbb, and using the loaded tbb module,  I get the 
> errors in the log attached. 
>
 This error is normal, we check for ALTIVEC which you don't have on your 
 machine (ALTIVEC is only for IBM machine) and since we can't find 
 altivec.h, we disable the altivec vectorization. What bothers me more is 
 that it seems that you have a problem with MKL:

 /gpfs/software/juwels/stages/2018b/software/imkl/2019.0.117-ipsmpi-2018b/mkl/lib/intel64/libmkl_core.a(mkl_get_mpi_wrappers_static.o):mkl_get_mpi_wrappers.c:function
  mkl_serv_get_mpi_wrappers: error: undefined reference to 
 'MKLMPI_Get_wrappers'


 What's the configuration you are trying to use?

 Best,

 Bruno

>>> -- 
>>> 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 a topic in the 
>>> Google Groups "deal.II User Group" group.
>>> To unsubscribe from this topic, visit 
>>> https://groups.google.com/d/topic/dealii/plJ81xhINQ0/unsubscribe.
>>> To unsubscribe from this group and all its topics, send an email to 
>>> dealii+un...@googlegroups.com.
>>> For more options, visit https://groups.google.com/d/optout.
>>>
>>

-- 
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.