Thread: [Libmesh-users] How to avoid complex numbers?

SourceForge Headquarters 1320 Columbia Street Suite 310 San Diego, CA 92101 +1 (858) 422-6466

Hi!

Is there a flag which I can add to my compilation process (Make.common) 
to disable complex arithmetic (or another way to achieve that goal)? 
Thanks!

Good morning Edgar,

using ``--enable-complex=no`` should do you that trick.
In the output of `configure --help` is a statement

Optional Features:
   --disable-option-checking  ignore unrecognized --enable/--with options
   --disable-FEATURE       do not include FEATURE (same as 
--enable-FEATURE=no)
   --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]

which is a bit hidden if one seeks for a particular option.

I hope, this help you.

All the best
Hubert

On 8/23/21 5:54 AM, edgar wrote:
> Hi!
>
> Is there a flag which I can add to my compilation process 
> (Make.common) to disable complex arithmetic (or another way to achieve 
> that goal)? Thanks!
>
>
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users

Hi Hubert!

Thanks. It seems that the option which you are sending is for the 
overall libMesh compilation (correct me if I'm wrong). I am interested 
in having my libMesh with the ability to process complex numbers, and 
just compiling a specific program (not the whole libMesh) without 
complex numbers. Do you know if there is a way for that? Thank you 
again.

On 2021-08-23 08:33, Hubert Weissmann wrote:
> Good morning Edgar,
> 
> using ``--enable-complex=no`` should do you that trick.
> In the output of `configure --help` is a statement
> 
> 
> Optional Features:
>   --disable-option-checking  ignore unrecognized --enable/--with 
> options
>   --disable-FEATURE       do not include FEATURE (same as 
> --enable-FEATURE=no)
>   --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
> 
> which is a bit hidden if one seeks for a particular option.
> 
> I hope, this help you.
> 
> All the best
> Hubert

Hi Edgar,

The way I think you'll want to handle this is using libMesh's built-in
numeric types. libMesh::Number will be a complex-valued quantity when
libMesh is compiled with complex numbers and a standard float when not.
libMesh::Real will always be a standard float, regardless of the complex
mode that libMesh is configured with. So the part of your program that you
wish to be real-valued, use libMesh::Real.

HTH,

Paul

On Mon, Aug 23, 2021 at 7:53 AM edgar <edg...@cr...> wrote:

> Hi Hubert!
>
> Thanks. It seems that the option which you are sending is for the
> overall libMesh compilation (correct me if I'm wrong). I am interested
> in having my libMesh with the ability to process complex numbers, and
> just compiling a specific program (not the whole libMesh) without
> complex numbers. Do you know if there is a way for that? Thank you
> again.
>
>
> On 2021-08-23 08:33, Hubert Weissmann wrote:
> > Good morning Edgar,
> >
> > using ``--enable-complex=no`` should do you that trick.
> > In the output of `configure --help` is a statement
> >
> >
> > Optional Features:
> >   --disable-option-checking  ignore unrecognized --enable/--with
> > options
> >   --disable-FEATURE       do not include FEATURE (same as
> > --enable-FEATURE=no)
> >   --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
> >
> > which is a bit hidden if one seeks for a particular option.
> >
> > I hope, this help you.
> >
> > All the best
> > Hubert
>
>
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>

Hi Edgar,

yes you are right. Sorry; than I misinterpreted your question.
Than I agree with Paul below; but AFAIK the `Systems` matrix/rhs (and 
thus solution) allways are of `Number` type!?

At least for the `Linear` systems, I am not aware of a builtin way to 
change this.

Best regards

On 8/23/21 2:58 PM, Paul T. Bauman wrote:
> Hi Edgar,
>
> The way I think you'll want to handle this is using libMesh's built-in
> numeric types. libMesh::Number will be a complex-valued quantity when
> libMesh is compiled with complex numbers and a standard float when not.
> libMesh::Real will always be a standard float, regardless of the complex
> mode that libMesh is configured with. So the part of your program that you
> wish to be real-valued, use libMesh::Real.
>
> HTH,
>
> Paul
>
> On Mon, Aug 23, 2021 at 7:53 AM edgar <edg...@cr...> wrote:
>
>> Hi Hubert!
>>
>> Thanks. It seems that the option which you are sending is for the
>> overall libMesh compilation (correct me if I'm wrong). I am interested
>> in having my libMesh with the ability to process complex numbers, and
>> just compiling a specific program (not the whole libMesh) without
>> complex numbers. Do you know if there is a way for that? Thank you
>> again.
>>
>>
>> On 2021-08-23 08:33, Hubert Weissmann wrote:
>>> Good morning Edgar,
>>>
>>> using ``--enable-complex=no`` should do you that trick.
>>> In the output of `configure --help` is a statement
>>>
>>>
>>> Optional Features:
>>>    --disable-option-checking  ignore unrecognized --enable/--with
>>> options
>>>    --disable-FEATURE       do not include FEATURE (same as
>>> --enable-FEATURE=no)
>>>    --enable-FEATURE[=ARG]  include FEATURE [ARG=yes]
>>>
>>> which is a bit hidden if one seeks for a particular option.
>>>
>>> I hope, this help you.
>>>
>>> All the best
>>> Hubert
>>
>> _______________________________________________
>> Libmesh-users mailing list
>> Lib...@li...
>> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>>
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users
>

