[deal.II] Re: shmem: mmap: shared memory error on Mac OS

Dear Bruno, Thank you very much for your reply and apologies for my late reply. I think that was exactly the issue. After upgrading to open-mpi 5.0.7, the problem seems to be gone. Thanks again and best regards, Nik On Thursday, February 6, 2025 at 6:37:59 AM UTC-8 bruno.t...@gmail.com wrote:

[deal.II] Re: Supported finite elements in the matrix-free infrastrucutre

Dear Peter, thanks a lot for your answer. Best, Giuseppe Il giorno lunedì 3 marzo 2025 alle 19:53:58 UTC+1 peterr...@gmail.com ha scritto: > The normal RT element is not working. But FE_RaviartThomasNodal is working > with MF. > > Regarding simplices: `FE_Simplex(DG)P` (also with FESystem) is

[deal.II] Re: Supported finite elements in the matrix-free infrastrucutre

The normal RT element is not working. But FE_RaviartThomasNodal is working with MF. Regarding simplices: `FE_Simplex(DG)P` (also with FESystem) is working. Best, Peter On Monday, 3 March 2025 at 19:38:08 UTC+1 gius...@gmail.com wrote: > Dear all, > I'm writing this message to ask some confirma

[deal.II] Re: Seeking Advice on Implementing Solver with Trillinos and PETSc Vectors

Hello, Thank you both for the quick response and useful advice! Wish you a great weekend Regards, Hussein On Friday, February 21, 2025 at 2:18:49 PM UTC+1 blais...@gmail.com wrote: > Dear Hussein, > It's not a big deal to have the VectorType as a class template. The only > thing you w

[deal.II] Re: Seeking Advice on Implementing Solver with Trillinos and PETSc Vectors

Dear Hussein, It's not a big deal to have the VectorType as a class template. The only thing you will get is that every function / member of the class will be compiled twice (once for every vector type). In the software my group develops (lethe https://github.com/chaos-polymtl/lethe), some class

[deal.II] Re: Matrix-Free and FE_Nothing

Hi Bruno, Thanks for your quick reply, that worked perfectly. Many thanks, Thomas On Thursday, February 20, 2025 at 2:40:52 PM UTC bruno.t...@gmail.com wrote: > Hello Thomas, > > If you use FE_Nothing, you don't want to loop over the cell_range. Instead > you create a subrange of cells using

[deal.II] Re: Matrix-Free and FE_Nothing

Hello Thomas, If you use FE_Nothing, you don't want to loop over the cell_range. Instead you create a subrange of cells using create_cell_subrange_hp_by_index() and then loop over the cells in t

Re: [deal.II] Re: Parallel computation fails to converge

On 2/14/25 00:07, Junxiang Wang wrote: Will the distribution process change the data? or there is anything I need to take care of ? The code also includes solution data interpolation and history data transfer to the new refined mesh. My general strategy for cases such as your is to make

Re: [deal.II] Re: Re-import an adaptively refined mesh (with hanging nodes)

On 2/14/25 04:51, Matteo Malvestiti wrote: But however there is the problem of the positing of the vertices, once placed thanks to the Manifold description, this is the result _after partitioning_: (only 1 process being used for test purposes) after_partitioning.png Any idea on how to solv

[deal.II] Re: Projecting solutions between distributed meshes

Hello Daniel, for interpolate I know that it is between two different finite element spaces on the same mesh, e.g. from Q2 to Q1 elements, or in your case from Q1 to dG0. What you will need to do is construct a Q1 DoFHandler on each mesh and use `interpolate_to_different_mesh` to get a Q1 soluti

Re: [deal.II] Re: Parallel computation fails to converge

On 2/11/25 03:17, Junxiang Wang wrote: Therefore, Can I interpret the issue I encounter in that it is troublesome to distribute the local_rhs when the hanging node is located on the shared edge of two elements and also the two elements belong to different processors respectively? Junxiang:

Re: [deal.II] Re: Parallel computation fails to converge

Sorry, the node distribution is as follows. On Tuesday, February 11, 2025 at 6:17:39 PM UTC+8 Junxiang Wang wrote: > Hi Bruno, > > I further print information about the RHS of the two elements that have > the abnormal residual value. It turns out the nodal values are all the same > between the

