[deal.II] PETSc and dealII clashes when compiling without MPI

I installed PETSc using this configuration line: Configure Options: --configModules=PETSc.Configure --optionsModule=config.compilerOptions PETSC_ARCH=linux-gnu-complex --with-scalar-type=complex --with-cc=gcc --with -cxx=g++ --with-fc=gfortran --with-mkl_pardiso=1 --with-blas-lapack-dir= /opt/int

[deal.II] Re: PETSc and dealII clashes when compiling without MPI

I intended to compile everything first without MPI (maybe easier?), but now I solved it by using an MPI compiler. Furthermore, my setup of PETSc was faulty, thus I had to reconfigure and -compile it in order to fix all the problems (which I could not see from the error messages during program co

[deal.II] Example for three coupled equations in dealII?

I have the following three PDEs (reduced from a more complex version): Now I am stuck. Usually I convert them to a discretized form (as for exam

[deal.II] Re: Example for three coupled equations in dealII?

Yes, that is a perfect example, thanks! Am Freitag, 7. Juli 2017 14:12:39 UTC+2 schrieb Bruno Turcksin: > > Hey, > > On Friday, July 7, 2017 at 5:03:05 AM UTC-4, Maxi Miller wrote: >> >> I have the following three PDEs (reduced from a more complex version): >> >> >>

[deal.II] Combination of Trilinos and deal.II fails at compilation

When compiling deal.II without Trilinos, the compilation works. When including Trilinos, I get a lot of errors, beginning with: In file included from /usr/local/include/ml_epetra_utils.h:42:0, > from /usr/local/include/ml_MultiLevelPreconditioner.h:86, > from >

[deal.II] Re: Combination of Trilinos and deal.II fails at compilation

Not tested yet. The compilation of both libraries separated works fine with the Intel MPI toolkit, but the combination does not. Should I test that (i.e. recompile all libraries using OpenMPI), or are there other possible fixes? Am Samstag, 8. Juli 2017 22:33:24 UTC+2 schrieb Bruno Turcksin: >

Re: [deal.II] Combination of Trilinos and deal.II fails at compilation

Bug also confirmed with OpenMPI. Is there a way to "uninstall" the conflicting files, such that I can install them in their own paths? Am Sonntag, 9. Juli 2017 18:19:03 UTC+2 schrieb Wolfgang Bangerth: > > On 07/08/2017 11:55 AM, 'Maxi Miller' via deal.II User Group

Re: [deal.II] Combination of Trilinos and deal.II fails at compilation

Installed it via make install . Now I get the error message: [ 61%] Building CXX object source/numerics/CMakeFiles/ obj_numerics_debug.dir/matrix_creator_inst3.cc.o In file included from /usr/local/include/Teuchos_Assert.hpp:46:0, from /usr/local/include/Teuchos_any.hpp:51,

Re: [deal.II] Combination of Trilinos and deal.II fails at compilation

Neither did Trilinos provide such a method, nor did the fix help, instead it invoked multiple errors in addition. Is there a temporary fix I can apply to continue work while this problem is fixed by the Trilinos-team? Am Sonntag, 9. Juli 2017 21:59:16 UTC+2 schrieb Wolfgang Bangerth: > > On 07/0

[deal.II] Is there a list of requirements for the different examples?

I was wondering why cmake does not generate anything for some examples, such as (in my case) step-17. My generator output is ### # # deal.II configuration: #CMAKE_BUILD_TYPE: DebugRelease #BUILD_SHARED_LIBS: ON #CMAKE_INSTALL_PREFIX: /usr/local #CMAKE_

[deal.II] Re: Is there a list of requirements for the different examples?

That is interesting. I got exactly that output (did not know that I can do that, btw.), but my PETSC-installation is configured with complex numbers. How can I then tell dealII during the configuration process that that is the case? Am Dienstag, 11. Juli 2017 14:10:47 UTC+2 schrieb Jean-Paul Pe

[deal.II] Re: Is there a list of requirements for the different examples?

> > Yes, but my problem is that I built PETSc using > --with-scalar-type=complex, and still dealII thinks that it was built > without complex support (according to the flag). > Am Dienstag, 11. Juli 2017 15:19:37 UTC+2 schrieb Jean-Paul Pelteret: > > Hi Maxi, > > To the best of my knowledge, d

[deal.II] Re: Is there a list of requirements for the different examples?

