- firedrake - imperial.ac.uk

Re: [firedrake] BC's - 1D Hermite elements
by Lawrence Mitchell 07 Apr '20

07 Apr '20

[Adding firedrake back in to cc] > On 7 Apr 2020, at 22:25, Roberto Federico Ausas <rfausas(a)icmc.usp.br> wrote: > > File "/home/usuario/Codes/newfire/firedrake/src/firedrake/firedrake/functionspacedata.py", line 514, in lgmap > nodes.append(tmp + i) > TypeError: can only concatenate list (not "int") to list This is the proximate error. Your definition from Patrick has: class FixHermiteDerivativeValue(DirichletBC): @utils.cached_property def nodes(self): if V.mesh().mpi_comm().size > 1: raise NotImplementedError return [1] # the derivative component of the left endpoint I think it might work if you use from pyop2.datatypes import IntType and then return numpy.asarray([1], dtype=IntType) Thanks, Lawrence

1 0

Re: [firedrake] BC's - 1D Hermite elements
by Roberto Federico Ausas 07 Apr '20

07 Apr '20

Dear Patrick, Thanks for your help! It essentially worked, however, for some reason the nonlinear solver does not allow me to use direct methods as the inner solver, i.e., ('ksp_type': 'preonly', 'pc_type': 'lu', 'pc_factor_mat_solver_type': 'mumps') produces some sort of error, possibly related to local_to_global_map, I do not understand the error actually. Thanks anyway! On Tue, Apr 7, 2020 at 8:00 AM <firedrake-request(a)imperial.ac.uk> wrote: > Send firedrake mailing list submissions to > firedrake(a)imperial.ac.uk > > To subscribe or unsubscribe via the World Wide Web, visit > https://mailman.ic.ac.uk/mailman/listinfo/firedrake > or, via email, send a message with subject or body 'help' to > firedrake-request(a)imperial.ac.uk > > You can reach the person managing the list at > firedrake-owner(a)imperial.ac.uk > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of firedrake digest..." > > > Today's Topics: > > 1. BC's - 1D Hermite elements (Roberto Federico Ausas) > 2. Re: BC's - 1D Hermite elements (Patrick Farrell) > 3. Problem installing Firedrake (Smail Merabet) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Mon, 6 Apr 2020 16:49:25 -0300 > From: Roberto Federico Ausas <rfausas(a)icmc.usp.br> > To: firedrake(a)imperial.ac.uk > Subject: [firedrake] BC's - 1D Hermite elements > Message-ID: > <CAH6Ec4nWnhmM=HeMtd8uU9VQ9NF0_hVdorMFdK= > p8Xc3Gus4xQ(a)mail.gmail.com> > Content-Type: text/plain; charset="utf-8" > > Hello, > > My name is Roberto. I'm currently a professor in the Applied Mathematics > Department of Univ. of S?o Paulo at S?o Carlos. > > I'm working on some fluid-structure interaction problems in which the solid > will be modeled as a 1D beam. Our reference integration domain will be the > interval [0,L]. > > We are planning to use firedrake and Hermite elements. The problem is > nonlinear. > > I already have a working example. Values degrees of freedom are easy to > impose by using DirichletBC(), however I'm not figuring out how to enforce > derivatives degrees of freedom as boundary conditions (let say at x=0), > which I need for a clamped beam. > > Any help or trick would be appreciated > Regards > > -- > Roberto F. Ausas > Instituto de Ci?ncias Matem?ticas e de Computa??o - ICMC/USP > S?o Carlos, SP - Brasil > *http://www.lmacc.icmc.usp.br/~ausas* <http://www.lmacc.icmc.usp.br/~ausas > > >