Re: [deal.II] Re: Parallel computation fails to converge

Hi Bruno, I further print information about the RHS of the two elements that have the abnormal residual value. It turns out the nodal values are all the same between the serial and parallel simulation. Therefore, Can I interpret the issue I encounter in that it is troublesome to distribute the

Re: [deal.II] Re: Parallel computation fails to converge

Hello Junxiang, It looks like the issue happens at a hanging node. In step-40, we show how to see which processor each cell belongs to. I think it would be interesting to output that image and see if the issue happens because the hanging node is shared between different processors. Best, Bruno

Re: [deal.II] Re: Parallel computation fails to converge

What happens if you don't refine the mesh? Is everything still the same? Bruno Le sam. 8 févr. 2025 à 06:26, Junxiang Wang a écrit : > In the first 80 steps, RHS and Stiffness are all the same. After several > step of refine mesh, the RHS then turns to different from the serial one. > > On Satu

Re: [deal.II] Re: Parallel computation fails to converge

On 2/8/25 04:24, Junxiang Wang wrote: It is really frastrating that increasing number of cores can make a difference... Yes, but at least you now have something that you know should not happen, but does happen. It is a very concrete thing (unlike "solver not converging") that you can explor

Re: [deal.II] Re: Parallel computation fails to converge

Are you making sure you’re initializing the vector to 0 in release mode. Subramanya. सुब्रह्मण्य . On Sat, Feb 8, 2025 at 4:56 PM Junxiang Wang wrote: > In the first 80 steps, RHS and Stiffness are all the same. After several > step of refine mesh, the RHS then turns to different from the seri

[deal.II] Re: Parallel computation fails to converge

In the first 80 steps, RHS and Stiffness are all the same. After several step of refine mesh, the RHS then turns to different from the serial one. On Saturday, February 8, 2025 at 7:24:07 PM UTC+8 Junxiang Wang wrote: > Dear Bruno, > > Thanks a lot for your advice. > > I tested it. The code ru

[deal.II] Re: Parallel computation fails to converge

Dear Bruno, Thanks a lot for your advice. I tested it. The code runing in parellel has exactly the same norm of stiffness mtx but different norm of RHS residual compared to running in serial. It is really frastrating that increasing number of cores can make a difference... Best, Junxiang.

[deal.II] Re: Parallel computation fails to converge

Hello Junxiang, It's very hard to know what the issue could be. What you could do is print some information about your matrix and right hand side (for instance, different norms) and check that they are the same in serial and in parallel. By comparing the serial and parallel simulations at diffe

[deal.II] Re: shmem: mmap: shared memory error on Mac OS

Hello Nik, This looks like an OpenMPI issue. I have you tried the suggestions from https://github.com/open-mpi/ompi/issues/12784 Best, Bruno On Friday, January 17, 2025 at 4:06:07 PM UTC-5 nikl...@stanford.edu wrote: > Dear all, > > Thanks to your help earlier, I was able to compile my progra

Re: [deal.II] Re: Changes in SolutionTransfer

Hello Winnifried, > With the old functionality it was possible to check if a refinement of > the mesh was performed between storage and loading and then interpolate > the vector onto the refined mesh used for computation. The new SolutionTransfer workflow performs the transfer of data during the

Re: [deal.II] Re: Using CharManifold to implement an Ellipsoid surface gives strange results

On 1/30/25 22:01, Amit Singh wrote: I was concerned about some elements of very high aspect ratio occurring at the chart boundaries. They appear like ridges on the surface. I have shown examples for both a sphere and an ellipsoid. Also, the mesh loses symmetry i.e the y-plus and the y-minus p

Re: [deal.II] Re: Using CharManifold to implement an Ellipsoid surface gives strange results

On 1/30/25 08:56, Amit Singh wrote: I am able to get Ellipsoidal shapes but the meshes have highly distorted elements at some points as shown below. Any suggestions for a better implementation for this manifold will be helpful. From the picture, it actually looks like a fairly decent mes

Re: [deal.II] Re: Changes in SolutionTransfer

Dear Marc, thanks for the information, I think I missed this merge in functionality. The serial SolutionTransfer class just had the responsibility to transfer vectors between differently refined meshes. How did you use it to load vectors from disk? I don't remember that it had such a functionali

[deal.II] Re: Changes in SolutionTransfer

Hello Winnifried, > I noticed that apparently a change in the SolutionTransfer class is coming with the next version. Particularly, the function 'prepare_for_pure_refinement' seems to be removed (was present in version 9.6.2 but is gone in git)? We had separate implementations of `SolutionTra