Yes, that explains it, thanks! Am Dienstag, 11. Juli 2017 15:33:18 UTC+2 schrieb Jean-Paul Pelteret: > > No, I think that you've misunderstood. The message is telling you what the > prerequisites are, i.e. that you must have PETSc enabled and PETSc cannot > have complex number support enabled. T

[deal.II] Best approach for multi-variable system with multiple grids

I have a coupled system of two equations:

Re: [deal.II] Best approach for multi-variable system with multiple grids

;t it? Am Dienstag, 11. Juli 2017 22:22:39 UTC+2 schrieb Wolfgang Bangerth: > > On 07/11/2017 08:29 AM, 'Maxi Miller' via deal.II User Group wrote: > > Which of those two approaches is more beneficial for me (especcially > after I > > would like to extend it afterward

[deal.II] Modification of example 33 correct, or not?

Having the following set of functions I would like to modify example 33 so that I can calculate them. All variables (N, TE and TL) are depending on x, y, (z)

[deal.II] Re: Modification of example 33 correct, or not?

N, T_E and T_L are a matrix, either 2d or 3d, which can be represented as points on a grid. k_N, k_E and k_L are depending on the value of (respective) N, T_E and T_L, i.e. k_{N, 11}=k_N(N_{11}). I am not sure if that qualifies the k-values as matrix/vector values, or as scalar. Am Samstag, 15.

[deal.II] Re: openmp flag in CMake?

Can you try CMake-GUI? At least there you can find all occurences of openmp and enable it. Am Sonntag, 16. Juli 2017 19:28:27 UTC+2 schrieb Kartik Jujare: > > Hello, > > Would be grateful if anyone could correct or direct me to the right > solution. > > As a test I have a following for loop is a

Re: [deal.II] Re: Modification of example 33 correct, or not?

The full origin

[deal.II] Time derivative on right hand side for equation

I have an equation system where the right hand side contains a time derivative, looking like Without that time derivative I can solve it, but I do not know

Re: [deal.II] Time derivative on right hand side for equation

So, you suggest to calculate first the first equation, and then use the result for the second equation? Am Dienstag, 18. Juli 2017 16:54:02 UTC+2 schrieb Wolfgang Bangerth: > > On 07/18/2017 07:21 AM, 'Maxi Miller' via deal.II User Group wrote: > > I have an equation syste

Re: [deal.II] Time derivative on right hand side for equation

That would result in (including my starting equations): But how do I then get the value for A after each iteration? Thank you! Am Dienstag, 18. Juli 2017

[deal.II] Modification of right hand side of example 52 for multiple equations leads to linker error

I tried to modify the right side of example 52 from one equation in parallel to three equations in parallel, each with own system_matrix, mass_matrix and source term. My code for that is template Table<1, Vector> Diffusion::evaluate_diffusion_full (const double time, const Table<1, Ve

[deal.II] Status of support for complex equations?

Hei, I was wondering if dealII already natively supports equations containing imaginary values, such as or if I still have to split them up, such as in exa

[deal.II] Usage of `extract_subvector_to`

I wanted to test how to use the function extract_subvector_to in order to slice a part of a vector and put the result into another vector. Using a short test program Vector vec_1 = Vector(9); Vector vec_2 = Vector(vec_1.size()/3); for(size_t i = 0; i < vec_1.size(); ++i) vec_

[deal.II] Expanding example 52 to multiple equations

I still have some problems trying to understand when to use multiple DoFHandler-variables/FE_Q-variables, and when to use only one. I am trying to expand example 52 to multiple coupled equations. My initial approach is to expand the functions from (here assemble_system) FEValues<2> fe_values(fe,

[deal.II] Re: Usage of `extract_subvector_to`

Slightly confusing (I would have expected something else), but that explains it. Thanks! Am Donnerstag, 20. Juli 2017 13:23:46 UTC+2 schrieb Jean-Paul Pelteret: > > Dear Maxi, > > The documentation for the alternate version of the function >

