Re: [deal.II] Re: dealii compilation on Cray machine

2019-04-06 Thread Matthias Maier 
Cross compilation should work - I have briefly tested it back in
2011 and have written a very minimalistic paragraph about what you have
to do:

  https://www.dealii.org/9.0.0/developers/porting.html

It would be great if someone could verify that it still works :-D

Best,
Matthias



On Fri, Apr  5, 2019, at 14:45 CDT, Bruno Turcksin  
wrote:

> Le ven. 5 avr. 2019 à 14:48, Phani Motamarri  a écrit :
>>
>> No login node is intel-broadwell and compute node is
>> intel-skylake. Problem is that the supercomputer I am compiling on
>> has no way to access compute node (interactive job) directly for
>> compiling. I am not sure if dealii allows for cross compilation.
>
> I've never tried that. Maybe Martin has experience with that?
>
> 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 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: dealii compilation on Cray machine

2019-04-06 Thread Martin Kronbichler 



On 05.04.19 21:45, Bruno Turcksin wrote:

Le ven. 5 avr. 2019 à 14:48, Phani Motamarri  a écrit :

No login node is intel-broadwell and compute node is intel-skylake. Problem is 
that the supercomputer I am compiling on has no way to access compute node 
(interactive job) directly for compiling. I am not sure if dealii allows for 
cross compilation.

I've never tried that. Maybe Martin has experience with that?


Unfortunately nothing that would help - I always managed to run at least 
the configuration (cmake) on the same architecture whereas I indeed 
built (make) on a different architecture once the compilation was 
finished through and with a fixed '-march=knl' for example. But we 
should definitely support cross-compilation. I'm not sure how our 
support was - I will test this before the release.


As to the present case: Phani, have you tried to submit the compilation 
job, i.e., "cmake -D  && make -j xx" through a job file? That should 
invoke the right microarchitecture; some machines do not have all 
compile tools such as cmake available as modules on the nodes, though.


Best,
Martin

--
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] project_boundary_values for components

2019-04-06 Thread Anton 
Thank you Wolfgang.

Basically if I interpolate a step function (which can be exactly presented 
in for example DGQ1 space if the discontinuity is at the node) I get a 
1-element wide transition.  In ASCII pictures, I expect  ___  but I get 
___/---  instead :)

So if I understand you correctly, one would need to (1) obtain 
ConstraintMatrix from project_boundary_values; (2) go over each row of the 
constraint matrix and check which component it is; (3) if it is the "right" 
component  then copy the row.  Could I ask how does one figure out the 
component number from the row number in the ConstraintMatrix - I cannot 
immediately figure this out from DoFHandler interface?

/Anton

On Saturday, April 6, 2019 at 5:11:09 PM UTC+2, Wolfgang Bangerth wrote:
>
> On 4/5/19 11:25 AM, Anton wrote: 
> > I need to apply a boundary condition (discontinuous function) to a few 
> > components of a dof_handler only.  I am currently using 
> > 
> > | 
> > voidVectorTools::interpolate_boundary_values 
> > <
> https://dealii.org/current/doxygen/deal.II/group__constraints.html#ga2376b9a282a2b62be5c55ab62a11a132>(constDoFHandlerType  
>
> >  >,consttypes::boundary_id 
> > <
> https://dealii.org/current/doxygen/deal.II/namespacetypes.html#aaf4eb6ec214fa642dfd956f11a9cd2d7>boundary_component,constFunction
>  
>
> >  >  
> >_function,ConstraintMatrix 
> > ,constComponentMask
> >  
>
> > _mask=ComponentMask
> >  
>
> > ()) 
> > | 
> > 
> > for this purpose, but because of the discontinuity I get unwanted 
> effects near 
> > the discontinuity - that is even when the discontinuity is aligned with 
> the 
> > mesh, and even though I am using discontinuous elements (FE_FaceQ), I 
> get 
> > unwanted effects because the interpolation point falls exactly between 
> the 
> > elements.  Is there a way of applying project_boundary_values instead to 
> fix 
> > this issue?  Basically I cannot seem to find support for component 
> selection 
> > in connection with project_boundary_values. 
>
> Can you explain what these unwanted effects are, for example in a picture? 
>
> You're right that the project_b_v() function (currently) doesn't support 
> selecting individual components. But you can just project onto all 
> components 
> into a separate object, and then you only transfer those elements that 
> correspond to the components you care about. 
>
> Best 
>   Wolfgang 
>
> -- 
>  
> Wolfgang Bangerth  email: bang...@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] project_boundary_values for components

2019-04-06 Thread Wolfgang Bangerth 
On 4/5/19 11:25 AM, Anton wrote:
> I need to apply a boundary condition (discontinuous function) to a few 
> components of a dof_handler only.  I am currently using
> 
> |
> voidVectorTools::interpolate_boundary_values 
> (constDoFHandlerType  
>  >,consttypes::boundary_id 
> boundary_component,constFunction
>  
>   >_function,ConstraintMatrix 
> ,constComponentMask
>  
> _mask=ComponentMask
>  
> ())
> |
> 
> for this purpose, but because of the discontinuity I get unwanted effects 
> near 
> the discontinuity - that is even when the discontinuity is aligned with the 
> mesh, and even though I am using discontinuous elements (FE_FaceQ), I get 
> unwanted effects because the interpolation point falls exactly between the 
> elements.  Is there a way of applying project_boundary_values instead to fix 
> this issue?  Basically I cannot seem to find support for component selection 
> in connection with project_boundary_values.

Can you explain what these unwanted effects are, for example in a picture?

You're right that the project_b_v() function (currently) doesn't support 
selecting individual components. But you can just project onto all components 
into a separate object, and then you only transfer those elements that 
correspond to the components you care about.

Best
  Wolfgang

-- 

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.