[deal.II] Re: Help understanding MappingQ(2) for curved elements

Thanks a lot, That solved the issue! I really appreciate your help. For future readers, here is the correct call to build_patches(): data_out.build_patches(mapping, 5, DataOut<2,2>::CurvedCellRegion ::curved_inner_cells); Best regards, Matteo Malvestiti Il giorno domenica 26 gennaio 2025 alle 11

[deal.II] Re: Help understanding MappingQ(2) for curved elements

Hi Matteo, it is probably only the output. DataOut::build_patches() has an additional parameter that specifies the curved region ( https://www.dealii.org/developer/doxygen/deal.II/classDataOut.html#a1303b5d99f10f3fae984e7d406cfeb37). Select CurvedCellRegion::curved_inner_cells. Best, Peter O

[deal.II] Re: Does multiple matrix free from one tria gives out the same cell batch id

Hi Subramanya, Thanks for the quick reply! So, the cell batch id does change between different dof_handlers. I think that's a caveat worth mentioning in the tutorial. Also, I'm curious about the partition algorithm, what is actually related to the cell partition, the Mesh, the FE, the partiti

[deal.II] Re: Does multiple matrix free from one tria gives out the same cell batch id

Hi ChengJiang , Martin answered this same question, when I asked this recently. https://groups.google.com/g/dealii/c/OWd7QsSSR18/m/q0j08zwLCQAJ?utm_medium=email&utm_source=footer Cheers, Subramanya On Thursday, January 23, 2025 at 8:18:51 AM UTC-8 richard...@gmail.com wrote: > Hi all, > > I

[deal.II] Re: Re-import an adaptively refined mesh (with hanging nodes)

Dear Matteo, The best solution here would be to serialize the mesh so you can re-read it anew. This would be identical to what you would do when you restart a simulation. We have a new tutorial that shows just how exactly this is done: https://www.dealii.org/current/doxygen/deal.II/step_83.html

[deal.II] Re: Get MueLu of Trilinos working on arm-based Mac Book M4

Dear Daniel, Thank you very much for your answer. I was indeed able to find a more detailed report of what was going wrong during the deal.ii test for MueLu by searching exactly as you mentioned. Thank you and best regards, Nik On Wednesday, January 8, 2025 at 7:26:00 PM UTC+1 d.arnd...@gmail.c

[deal.II] Re: Get MueLu of Trilinos working on arm-based Mac Book M4

Dear James, Thank you very much for your very kind and detailed reply. Using your recommendations, I managed to install Trilinos + MueLu working with deal.ii master. Thank you very much for your help with this! The only thing I had to change from the default recipe here https://github.com/deal

Re: [deal.II] Re: Error with Simplex and adaptive mesh with FESubfaceValues

On 1/8/25 22:00, Sanghyun Lee wrote: *Now, here is how I fixed the issue. * > [...] Would you mind submitting a pull request with your changes? That's the easiest way to discuss the correctness of what you have. Best W. -- --

[deal.II] Re: Error with Simplex and adaptive mesh with FESubfaceValues

I fixed the issue; but before I discuss that, = first, sorry there was a typo in my above email. The correct discussion is ... Assert (neighbor->neighbor_child_on_subface(neighbor_face_subface.first, neighbor_face_subface.second) == cell, ExcInternalError()); Please see the figure: 1.

[deal.II] Re: Error with Simplex and adaptive mesh with FESubfaceValues

Dear Peter Thanks for the reply and pointing the direction that I was exactly looking for. I downloaded and installed with the modified files on the PR, and checked with my code. The initial concern is now removed. However, now I found one more discrepancy for the *simplex case* in the foll

[deal.II] Re: Error with Simplex and adaptive mesh with FESubfaceValues

Hi Sanghyun, there is an open PR that should (hopefully) fix the issue: https://github.com/dealii/dealii/pull/17908. It is not merged yet. But you could try it out and gives us feedback! Best, Peter On Wednesday, 8 January 2025 at 21:10:08 UTC+1 shl@gmail.com wrote: > Hello! > Happy New Y

[deal.II] Re: Get MueLu of Trilinos working on arm-based Mac Book M4

The problem with boost should have been resolved in 9.6.0. For -- MueLu was found but is not usable through Epetra! Disabling MueLu support. You should be able to get more information searching through build/CMakeFiles/ for TRILINOS_MUELU_IS_FUNCTIONAL. Best, Daniel On Wednesday, January 8, 2