Re: [deal.II] Re: Usage of `extract_subvector_to`

I do not know how to formulate it, but I can say what I expected: Using std::vector I can write the vector-extraction in the following form: //Using std::vector std::vector std_vec_1(9); std::vector std_vec_2(std_vec_1.size()/3); for(size_t i = 0; i < std_vec_1.size(); ++i) std_

Re: [deal.II] Re: Usage of `extract_subvector_to`

Yes, that makes sense! I did not see that you also can extract non-continuous values using that function. Thank you for the explanation! Am Freitag, 21. Juli 2017 20:57:35 UTC+2 schrieb Wolfgang Bangerth: > > On 07/21/2017 01:25 AM, Maxi Miller wrote: > > > > i.e. the begin()-iterator of the de

[deal.II] Need help setting up the left hand side in a modification of example 52

I am following example 52 for writing code which should solve a system of three equations. In it I can rewrite my equations such that my source term is zero, but my matrices on the left side are depending on time, which means that I have to reassemble the system matrix and the inverse system ma

Re: [deal.II] Need help setting up the left hand side in a modification of example 52

Why is evaluate_diffusion() declared as const? When removing the const-keyword, compilation works fine (running not tested yet). Am Donnerstag, 3. August 2017 13:08:47 UTC+2 schrieb Wolfgang Bangerth: > > On 08/03/2017 03:16 AM, 'Maxi Miller' via deal.II User Group wrote: &g

[deal.II] FESystem with FE_values_view returns wrong dimensions

I created a FESystem containing three FE_Q-elements with -dimensions each, using the suggested approach in https://www.dealii.org/8.4.0/doxygen/deal.II/classFESystem.html: template std::vector*> Diffusion::create_fe_list( const unsigned int polynomial_degree) { std::vector

Re: [deal.II] FESystem with FE_values_view returns wrong dimensions

Why do I have a scalar base element when initializing FE_Q with dim=3? I thought that I then have the finite element as 3d-element? Am Donnerstag, 3. August 2017 16:32:03 UTC+2 schrieb Wolfgang Bangerth: > > > > > > | > > template > > > > > std::vector*>Diffusion::create_fe_list(constu

Re: [deal.II] FESystem with FE_values_view returns wrong dimensions

I.e. I should rather choose FESystem a(FE_Q(2), 3*dim)? Am Donnerstag, 3. August 2017 16:50:23 UTC+2 schrieb Wolfgang Bangerth: > > On 08/03/2017 08:47 AM, Maxi Miller wrote: > > Why do I have a scalar base element when initializing FE_Q with > dim=3? I > > thought that I then have the finite e

[deal.II] Function for VectorTools::interpolate_boundary_values with different boundary values in multi-variable system

I have a system with three 3d-variables, and would like to apply different boundary conditions on the constraint matrix, depending on the variable. According to the examples I have to create my own function: template class BoundaryValues : public Function { private: equatio

Re: [deal.II] Function for VectorTools::interpolate_boundary_values with different boundary values in multi-variable system

;3*dim>(carrier_density),constraint_matrix); Am Freitag, 4. August 2017 11:14:32 UTC+2 schrieb Wolfgang Bangerth: > > On 08/04/2017 01:38 AM, 'Maxi Miller' via deal.II User Group wrote: > > > > > > but how do I tell that in > > | > > >

Re: [deal.II] Function for VectorTools::interpolate_boundary_values with different boundary values in multi-variable system

t;> dealii::DoFHandler<2>&, int, Step52::BoundaryValues<6>, dealii:: >> ConstraintMatrix&)’ >> VectorTools::interpolate_boundary_values(dof_handler,1, >> BoundaryValues<3*dim>(carrier_density),constraint_matrix); >> >> >>

[deal.II] Component Mask for setting initial conditions

When creating boundary conditions, I can use the parameter component_mask in order to set boundary conditions only for a part of my system. Now I want to set initial conditions for the same system. One approach is using VectorTools::project but there I can not see an option for component_mask in

Re: [deal.II] Component Mask for setting initial conditions

equations.ambient_temperature; } Am Freitag, 4. August 2017 12:34:30 UTC+2 schrieb Wolfgang Bangerth: > > On 08/04/2017 04:23 AM, 'Maxi Miller' via deal.II User Group wrote: > > When creating boundary conditions, I can use the parameter > component_mask in > > order