2 1

Problem installing Firedrake
by Smail Merabet 07 Apr '20

07 Apr '20

Hi firedrake users, I have a problem with the installation of firedrake, I work on Mojave 10.14.6 I did the following steps: url -O https://raw.githubusercontent.com/firedrakeproject/firedrake/master/scripts… unset PYTHONPATH unset PETSC_DIR unset PETSC_ARCH; python3 firedrake-install the installation see to be done, I attached the log file to this email, but when I excuse any script I get the following error message (firedrake) (base) Merabets-MacBook-Pro:examples merabet$ python3 fi.py Traceback (most recent call last): File "/Users/merabet/firedrake/src/h5py/h5py/__init__.py", line 26, in <module> from . import _errors ImportError: dlopen(/Users/merabet/firedrake/src/h5py/h5py/_ errors.cpython-37m-darwin.so, 2): Library not loaded: @rpath/libhdf5.103.dylib Referenced from: /Users/merabet/firedrake/src/h5py/h5py/_ errors.cpython-37m-darwin.so Reason: image not found During handling of the above exception, another exception occurred: Traceback (most recent call last): File "fi.py", line 1, in <module> import firedrake File "/Users/merabet/firedrake/src/firedrake/firedrake/__init__.py", line 53, in <module> from firedrake.checkpointing import * File "/Users/merabet/firedrake/src/firedrake/firedrake/checkpointing.py", line 3, in <module> from firedrake.cython import hdf5interface as h5i File "firedrake/cython/hdf5interface.pyx", line 2, in init firedrake.cython.hdf5interface File "/Users/merabet/firedrake/src/h5py/h5py/__init__.py", line 30, in <module> raise ImportError("You cannot import h5py from inside the install directory.\nChange to another directory first.") ImportError: You cannot import h5py from inside the install directory. Change to another directory first. (firedrake) (base) Merabets-MacBook-Pro:examples merabet$ Please, could any one help me to solve this problem ? Best regards

1 0

BC's - 1D Hermite elements
by Roberto Federico Ausas 06 Apr '20

06 Apr '20

Hello, My name is Roberto. I'm currently a professor in the Applied Mathematics Department of Univ. of São Paulo at São Carlos. I'm working on some fluid-structure interaction problems in which the solid will be modeled as a 1D beam. Our reference integration domain will be the interval [0,L]. We are planning to use firedrake and Hermite elements. The problem is nonlinear. I already have a working example. Values degrees of freedom are easy to impose by using DirichletBC(), however I'm not figuring out how to enforce derivatives degrees of freedom as boundary conditions (let say at x=0), which I need for a clamped beam. Any help or trick would be appreciated Regards -- Roberto F. Ausas Instituto de Ciências Matemáticas e de Computação - ICMC/USP São Carlos, SP - Brasil *http://www.lmacc.icmc.usp.br/~ausas* <http://www.lmacc.icmc.usp.br/~ausas>

2 1

Help me
by Emmanuel Yarleque Medina 05 Apr '20

05 Apr '20

Hi users, I have a problem installing Firedrake with Petsc. Initially install using the following steps: curl -O https://raw.githubusercontent.com/firedrakeproject/firedrake/master/scripts… unset PYTHONPATH; unset PETSC_DIR; unset PETSC_ARCH; python3 firedrake-install and it works on installation, only when testing my code with the discontinuous Galerkin hybrid method (HDG) it gave the following error: (firedrake) hpc@hpc:~/Downloads/Exemplos_Firedrake/HDG/poroso/MWE/SDHM$ python3 sdhm_monolithic.py Traceback (most recent call last): File "sdhm_monolithic.py", line 1, in <module> from firedrake import * File "/home/hpc/firedrake/src/firedrake/firedrake/__init__.py", line 8, in <module> raise ImportError("PETSC_DIR is set, but you did not install with --honour-petsc-dir.\n" ImportError: PETSC_DIR is set, but you did not install with --honour-petsc-dir. Please unset PETSC_DIR (and PETSC_ARCH) before using Firedrake. try installing with python3 firedrake-install --honor-petsc-dir, but I have the following error: hpc @ hpc: ~ $ python3 firedrake-install --honor-petsc-dir Running firedrake-install --honor-petsc-dir The --honor-petsc-dir is set, but PETSC_DIR environment variable is not defined. If you have compiled PETSc manually, set PETSC_DIR (and optionally PETSC_ARCH) variables to point to the build directory. could you help me on how to solve that problem, please. Best regards, Emmanuel Y. Medina Ph.D Student in UFJF-Brazil