[deal.II] Re: Get MueLu of Trilinos working on arm-based Mac Book M4

Dear Nik, I’ve been working with deal.II and Trilinos (including MueLu) across various platforms, including macOS, for a while now. I understand your frustration, as the installation and configuration process can be tricky. Let me briefly share some insights based on my experience. *Key Is

[deal.II] Re: merging unmatching triangulations

Dear Abbas, I appreciated your answer as I did not know about step 46. Thank you. I already merged cells to remove the hanging node, I think there is no other solution for it now. Some studies are out there about gluing different faces that do not match. Not sure if that would help anyone sear

[deal.II] Re: merging unmatching triangulations

Dear Njwa, Sorry I misinterpreted your question sorry my bad. Yeah I don't know how to answer that one. Never done something like that before. On Wednesday, December 25, 2024 at 9:55:57 PM UTC+1 najwaa...@gmail.com wrote: > Dear Abbas, > > Thank you for your reply. I quickly looked at step 4

[deal.II] Re: merging unmatching triangulations

Dear Abbas, Thank you for your reply. I quickly looked at step 46 and did not notice that they had the solution to my question. I will read it more carefully tomorrow. I forgot to say that I know that a trivial solution is to refine the corser side. However, my question is, can this be avoid

[deal.II] Re: merging unmatching triangulations

Maybe something like step 46? Not sure if that would work so maybe someone would add to this or give you something better. best, Abbas On Wednesday, December 25, 2024 at 6:57:02 PM UTC+1 najwaa...@gmail.com wrote: > Dear all, > > I have a quick question here. > > I know that merge triangul

Re: [deal.II] Re: Convergence issues faced in strain L2 function

Hi Dr. Bangerth, Thank you for the explanation. Yes I just did the problem with 1 element and figured out that because I was using vector to output the strain the errors come up. I should have checked this before. Now with scalar definition my problem is converging even with multiple elements an

Re: [deal.II] Re: Convergence issues faced in strain L2 function

On 12/15/24 08:40, Pratyusa Kar wrote: Also I checked with a 4 element mesh and 1 quadrature point. However, I see a noninvertible mass matrix. I do not understand what is missing in the code. Pratyusa: OK, so you already determined that the matrix is not invertible. The question is why. What

[deal.II] Re: Convergence issues faced in strain L2 function

Also I checked with a 4 element mesh and 1 quadrature point. However, I see a noninvertible mass matrix. I do not understand what is missing in the code. Thank you. On Saturday, December 14, 2024 at 6:41:36 PM UTC-6 Pratyusa Kar wrote: > Hi, > > I have a simple 2D linear elasticity problem,

[deal.II] Re: mpirun failured for step-17 on WSL

Hello Pei-Liang, Do you have the same problem if you use a simple hello_world code that doesn't use deal.II? This will tell us if the problem is with deal.II or with mpi. Best, Bruno On Tuesday, December 3, 2024 at 3:25:30 AM UTC-5 peilia...@gmail.com wrote: > Hello, > > I am now trying to r

[deal.II] Re: why step-24 result is more "angular" than step-89

for example, i created a 4 grids mesh with 3-order DGQ<2> in step-24, GaussPluse expression: exp(-1000 * ( (x-x0)^2 + (y-y0) ^ 2 )). distribution of the quadrature points distribution of : [image: quad_points.png] The initial value plot doesn't seem to have that many control points: [image: ini

[deal.II] Re: Error installing deal.II in the University of Michigan Great Lakes cluster

Hello David, It seems that the error is from TBB. Is TBB part of the intel/2022.1.2 module? Best, Bruno On Sunday, November 24, 2024 at 5:36:15 PM UTC-5 dmon...@umich.edu wrote: > Hello, > > I am attempting to install deal.II using candi in the University of > Michigan Great Lakes cluster (2

Re: [deal.II] Re: access coordinates of quadrature points

On 10/28/24 15:24, Yann Jobic wrote: I found my mistake. The written code works fine, but i cannot inspect the values directly in the debugger. That confused me. Sorry for polluting the mailing list. Shall both email erased ? You can, but you don't need to. I'm glad you found the issue. My