Re: [deal.II] Component Mask for setting initial conditions

I am filling it up like that, and then (when filling system_matrix) I would like to take a look at the values: for (unsigned int q_point=0; q_point class InitialValues : public Function { private: equation_class eq_class; physics_equations equations; public:

Re: [deal.II] Component Mask for setting initial conditions

I tried to debug it (thus the reason for "volatile"), and reduced the whole code down to (called in SetupSystem()): VectorTools::project(dof_handler, constraint_matrix, QGauss(3), InitialValues(carrier_density), solution); //For checking of content const FEValuesExtracto

Re: [deal.II] Component Mask for setting initial conditions

I think I understand my problem (slowly): Using the project-function means that I write the initial conditions into my solution-vector. But when I call fe_values[N_density] I get a view on the finite element-variable, which is not connected to the solution-vector. Thus I get (in the last line) t

Re: [deal.II] Component Mask for setting initial conditions

I needed rather the value of the vector field at that position, i.e. for fe_values[N_density].value(i, q_point) the value 300 for all q_point-values. Might be that I have to take a look at the tutorials again... Am Samstag, 5. August 2017 16:14:32 UTC+2 schrieb Jean-Paul Pelteret: > > Yeah, tha

[deal.II] Setting off-diagonal elements in the matrix

Having a coupled set of poisson-like equations \partial_t a = \nabla^2 a \partial_t b = \nabla^2 b + \nabla^2 a I can write the weak form(s) as (while ignoring boundary conditions) \partial_t a = (\nabla v_1, \nabla a) \partial_t b = (\nabla v_2, \nabla b) + (\nabla v_2, \nabla a) Now in deal.II I

[deal.II] Explanation of right hand side in example 33?

I try to follow the math in example 33, but am not fully able to understand how the right hand side is included. According to my understanding, I have the function/vector G, multiply it with the test function z, and integrate over both over the whole area. After there are no derivatives included

Re: [deal.II] Explanation of right hand side in example 33?

, 16. August 2017 15:38:44 UTC+2 schrieb Wolfgang Bangerth: > > On 08/15/2017 01:15 AM, 'Maxi Miller' via deal.II User Group wrote: > > I try to follow the math in example 33, but am not fully able to > understand > > how the right hand side is included. According to

[deal.II] Usage of the laplace-matrix in example 23

I am trying to develop my own code based on example 23 (\partial_t U = i/(2k)\nabla^2U), and am trying to understand the purpose of the laplace-matrix. Why is it used in order to multiply in both sides, in comparison with the weak formulation (\nabla\phi_i,\nabla\phi_j) and the assembly of the

[deal.II] Assembly of system matrix for schrödinger equation with complex numbers

I have the following schrödinger equation:

Re: [deal.II] Assembly of system matrix for schrödinger equation with complex numbers

Thanks! Am Montag, 21. August 2017 18:16:30 UTC+2 schrieb Wolfgang Bangerth: > > On 08/21/2017 06:51 AM, 'Maxi Miller' via deal.II User Group wrote: > > > > As far as I am concerned, that looks wrong, but at which step did I make > > a mistake here? >

[deal.II] Solving space-dependent equations using dealII?

Is it possible to solve not only time-, but also space-dependent equations in dealII? As example the nonlinear schrödinger equation: \partial_z U = i(\partial_x^2+\partial_y^2) U with U=U(x, y, z) and U_0=U(x, y, 0)? As far as I know until now, all the equations are based on time, but stationary

[deal.II] Re: Solving space-dependent equations using dealII?

I think my question was badly formulated. When solving the equation above, I get 2d-slices for every step, which are then assembled to a 3d-model in z-direction. Is something similar possible with dealII, or do I have to use external programs for that? And yes, the same can be done with time ins

Re: [deal.II] Re: Solving space-dependent equations using dealII?

t; > Best, > Daniel > > [1] https://www.dealii.org/8.5.0/doxygen/deal.II/classDataOutStack.html > > > Am Dienstag, 22. August 2017 18:01:38 UTC+2 schrieb Wolfgang Bangerth: >> >> On 08/22/2017 09:30 AM, 'Maxi Miller' via deal.II User Group wrote: >

[deal.II] Using Sacado with example 15

I would like to get a feeling in working with Sacado, and would like to try it on example 15. According to my understanding I now have to treat the boundary conditions on my own, and not by using the ConstraintMatrix. Is that still true, or was there an update since 2008? Furthermore, now the f