2 1

sprint to merge complex branch
by Sagiyama, Koki 01 Apr '20

01 Apr '20

Dear all, As we are ready to merge the `complex` branch to master except that we have yet to: * ensure type safety in each part of Firedrake, * reduce the size of tests to make complex tests run in a reasonable amount of time, we would like to run a sprint to have these efficiently done, each participant working on their favorite parts of Firedrake. We plan to have this sprint in the afternoons (UK time, GMT+1) of the week of 20th April on Webex. We will also be hosting a pre-sprint Webex meeting on Friday 17th April at 16:00 (UK time, GMT+1), to discuss the workflow. (We will provide the links to the sprint and the pre-sprint meeting on Slack http://firedrakeproject.slack.com beforehand.) I hope that many of you could attend and help us. If you could, please add your name to the list of participants on the github wiki page for this sprint: https://github.com/firedrakeproject/firedrake/wiki/MergeComplexSprint Any updates regarding this sprint will also be posted on this page. Regards, Koki

1 0

Re: [firedrake] Firedrake in a project at ATI
by Ham, David A 24 Mar '20

24 Mar '20

Dear Eric, Thanks for the email. Since you indicated that you tried to join the mailing list, I have taken the liberty of subscribing you to the list and copying the list in on my answer. We currently have what you want for the cases of point evaluation node elements (i.e. CG and DG elements). See: https://www.firedrakeproject.org/interpolation.html#interpolator-objects . The interpolator object contains the interpolation matrix in appropriate cases. There are people working on generalising this to all the elements we have, and making the differentiation work cleanly so that things like point forcing and point measurements work seamlessly. I hope we have more progress on this in the coming months. Cheers, David -- Dr David Ham Department of Mathematics Imperial College London https://www.imperial.ac.uk/people/david.ham From: Eric Daub <edaub(a)turing.ac.uk> Date: Wednesday, 11 March 2020 at 12:19 To: "Ham, David A" <david.ham(a)imperial.ac.uk> Subject: Firedrake in a project at ATI Dear David, I am an RSE at the Turing working with Mark Girolami (and Alastair Gregory before he moved to a new position) on implementing a probabilistic version of the Finite Element Method that he has developed over the past couple of years. The method provides a Bayesian framework for finding FEM solutions conditioned on sensor data, and Mark hopes that this tool will become a part of several research projects in the Turing Data-Centric Engineering Programme. Some other ICL researchers (Craig Buchanan and Leroy Gardner in Civil Engineering) are also involved with this research program. I am building our software on top of Firedrake as suggested by Alastair, as it nicely supports much of what we need to do straight out of the box (in particular, easily integrating with PETSc and ensemble parallelism have made my job relatively straightforward, as those have proven to be the difficult bits of the implementation thus far). However, one additional piece that is necessary for the FEM solves required by the method is the ability to access the matrix that interpolates between the FEM DOF nodes and arbitrary sensor locations where data has been collected. This involves evaluating the FEM basis functions at all of the sensor locations and collecting them in a sparse matrix. I did not see a simple way to do this using the existing interfaces, but it is highly likely that I missed something. I have a simplified version that works for linear CG elements on triangular/tetrahedral meshes that I have been using for testing purposes, but eventually want to implement this for the more general case. Any advice on how to proceed would be greatly appreciated. I attempted to join the Firedrake listserver, as I suspect that some of the discussions arising there may be relevant to my work, though it seems that my request never went through. If I can join the list I would be happy to re-pose my question there for the development team if you feel that would be more appropriate. Thanks for your help, and I hope that my work can help get more people using Firedrake in various research threads here at the Turing. Cheers, Eric The Alan Turing Institute is a limited liability company, registered in England with registered number 09512457. Our registered office is at British Library, 96 Euston Road, London, England, NW1 2DB. We are also a charity registered in England with charity number 1162533. DISCLAIMER: Although The Alan Turing Institute has taken reasonable precautions to ensure no viruses are present in this email, The Alan Turing Institute cannot accept responsibility for any loss or damage sustained as a result of computer viruses and the recipient must ensure that the email (and attachments) are virus free. While we take care to protect our systems from virus attacks and other harmful events, we give no warranty that this message (including attachments) is free of any virus or other harmful matter, and we accept no responsibility for any loss or damage resulting from the recipient receiving, opening or using it. E-mail transmission cannot be guaranteed to be secure or error-free as information could be intercepted, corrupted, lost, destroyed, arrive late or be incomplete. If you think someone may have interfered with this email, please contact the Alan Turing Institute by telephone only and speak to the person dealing with your matter or the Accounts Department. Fraudsters are increasingly targeting organisations and their affiliates, often requesting funds to be transferred to a different bank account. The Alan Turing’s bank details are contained within our terms of engagement. If you receive a suspicious or unexpected email from us, or purporting to have been sent on our behalf, particularly containing different bank details, please do not reply to the email, click on any links, open any attachments, or comply with any instructions contained within it, but contact our Accounts department by telephone. Our Transparency Notice found here - https://www.turing.ac.uk/transparency-notice sets out how and why we collect, store, use and share your personal data and it explains your rights and how to raise concerns with us.