[deal.II] Re: access coordinates of quadrature points

I found my mistake. The written code works fine, but i cannot inspect the values directly in the debugger. That confused me. Sorry for polluting the mailing list. Shall both email erased ? Regards, Yann Le 10/28/2024 à 6:36 PM, Yann Jobic a écrit : Hello, I'm trying to access the coordinates o

Re: [deal.II] Re: cmake error when building with arbox

Kaushik, the last error message you showed was observed also here: https://github.com/dealii/dealii/issues/15749#issuecomment-1636841767 Best, Marco Il giorno giovedì 17 ottobre 2024 alle 21:48:39 UTC+2 k.d...@gmail.com ha scritto: > Hello again, I may have declared victory a little too early

Re: [deal.II] Re: cmake error when building with arbox

Hello again, I may have declared victory a little too early. cmake configure completes but I am getting some build error. This may be due to version compatibility. Please let me know what versions of Trilinous, Kokkos (that comes with Trilinous), ArborX are compatible with dealii 9.4? I am using de

Re: [deal.II] Re: cmake error when building with arbox

Thanks for the tip. Using kokkos that came with trillions solved my problem. On Wednesday, October 16, 2024 at 2:08:20 PM UTC-4 d.arnd...@gmail.com wrote: > Kaushik, > > You can use the same flags you are using to configure the standalone > Kokkos to configure Kokkos inside Trilinos. > > Best,

Re: [deal.II] Re: cmake error when building with arbox

Kaushik, You can use the same flags you are using to configure the standalone Kokkos to configure Kokkos inside Trilinos. Best, Daniel On Wed, Oct 16, 2024 at 12:37 PM Kaushik Das wrote: > > I see. The only reason I installed Kokkos separately was I wanted to use > Kokkos GPU and LAMDA support

Re: [deal.II] Re: cmake error when building with arbox

I see. The only reason I installed Kokkos separately was I wanted to use Kokkos GPU and LAMDA support for other parts of my code. But I could not find how to do that for kokkos that come with Trilinos. Specifically, I want to set these kokkos cmake configs -DKokkos_ENABLE_CUDA=ON, -DKokkos_ARCH_PAS

Re: [deal.II] Re: cmake error when building with arbox

The issue is that you have two versions of Kokkos installed. You can only have a single version of Kokkos. Remove the version of Kokkos you installed yourself and reinstall ArborX . Set the Kokkos directory to the Trilinos directory when configuring ArborX. Bruno Le mer. 16 oct. 2024 à 10:04, Kau

Re: [deal.II] Re: cmake error when building with arbox

Thanks for the clarification. I installed Kokkos separately, and I also have Trilinos support. Can this issue be related to Kokkos's version? It seems the latest Arborx needs Kokkos 4.3.00, but I have 4.3.01. It's not a major version mismatch, though. On Wed, Oct 16, 2024 at 9:58 AM Bruno Turcksin

[deal.II] Re: cmake error when building with arbox

Kaushik, The message you see is the auto-detect trying to find ADOLC. It's OK if the auto-detect does not find ADOLC. The real issue is with Kokkos. When you installed ArborX, did you use the Kokkos library from Trilinos or did you install Kokkos yourself too? Best, Bruno On Tuesday, October

Re: [deal.II] Re: FSI with separate grids.

If you want to use BEM, then also take a look at step-34, and at WaveBEM. Luca > Il giorno 9 ott 2024, alle ore 01:45, Wolfgang Bangerth > ha scritto: > > On 10/8/24 12:19, Tim Braun wrote: >> ** >> I'll add that I see step-70 as an example, which is time-dependent but does >> address the c

Re: [deal.II] Re: FSI with separate grids.

On 10/8/24 12:19, Tim Braun wrote: ** I'll add that I see step-70 as an example, which is time-dependent but does address the coupling process between a solid and fluid. Is that a good place to work from? I would suggest looking through all the programs in the range 60-90. Some of these use

[deal.II] Re: FSI with separate grids.