Re: [deal.II] Using Sacado with example 15

Concerning the constraints: I was refering to this line: In the first case, there is nothing we need to do: we are using a > continuous finite element, and face terms do not appear in the bilinear > form in this case. The second case usually does not lead to face terms > either if we enforce ha

[deal.II] Assembly of matrices for the heat equation when using Sacado generates -nan-values

Based on the answers on my last question (Using sacado in example 15) I thougth that I understood enough in order to extend it to the next step, using Sacado for solving the time-dependent heat-equation. Thus, the function F() is F(u_{n+1}, u_n) = u_{n+1}-u_n - dt*theta*nabla^2 u_{n+1} - dt*(1-

[deal.II] Re: Assembly of matrices for the heat equation when using Sacado generates -nan-values

Found my problem, it was the remeshing. Whenever I was remeshing, I basically killed the old solution, which then resulted in nan-values for it. That in turn generated -nan-values for the system matrix. Am Montag, 28. August 2017 09:52:45 UTC+2 schrieb Uwe Köcher: > > looks like your system matr

[deal.II] Best way to transfer old solutions in time stepping over mesh refinement?

I am trying to implement example 26 using Newton's method and Sacado, in order to get a better feeling for the method. As mentioned before, I had problems with retaining the old solution when remeshing. Thus I rewrote my remeshing sequence as template void HeatEquation::refine_mesh ()

[deal.II] .add() for PETSc::MPI::SparseMatrix

While trying to port step-15 to a MPI-capable version of the program, I came to the following point in the code: for (unsigned int i=0; ihttps://www.dealii.org/8.4.0/doxygen/deal.II/namespaceLocalIntegrators_1_1Advection.html#aa57fdeca62a0708d77768a3bb2aeb826> (i,j)); system_rhs(local_dof_i

[deal.II] Using LinearAlgebraTrilinos::MPI::Vector.l2_norm leads to an MPI_error for multiple nodes

I tried to implement the example from step 15, but using MPI, so that I can get a better grasp on how to write such programs. My residual calculation looks quite similar to the original: const QGauss quadrature_formula(fe.degree+1); FEValues fe_values (fe, quadrature_formula,

[deal.II] Re: Using LinearAlgebraTrilinos::MPI::Vector.l2_norm leads to an MPI_error for multiple nodes

In addition: It fails when declaring the vector residual as global vector via LinearAlgebraTrilinos::MPI::Vector residual; IndexSet solution_relevant_partitioning(dof_handler.n_dofs()); DoFTools::extract_locally_relevant_dofs(dof_handler, solution_relevant_partitioning); residual.reinit(soluti

[deal.II] Re: Using LinearAlgebraTrilinos::MPI::Vector.l2_norm leads to an MPI_error for multiple nodes

Running in debug mode -> make debug, or using deal in Debug configuration? If the former, it did not show anything for that. If the latter, I thought I compiled deal.II both in Release and Debug (ReleaseDebug in CMake). Nevertheless, it did not fix the problem... Thanks! Am Freitag, 8. September

Re: [deal.II] Re: Using LinearAlgebraTrilinos::MPI::Vector.l2_norm leads to an MPI_error for multiple nodes

in communicator MPI_COMM_WORLD with errorcode 255. Looks as if I have to check both vectors for ghost elements before... Thanks! Am Freitag, 8. September 2017 14:48:34 UTC+2 schrieb Bruno Turcksin: > > 2017-09-08 8:35 GMT-04:00 'Maxi Miller' via deal.II User Group > >: &

[deal.II] Access to Trilinos::MPI::Vector::owned_elements() for debugging purposes?

In my program I have a vector "present_solution", defined as LinearAlgebraTrilinos::MPI::Vector present_solution; const size_t dof_numbers = dof_handler.n_dofs(); IndexSet solution_partitioning(dof_numbers), solution_relevant_partitioning( dof_numbers); solution_partitioning = dof_handler.l

Re: [deal.II] Access to Trilinos::MPI::Vector::owned_elements() for debugging purposes?