Dear Paul and Hubert,

Thank you very much. I had tried by changing libMesh::Number to 
libMesh::Real, and I was getting an error. I thought that it was related 
to PETSc, but now that Hubert mentions about the solvers, I guess that 
he is right. I am getting this:

     error: no matching function for call to 
'libMesh::DirichletBoundary::DirichletBoundary(std::set<short int>&, 
std::vector<unsigned int>&, libMesh::ConstFunction<double>, 
libMesh::VariableIndexing)'
        66 |       libMesh::LOCAL_VARIABLE_ORDER);
     ... --- 8< snip ----
     error: cannot convert 'libMesh::DenseMatrix<double>' to 
'libMesh::DenseMatrix<std::complex<double> >&'
      150 |       
dof_map.heterogenously_constrain_element_matrix_and_vector (Ke, Fe, 
dof_indices);

Thank you both. I will live with my complex numbers :D .

On 2021-08-23 13:02, Hubert Weissmann wrote:
> Hi Edgar,
> 
> yes you are right. Sorry; than I misinterpreted your question.
> Than I agree with Paul below; but AFAIK the `Systems` matrix/rhs (and
> thus solution) allways are of `Number` type!?
> 
> At least for the `Linear` systems, I am not aware of a builtin way to
> change this.
> 
> Best regards
> 
> 
> On 8/23/21 2:58 PM, Paul T. Bauman wrote:
>> Hi Edgar,
>> 
>> The way I think you'll want to handle this is using libMesh's built-in
>> numeric types. libMesh::Number will be a complex-valued quantity when
>> libMesh is compiled with complex numbers and a standard float when 
>> not.
>> libMesh::Real will always be a standard float, regardless of the 
>> complex
>> mode that libMesh is configured with. So the part of your program that 
>> you
>> wish to be real-valued, use libMesh::Real.
>> 
>> HTH,
>> 
>> Paul
> 
> 
> _______________________________________________
> Libmesh-users mailing list
> Lib...@li...
> https://lists.sourceforge.net/lists/listinfo/libmesh-users

On Mon, Aug 23, 2021 at 7:53 AM edgar <edg...@cr...> wrote:

> Hi Hubert!
>
> Thanks. It seems that the option which you are sending is for the
> overall libMesh compilation (correct me if I'm wrong). I am interested
> in having my libMesh with the ability to process complex numbers, and
> just compiling a specific program (not the whole libMesh) without
> complex numbers. Do you know if there is a way for that? Thank you
> again.
>

The decision to use complex numbers is a compile-time one. If your program
links against a libmesh with complex numbers enabled, then the Number type
is #defined to be std::complex<double>. Furthermore certain libmesh
templates, in particular the NumericVector and SparseMatrix classes are
only instantiated for type Number, so these will all use
std::complex<double> internally in such builds.

-- 
John

On 2021-08-23 13:02, John Peterson wrote:
> On Mon, Aug 23, 2021 at 7:53 AM edgar <edg...@cr...> wrote:
> 
>> Hi Hubert!
>> 
>> Thanks. It seems that the option which you are sending is for the
>> overall libMesh compilation (correct me if I'm wrong). I am interested
>> in having my libMesh with the ability to process complex numbers, and
>> just compiling a specific program (not the whole libMesh) without
>> complex numbers. Do you know if there is a way for that? Thank you
>> again.
>> 
> 
> The decision to use complex numbers is a compile-time one. If your 
> program
> links against a libmesh with complex numbers enabled, then the Number 
> type
> is #defined to be std::complex<double>. Furthermore certain libmesh
> templates, in particular the NumericVector and SparseMatrix classes are
> only instantiated for type Number, so these will all use
> std::complex<double> internally in such builds.

Thank you very much, John. Indeed, my suspicion was in the right track 
(see previous e-mail where I answer to Hubert and Paul). If I may add to 
your comment in a joyous way: ...and if you try to modify the 
LIBMESH_USE_COMPLEX_NUMBERS=0, it will die a terrible death :P .

Just to add on this out of curiosity: Due to the missing instantiation 
of NumericVector<Real> etc.
it will be even not possible to setup some custom `System`-type with 
double-variables, I guess?

Best regards
Hubert

On 8/23/21 3:02 PM, John Peterson wrote:
> On Mon, Aug 23, 2021 at 7:53 AM edgar <edg...@cr...> wrote:
>
>> Hi Hubert!
>>
>> Thanks. It seems that the option which you are sending is for the
>> overall libMesh compilation (correct me if I'm wrong). I am interested
>> in having my libMesh with the ability to process complex numbers, and
>> just compiling a specific program (not the whole libMesh) without
>> complex numbers. Do you know if there is a way for that? Thank you
>> again.
>>
> The decision to use complex numbers is a compile-time one. If your program
> links against a libmesh with complex numbers enabled, then the Number type
> is #defined to be std::complex<double>. Furthermore certain libmesh
> templates, in particular the NumericVector and SparseMatrix classes are
> only instantiated for type Number, so these will all use
> std::complex<double> internally in such builds.
>