I'll add that I see step-70 as an example, which is time-dependent but does address the coupling process between a solid and fluid. Is that a good place to work from? On Tuesday, October 8, 2024 at 12:55:50 PM UTC-4 Tim Braun wrote: > I am designing a study of elastic responses of structures un

[deal.II] Re: Using PetscWrappers::TimeStepper with LinearAlgebra::distributed::Vector

Hi Praveen, PETScWrappers::TimeStepper has only been explicitly instantiated for the PETScWrappers Vectors: https://github.com/dealii/dealii/blob/25495718f4bb6aa45a8f6c8cc4f2aedbe8d62231/source/lac/petsc_ts.cc#L82-L89 Will deal.II compile if you add the corresponding instantiations for the Line

[deal.II] Re: Announcing deal.II 9.6

All: I have two follow-ups to the 9.6 announcement. This is the first: To every deal.II release, many many people contribute. The release paper at https://www.dealii.org/deal96-preprint.pdf has 17 authors that have contributed in a major way to this release. It also lists 38 others who have co

[deal.II] Re: Announcing deal.II 9.6

The other announcement about this release: There are numerous new tutorial and code gallery programs in this release that are perhaps interesting to some of you: * step-83 demonstrates how one can implement checkpoint/restart functionality in deal.II-based programs, using the BOOST serializ

[deal.II] Re: Issues with BOOST when compiling deal.II using candi

Hi all, Sorry, turns out Ninja was picking files from the clusters module list. Thanks, Greg On Wednesday 18 September 2024 at 14:09:04 UTC+1 Gregory Christian wrote: > Hi all, > > I'm trying to install deal.II using candi on a linux cluster, but when I > try to compile conservation-law/ryuji

Re: [deal.II] Re: Q&A: Issue with Adaptive Refinement for Higher-Order Elements Q_k+1

Thanks On Monday, September 16, 2024 at 5:14:03 PM UTC+3 Wolfgang Bangerth wrote: > On 9/16/24 05:38, Najwa Alshehri wrote: > > > > It could be an issue related to having one of the meshes circle > approximated > > by a polygon. I have a guess that the error of not fitting the circle > exactl

Re: [deal.II] Re: Q&A: Issue with Adaptive Refinement for Higher-Order Elements Q_k+1

On 9/16/24 05:38, Najwa Alshehri wrote: It could be an issue related to having one of the meshes circle approximated by a polygon. I have a guess that the error of not fitting the circle exactly is seen worse using higher order elements. You may be interested in this discussion: https://deal

Re: [deal.II] Re: dealii installation error : git clone failed

thank you for this hint ! it works and the installation goes forward and ends correctly the compilation of deal.II... until an other compilation error with unknown Tpetra objects not found at link step but that's an other story ! Regards, Le vendredi 13 septembre 2024 à 17:27:27 UTC+2, bruno.t.

Re: [deal.II] Re: FESystem and FEInterfaceValues: how to get component

On 9/14/24 08:04, Praveen C wrote:          unsigned int ii = (i < n_cell_dofs) ? i : i - n_cell_dofs;          const auto c = fe_face_values.get_fe().system_to_component_index(ii).first; Praveen: this makes the assumption that you first have the DoFs from one side, and then the ones from t

Re: [deal.II] Re: Q&A: Issue with Adaptive Refinement for Higher-Order Elements Q_k+1

Najwa: I ran step-22 with Q2^d-Q1 and Q3^d-Q2, and I observed the same behavior. Additionally, I found that in my code, even when using Q2, these jumps also appear, though on a smaller scale and in fewer locations. This brings me to an updated question: * Is this behavior not significant

[deal.II] Re: Q&A: Issue with Adaptive Refinement for Higher-Order Elements Q_k+1

Dear all, I have an important update on the situation. I ran step-22 with Q2^d-Q1 and Q3^d-Q2, and I observed the same behavior. Additionally, I found that in my code, even when using Q2, these jumps also appear, though on a smaller scale and in fewer locations. This brings me to an updated

[deal.II] Re: FESystem and FEInterfaceValues: how to get component

I am doing the following in my face worker, which seems to work. best praveen const unsigned int n_cell_dofs = fe_face_values.get_fe().n_dofs_per_cell(); const unsigned int n_face_dofs = fe_face_values.n_current_interface_dofs(); for(unsigned int q=0; q num_flux(nvar); PDE::numeri

Re: [deal.II] Re: dealii installation error : git clone failed

Guy, In deal.II-toolchain/packages/dealii.package try changing SOURCE= https://github.com/dealii/ to SOURCE=https://dealii.org/downloads/ and PACKING=git to PACKING=.tar.gz I haven't used CANDI in a long time so I am not sure this will work. Best, Bruno Le ven. 13 sept. 2024 à 11:22, Guy Moebs