2017 23:25:07 UTC+2 schrieb Timo Heister: > > Maxi, > > can you try replacing > > present_solution.reinit(solution_relevant_partitioning, MPI_COMM_WORLD); > by > > present_solution.reinit(solution_partitioning, > solution_relevant_partitioning, MPI_COMM_WORLD); > ? &g

[deal.II] AztecOO::Iterate error code -3: loss of precision for offset values in step-15

I rewrote example 15 for a MPI-environment using Trilinos, and solve it with IndexSet solution_relevant_partitioning(dof_handler.n_dofs()); DoFTools::extract_locally_relevant_dofs(dof_handler, solution_relevant_partitioning); LinearAlgebraTrilinos::MPI::Vector completely_distribute

Re: [deal.II] AztecOO::Iterate error code -3: loss of precision for offset values in step-15

l_matrix, cell_rhs, local_dof_indices, system_matrix, system_rhs); in order to keep the zero boundary conditions for the newton update. Am Dienstag, 26. September 2017 03:15:47 UTC+2 schrieb Wolfgang Bangerth: > > On 09/25/2017 01:25 PM, 'Maxi Miller' via deal.II User Group wr

Re: [deal.II] AztecOO::Iterate error code -3: loss of precision for offset values in step-15

ove, but still the sharp drop remains. Why? Thank you! Am Dienstag, 26. September 2017 03:15:47 UTC+2 schrieb Wolfgang Bangerth: > > On 09/25/2017 01:25 PM, 'Maxi Miller' via deal.II User Group wrote: > > > > I do not understand the reason why that happens.

[deal.II] Build of example 9 fails in combination with PETSC

I tried to install dealII together with PETSC, which was installed using the configuration line ./configure PETSC_ARCH=linux-gnu-openmpi-complex --CXX_LINKER_FLAGS=-lc - CC_LINKER_FLAGS=-lc --with-mkl_pardiso=1 --with-blaslapack-dir=/opt/intel/mkl --with-debugging=0 FOPTFLAGS="-fPIC -O3 -march=na

[deal.II] Re: Build of example 9 fails in combination with PETSC

Thanks for the answer, then I am sure that it is/was a bug. I continued by using the non-complex version of PETSC, after I do not need complex numbers here, and deal.II does not support complex values, either. Am Montag, 2. Oktober 2017 16:24:44 UTC+2 schrieb Daniel Arndt: > > Maxi, > > I also s

[deal.II] SystemMatrix gets large fast, error or to be expected?

I have a system of coupled heat equations, written for MPI using LinearAlgebraTrilinos::MPI::SparseMatrix system_matrix; as the sparse system matrix. Now for certain values the program runs fine, but for other values (which I am interested in) the RAM consumption goes up rapidly after several

Re: [deal.II] SystemMatrix gets large fast, error or to be expected?

10/03/2017 01:04 AM, 'Maxi Miller' via deal.II User Group wrote: > > I have a system of coupled heat equations, written for MPI using > > > > | LinearAlgebraTrilinos::MPI::SparseMatrixsystem_matrix; | > > > > as the sparse system matrix. Now for certain

Re: [deal.II] SystemMatrix gets large fast, error or to be expected?

The system has ~27000 DoF, thus there must be an error on my side. Thanks! -- 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

Re: [deal.II] SystemMatrix gets large fast, error or to be expected?

I rechecked after letting the program run for several time. Apparently the DoF-number rises after some time, with a current value of 2223802 DoFs. Can that make sense? If it makes sense, it also explains the memory consumption. Am Dienstag, 3. Oktober 2017 17:24:42 UTC+2 schrieb Maxi Miller: >

[deal.II] Initial setup of dealII fails for Intel MPI, but not for OpenMPI

I tried to set up dealII with the Intel MPI compilers (using gcc 6.3.0 in the background), but already in the initial setup before I can choose any values I get the error Building CXX object CMakeFiles/cmTC_8341d.dir/testCXXCompiler.cxx.o /share/apps/software/Compiler/intel/2017.1.132-GCC-6.3.

Re: [deal.II] SystemMatrix gets large fast, error or to be expected?

