Re: [deal.II] Re: New step tutorial elastic wave equation with PMLs, complex-valued algebra

2018-04-30 Thread Daniel Garcia 


On Monday, April 30, 2018 at 7:13:26 AM UTC+2, Denis Davydov wrote:
>
>
>
> > On 30 Apr 2018, at 04:54, Wolfgang Bangerth  > wrote: 
> > 
> > On 04/28/2018 10:36 PM, Denis Davydov wrote: 
> >> As far as as recall there just a few things left related to complex 
> numbers. 
> >> The biggest one is complex-valued Dirichlet constraints in 
> ConstraintMatrix object 
> >> https://github.com/dealii/dealii/issues/1760 
> >> That's not possible yet. Otherwise all the solvers and linear algebra 
> is ready. So you can certainly assemble system and solve it. 
> > 
> > Denis -- out of curiosity, what solver are you using for complex-valued 
> linear systems? I don't think the deal.II SolverGMRES class works, and 
> obviously the other iterative solvers don't either. Nor does UMFPACK. 
>
> I would not be so sure about it, but I have not tried either. 
> I am quite certain complex-valued PETSc will work. 
>
> ps. I don’t use any of them, some time ago I invested a lot of efforts to 
> push forward 
> complex-valued algebra in deal.II (which was also followed by others), but 
> strangely enough I never used it. 
>
>
>  
I will write the first version of the code with real/imag separation and 
UMFPACK. This works well with PML matrices.

Then I will modify the code and use complex-valued algebra with PETSc.

This is an extra-step, but it will allow to compare the results. 

Best,
Daniel

-- 
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: New step tutorial elastic wave equation with PMLs, complex-valued algebra

2018-04-29 Thread Denis Davydov 


> On 30 Apr 2018, at 04:54, Wolfgang Bangerth  wrote:
> 
> On 04/28/2018 10:36 PM, Denis Davydov wrote:
>> As far as as recall there just a few things left related to complex numbers.
>> The biggest one is complex-valued Dirichlet constraints in ConstraintMatrix 
>> object
>> https://github.com/dealii/dealii/issues/1760
>> That's not possible yet. Otherwise all the solvers and linear algebra is 
>> ready. So you can certainly assemble system and solve it.
> 
> Denis -- out of curiosity, what solver are you using for complex-valued 
> linear systems? I don't think the deal.II SolverGMRES class works, and 
> obviously the other iterative solvers don't either. Nor does UMFPACK.

I would not be so sure about it, but I have not tried either.
I am quite certain complex-valued PETSc will work.

ps. I don’t use any of them, some time ago I invested a lot of efforts to push 
forward
complex-valued algebra in deal.II (which was also followed by others), but 
strangely enough I never used it. 

> 
> 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 a topic in the 
> Google Groups "deal.II User Group" group.
> To unsubscribe from this topic, visit 
> https://groups.google.com/d/topic/dealii/N-V-8geRLrA/unsubscribe.
> To unsubscribe from this group and all its topics, 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.

[deal.II] Re: New step tutorial elastic wave equation with PMLs, complex-valued algebra

2018-04-29 Thread Daniel Garcia 
Thanks Denis,

I will try to assemble and solve the system using complex-valued algebra.

Best,
Daniel

On Sunday, April 29, 2018 at 6:36:03 AM UTC+2, Denis Davydov wrote:
>
> Hi Daniel,
>
> Thanks for willing to contribute the tutorial.
>
> As far as as recall there just a few things left related to complex 
> numbers. 
> The biggest one is complex-valued Dirichlet constraints in 
> ConstraintMatrix object
> https://github.com/dealii/dealii/issues/1760 
> That's not possible yet. Otherwise all the solvers and linear algebra is 
> ready. So you can certainly assemble system and solve it.
>
> Denis.
>
> On Saturday, April 28, 2018 at 11:49:40 AM UTC+2, Daniel Garcia wrote:
>>
>> Hi,
>>
>> I'm a researcher at the CNRS (Paris). My nickname in GitHub is dangars (I 
>> contributed with a couple of pull requests)
>>
>> I work in optomechanics/phononics. I've been using deal.ii for 6 months 
>> and it works great! I developed a model of the elastic wave equation in 2D 
>> that uses PMLs.
>>
>> I'm preparing a contribution of a new step tutorial of the elastic wave 
>> equation with PMLs in 3D to GitHub/dealii. This will be useful for the 
>> phononics and optomechanics communities.
>>
>> The separation real/imag works very well with PMLs.
>>
>> The elastic/PML matrix with real/imag separation is big. If I use complex 
>> numbers the matrix will be compact and will fit in one line (much better 
>> for a step tutorial)
>>
>> What is the status of complex valued calculations? Is it ready to solve a 
>> differential equations? Is there a performance penalty with complex numbers 
>> vs separate real/imag parts?
>>
>> According to this, it seems to be almost ready:
>> https://github.com/dealii/dealii/projects/1
>>
>> Thanks,
>> Daniel Garcia
>>
>

-- 
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.

[deal.II] Re: New step tutorial elastic wave equation with PMLs, complex-valued algebra

2018-04-28 Thread Denis Davydov 
Hi Daniel,

Thanks for willing to contribute the tutorial.

As far as as recall there just a few things left related to complex 
numbers. 
The biggest one is complex-valued Dirichlet constraints in ConstraintMatrix 
object
https://github.com/dealii/dealii/issues/1760 
That's not possible yet. Otherwise all the solvers and linear algebra is 
ready. So you can certainly assemble system and solve it.

Denis.

On Saturday, April 28, 2018 at 11:49:40 AM UTC+2, Daniel Garcia wrote:
>
> Hi,
>
> I'm a researcher at the CNRS (Paris). My nickname in GitHub is dangars (I 
> contributed with a couple of pull requests)
>
> I work in optomechanics/phononics. I've been using deal.ii for 6 months 
> and it works great! I developed a model of the elastic wave equation in 2D 
> that uses PMLs.
>
> I'm preparing a contribution of a new step tutorial of the elastic wave 
> equation with PMLs in 3D to GitHub/dealii. This will be useful for the 
> phononics and optomechanics communities.
>
> The separation real/imag works very well with PMLs.
>
> The elastic/PML matrix with real/imag separation is big. If I use complex 
> numbers the matrix will be compact and will fit in one line (much better 
> for a step tutorial)
>
> What is the status of complex valued calculations? Is it ready to solve a 
> differential equations? Is there a performance penalty with complex numbers 
> vs separate real/imag parts?
>
> According to this, it seems to be almost ready:
> https://github.com/dealii/dealii/projects/1
>
> Thanks,
> Daniel Garcia
>

-- 
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.