[deal.II] Re: dealii installation error : git clone failed

Hi Bruno, thank you for your answer and the link. I try the several solutions proposed of the "stackoverflow" page but they fail. I have downloaded the .tar.gz package and put it in the candi tree next to the others, it doesn't work too. candi.sh doesn't seem to see it and always try to git clon

[deal.II] Re: dealii installation error : git clone failed

Guy, According to this , it might be a network issue. It seems that using `git config --global pack.window 1` might solve the issue. You could also download the release tarball instead of

Re: [deal.II] Re: Distributed Block-Gauss-Seidel (BGS) Preconditioning

Nik, it sounds like you've already figured out the big building blocks. Indeed, the right way to deal with this is to use different finite elements on the different parts of the domain (like in step-46), and then to sort DoFs by vector component -- which then automatically produces the block

[deal.II] Re: Distributed Block-Gauss-Seidel (BGS) Preconditioning

Quick update on the post above after some more reading of the documentation - I apologize that I didnt't wait with the post until I had fully read into these things. On Thursday, September 5, 2024 at 11:43:18 PM UTC+2 Nik Markus Leuenberger wrote: Dear all, It has been a very nice experience

[deal.II] Re: dealii installation error - non-existent path when running "make test"

I am closing this conversation since it's an issue on the server side of my department. On Tuesday, August 27, 2024 at 3:24:56 PM UTC-5 Mark Simmons wrote: > I think I found where the issue is but I am not sure how to solve it. When > I am running cmake to generate the build files, I see this l

[deal.II] Re: dealii installation error - non-existent path when running "make test"

I think I found where the issue is but I am not sure how to solve it. When I am running cmake to generate the build files, I see this line in green. INCLUDE_DIRECTORIES: /storage/mbsimmon/EXTERNAL_SOFTWARES/COMPUTE_12/petsc/include;/storage/mbsimmon/EXTERNAL_SOFTWARES/COMPUTE_12/ petsc/arch-

[deal.II] Re: Assemble of a matrix in a parallel code

Dear Giacomo, I would suggest you take a look at step-40. This step does a pretty good job of assembling a matrix in parallel on a distributed triangulation. Here it's hard to tell if you are doing something bad / good without looking at the full code itself :) If you have any questions, please

[deal.II] Re: dealii installation error - non-existent path when running "make test"

the commands for petsc are done in a separate directory... the first message commands are done in the dealii directory. Sorry for the confusion. -- 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 --- Y

[deal.II] Re: dealii installation error - non-existent path when running "make test"

Also note that I have installed petsc using the following commands: ./configure --download-f-blaslapack=1 --with-shared-libraries=1 --download-hypre=1 --download-mumps=1 --download-spooles=1 --download-scalapack=1 --download-metis=1 --download-parmetis=1 --download-blacs=1 --with-debugging=0 -

Re: [deal.II] Re: Crack propagation

On 8/26/24 06:35, Junxiang Wang wrote: However, In all the tutorials steps of deal.II , the test function is indexed by "i" loop, right? For example, in the form of  local_matrix( i, j ). The difference is whether you multiply the equation from the left, or from the right, by a test funct

Re: [deal.II] Re: Crack propagation

Dear Prof. Thomas Wick, Thanks a lot for your answer. Now, I understand the purpose of your code. However, In all the tutorials steps of deal.II , the test function is indexed by "i" loop, right? For example, in the form of local_matrix( i, j ). Best regards, Junxiang. On Monday, August 26

Re: [deal.II] Re: Crack propagation

Dear Junxiang Wang, the reason is that we are dealing with a non-symmetric problem and the test function determines the row, while trial functions determine columns of the system matrix. The quickest and easiest way is to switch those indices to achieve this. In general, actually, the notatio

Re: [deal.II] Re: Crack propagation

Dear Timo Heister Thomas Wick, I have been reading your paper about phase-field crack propagation and your code recently. May I ask why your index of assembly subroutine has the form of local_matrix( j, i ) rather than the normal form of local_matrix( i, j )? Thanks a lots On Tuesday, Febru

[deal.II] Re: Lifting a Triangulation<2> to Triangulation<2,3>

Jasper, You don't want to use GridGenerator::hyper_ball. Instead try: Triangulation<2,3> my_triangulation; GridGenerator::hyper_shell(my_triangulation); Note that you need to use a triangulation where dim is different than spacedim. Best, Bruno On Monday, August 19, 2024 at 8:48:54 AM UTC-4