Of course, having an (former) undetected bug in the equations itself also can lead to unwanted memory consumption... Thanks for the help! Am Mittwoch, 4. Oktober 2017 17:23:05 UTC+2 schrieb Wolfgang Bangerth: > > On 10/04/2017 01:03 AM, 'Maxi Miller' via deal.II User Gro

[deal.II] Re: Initial setup of dealII fails for Intel MPI, but not for OpenMPI

When removing everything, and starting from a clean build-directory, I get LANG=eng cmake .. -DCMAKE_CXX_COMPILER=/share/apps/software/Compiler/intel/ 2017.1.132-GCC-6.3.0-2.27/impi/2017.1.132/bin64/mpicxx -DCMAKE_C_COMPILER= /share/apps/software/Compiler/intel/2017.1.132-GCC-6.3.0-2.27/impi/2017.1

[deal.II] L2-norm gets decreased to certain value, but stays above threshold

Having a nonlinear, time-dependent heat equation, I tried to implement it as in step-15 and in step-33. Everytime my l2-value goes below a certain threshold, I make one step in time (or I refine the mesh). At the beginning the L2-value goes down rather fast, but later it slows down, and even sto

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

correct). Can it be related to (as mentioned) large value differences? For small value differences the l2-norm goes below the threshold, for large it does not. Am Donnerstag, 5. Oktober 2017 16:59:30 UTC+2 schrieb Wolfgang Bangerth: > > On 10/05/2017 08:52 AM, 'Maxi Miller' via de

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

roblemcompute_residual > > > A > > On 5 Oct 2017, at 16:13, 'Maxi Miller' via deal.II User Group < > dea...@googlegroups.com > wrote: > > Yes, I forgot to clarify. I am calculating the l2-norm of the right hand > side, i.e. which contains F(U) (when staying

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

eroed: > > for (unsigned int i=0; i <http://www.dealii.org/developer/doxygen/deal.II/classDoFHandler.html#a54e7270f2ba6206604f794114b39a2aa>(); > > ++i) > if (boundary_dofs[i] == true) > residual(i) = 0; > > > On 5 Oct 2017, at 17:01, 'Maxi Miller' via deal

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

Compared to the target value? Depending on the values between ~1e-6 and 2e-3, but it can be larger, too. I will check the latter function and compare it to the current solution, but I thought by using ZeroFunction as boundary condition that then all boundary_dofs are set to zero. Am Donnerstag,

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

loop over all of those and set the residual entry to zero. This > happens in the following lines which we have already seen used in step-11: > > I’m not sure if this actually your question, but based on the above and > Timo’s response, I hope this helps > > A > >

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

Concerning the first point: Currently I am controlling the solver as in SolverControl solver_control (dof_handler.n_dofs(), (system_rhs.l2_norm() > 0) ? 1e-6 * system_rhs.l2_norm() : 1e-6); Is your suggestion to decrease that threshold? Concerning the

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

Short update: Changing the threshold from 1e-6 to 1e-8 did not change anything, it still loops at the same values as final values. Am Freitag, 6. Oktober 2017 11:56:59 UTC+2 schrieb Maxi Miller: > > Concerning the first point: > Currently I am controlling the solver as in > SolverControl

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

Isn't that basically the method done in step-15 and step-33 (line search)? But why should my step length (I assume it is alpha) go to zero? Or is it the newton-update (which also should go to zero, when being close to the solution)? Am Freitag, 6. Oktober 2017 15:13:08 UTC+2 schrieb Wolfgang Ba

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

Bangerth: > > On 10/06/2017 07:20 AM, 'Maxi Miller' via deal.II User Group wrote: > > Isn't that basically the method done in step-15 and step-33 (line > search)? But > > why should my step length (I assume it is alpha) go to zero? Or is it > the > > newt

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

ktober 2017 15:27:40 UTC+2 schrieb Wolfgang Bangerth: >>> >>> On 10/06/2017 07:20 AM, 'Maxi Miller' via deal.II User Group wrote: >>> > Isn't that basically the method done in step-15 and step-33 (line >>> search)? But >>> > why sho

Re: [deal.II] L2-norm gets decreased to certain value, but stays above threshold

p-15 uses a fixed alpha = 0.1, but in the suggestions it is proposed >>> to increase alpha to 1, when suitable in order to increase convergence. I >>> am also using a fixed alpha of 0.1 at the moment, thus it will never go to >>> zero. >>> >>> Am Freit