2 1

Returning eigenvectors of a solution to a PDE at each node
by Andrew Hicks 20 Mar '20

20 Mar '20

Hello, I am currently using firedrake to solve a PDE. The solution to this PDE is vector-valued, and I have no problem finding the solution. However, I'd like to go a step further. I have named the solution "q_soln" and this is a 2-dimesional vector. I have used this vector to construct a 2x2 matrix at the node [0.5,0.5]. I then extracted an eigenvector from this matrix, and then I printed it: import numpy as np from numpy import linalg as la E1 = (1 / np.sqrt(2)) * np.array([[1,0],[0,-1]]) E2 = (1 / np.sqrt(2)) * np.array([[0,1],[1,0]]) # as an example, we evaluate the solution at the point [0.5,0.5], and pick the first eigenvector q_soln_atpoint = q_soln.at([0.5,0.5]) Q_soln = q_soln_atpoint[0] * E1 + q_soln_atpoint[1] * E2 evals, evecs = la.eig(Q_soln) # evec is a 2x2 array with the columns being the e-vectors q_soln_evec = np.zeros((1,2)) q_soln_evec += evecs[0,:] print(q_soln_evec) My question is this: how can I get firedrake to do this automatically for every node in my mesh, and then store it in as a Function "q_soln_evec"? The only reason I used numpy to get this one point is because I'm unfamiliar with firedrake. I bet firedrake has some built-in way of doing what I need here. Thanks, Andrew Hicks

3 2

Getting on mailing list
by Andrew Hicks 19 Mar '20

19 Mar '20

I cannot seem to get on the mailing list. I entered my email (ahick17(a)lsu.edu<mailto:ahick17@lsu.edu>) into the system two different times, expecting a confirmation email, but never received one. I am now attempting to email the list directly. Andrew Hicks

2 1

Solving Decoupled Problems over Elements of Mesh of Global Problem
by Austin Kaul 12 Mar '20

12 Mar '20

Hello, I am working on developing a method where optimal discontinous test functions are computed for the global problem. The issue is that these optimal test function computations are decoupled from the global problem and are over each element individually. Has this been done or is there a way with par_loop() to perform a solve over each element of a mesh and have its solutions contribute to the global solve at the end? Each element problem has its own variational statement with its own test functions which we would define. Using these results we would construct the global stiffness matrix(function) as we solve this optimal test function problem over each element. We have implemented a very simple version of this with an in house code but your framework allows for so much flexibility we would love to be able to do this in Firedrake. Regards, -- Austin Kaul

2 1