[deal.II] Re: dealii dmg for MacOS Monterey 12.7.4, clang version 14.0.0

I had a similar issue for almost 10 months (I was using remote station or VM to run dealii as an alternative, but they have their own restrictions). Finally, I was able to resolve it just yesterday. After Talking with Timo and Luca, I realized that Mac OS keeps different versions of updated pac

[deal.II] Re: Sparsity pattern building issue (dealii version 9.2.0)

Hello Marc, Thanks for that ! I forgot to account for the fact that I was printing the matrix to a text file naively using nested loops. I just came across the function print_formatted() in the SparseMatrix class which let's me accomplish what I was looking to do. Best Regards, Sudip Kunda On

[deal.II] Re: Sparsity pattern building issue (dealii version 9.2.0)

Hi Sudip, this is the only place in assemble_system() in which you call SparseMatrix::operator() /*std::cout << "Before applying boundary conditions" << std::endl;*/ for (unsigned int i = 0; i < dof_handler.n_dofs(); i++) { for (unsigned int j = 0; j < dof_handler.n_dofs(); j++) { text_ou

Re: [deal.II] Re: Does MatrixFree need to reinit when boundary_id changes?

Hi Martin, Thanks for the quick reply! Not very good news for me :-( since I need to change the boundary condition multiple times during evolution and many reinit is definitely expensive. Is there a better way that I can just change the face group and avoid reinit everything else in the MatrixF

Re: [deal.II] Re: Does MatrixFree need to reinit when boundary_id changes?

Dear Chengjiang, Yes, you need to run reinit() again when the boundary id on certain boundaries changes, because the boundary id gets extracted to the data structures of matrix-free face integrals, and faces of the same ID will get grouped together. Best, Martin Am 01.08.24 um 04:56 schrie

[deal.II] Re: Does MatrixFree need to reinit when boundary_id changes?

BTW, the changing boundary is not Dirichlet BC so that I do not need to rebuild my AffineConstraints. 在2024年7月30日星期二 UTC+8 10:41:34 写道： > Hi all, > > I need to change my boundary id on part of the mesh boundary when evolving > the equations. And I also have a boundary operator that acts on spec

[deal.II] Re: MPI and DataOut error

After digging into the source file, it looks like the error arises when the owned DoFs of a partition is empty. Some more digging looks like I had an off by one error while constructing the linear system which introduced an extra partitioning with range [0, 0) tacked on the end. This partition

[deal.II] Re: Speedup of the direct solver in parallel computing

If you are using PETScWrappers for MPI parallelization, there's a PETScWrappers::SparseDirectMUMPS that works with PETScWrappers out of the box. If you are familiar with PETSc or Trilinos, you can also manually choose MUMPS or SuperLU_DIST by tuning ksp/pc or amesos/amesos2 package. 在2024年7月19日

[deal.II] Re: Speedup of the direct solver in parallel computing

I am using UMFPACK and the code is parallelized based on MPI. Do you have a code example or tutorial to use MUMPS or SuperLU_DIST in deal.ii as you mentioned? Thank you so much for your help. Best, Giang On Thursday, July 18, 2024 at 10:22:16 AM UTC-4 avfa...@gmail.com wrote: > Are you using m

[deal.II] Re: Speedup of the direct solver in parallel computing

Are you using multi-threading or multi-processing? The bundled UMFPACK solver is single threaded, the theoretical speedup is about 1. I see 2 choices: (a) Use external multi-threaded direct solver like PARDISO, you can pass the reordered CSR data to it and (b) use MPI multi-processing and distr

[deal.II] Re: unit_gradient causing oscillations

Sorry I realized that image is unhelpful without what it is supposed to be. It should be a straight, flat line of zero values. Now attached is the same image but with the correct value plotted too as a black line. Best, Sean On Wednesday, July 17, 2024 at 4:38:12 PM UTC-6 Sean Johnson wrote: >

[deal.II] Re: More CutFEM examples

Hi, > Is there a similar tutorial for solving linear static analysis using CutFEM, especially with illustration of how to setup boundary conditions. There is no tutorial for this. You would set the boundary conditions weakly at the immersed boundaries. Here are two papers on linear elasticity w

  1   2   3   4   5   6   7   8   9   10   >