[deal.II] Re: Initial setup of dealII fails for Intel MPI, but not for OpenMPI

I would like to try to help bug fixing, thus I was wondering where I can find the commands which are run when building deal.II the first time (which compile the test programs)? Thanks! Am Donnerstag, 5. Oktober 2017 15:53:00 UTC+2 schrieb Maxi Miller: > > When removing everything, and starting f

[deal.II] set_zero fails for multiple threads

When trying to run my program on an i7-CPU with 4/8 threads, I get the following error when calling set_zero() from the ConstraintMatrix-class: An error occurred in line <565> of file in function void dealii::TrilinosWrappers::MPI::V

Re: [deal.II] Re: Initial setup of dealII fails for Intel MPI, but not for OpenMPI

17 18:00:09 UTC+2 schrieb Wolfgang Bangerth: > > On 10/06/2017 03:54 PM, 'Maxi Miller' via deal.II User Group wrote: > > I would like to try to help bug fixing, thus I was wondering where I can > find > > the commands which are run when building deal.II the first time (wh

Re: [deal.II] set_zero fails for multiple threads

h > > wrote: > > On 10/10/2017 05:15 AM, 'Maxi Miller' via deal.II User Group wrote: > >> > >> > >> When trying to run my program on an i7-CPU with 4/8 threads, I get the > >

Re: [deal.II] set_zero fails for multiple threads

10. Oktober 2017 22:54:46 UTC+2 schrieb Timo Heister: > > On Tue, Oct 10, 2017 at 3:49 PM, 'Maxi Miller' via deal.II User Group > > wrote: > > I am coming out of a loop where I am doing > > hanging_node_constraints.distribute_local_to_global(cell_residual, > >

[deal.II] Error when applying initial values to MPI::Vector in multiple dimensions

I try to apply initial values to a vector defined as *LinearAlgebraTrilinos::MPI::Vector *using VectorTools::project (dof_handler, hanging_node_constraints, QGauss(fe.degree+1), InitialValues(), local_solution); When initializing the variable fe (as FESystem) with one or two components, i

[deal.II] Re: Error when applying initial values to MPI::Vector in multiple dimensions

Additional: Even though it compiles in release mode, as soon as I run it with multiple nodes, I get a segfault at that place: mpirun noticed that process rank 0 with PID 0 on node linux-lb8c exited on signal 11 (Segmentation fault). Am Freitag, 13. Oktober 2017 11:05:01 UTC+2 schrieb Maxi Mil

[deal.II] Re: Error when applying initial values to MPI::Vector in multiple dimensions

Yes, .interpolate() works fine. But what is the difference between interpolate() and project()? Am Montag, 16. Oktober 2017 08:54:02 UTC+2 schrieb Denis Davydov: > > Or you may want to use interpolate if this is enough > > https://www.dealii.org/developer/doxygen/deal.II/namespaceVectorTools.html

[deal.II] Re: Error when writing files using write_vtu_header

The explanation makes sense (race condition), but I am using exactly the code from example 40, which should prevent such problems... Am Dienstag, 17. Oktober 2017 08:49:55 UTC+2 schrieb Denis Davydov: > > AFAIK there is no know problem. > Make sure you write this only from a single MPI core, tho

Re: [deal.II] Using Sacado with example 15

How would I use the function get_function_gradients_from_local_dof_values with a vector_extractor? When trying to compile it with the vector shown above (including changing the first value) it fails with a rather long list of template errors. Thanks! Am Freitag, 25. August 2017 09:51:37 UTC+2 s

[deal.II] Re: Error when writing files using write_vtu_header

I tried, but I did not manage to reproduce it until now. Based on my recent experience changing the amount of CPUs does not help, sometimes it crashes, sometimes it does not. Am Dienstag, 17. Oktober 2017 14:36:55 UTC+2 schrieb Bruno Turcksin: > > Hi, > > It could be a problem with the file syst

Re: [deal.II] Re: Error when writing files using write_vtu_header

As written above, I am using the code from example 40, so it should not be a problem of the code. But I will talk to the administrator about that, thanks! Am Dienstag, 17. Oktober 2017 18:41:57 UTC+2 schrieb Wolfgang Bangerth: > > On 10/17/2017 09:46 AM, 'Maxi Miller' via de

  1   2   3   4   >