March 2017 - firedrake - imperial.ac.uk

Deactivate UFL Warning
by Buesing, Henrik 23 Mar '17

23 Mar '17

Dear all, I am getting a lot of UFL:WARNING Discontinuous Lagrange element requested on quadrilateral * interval, creating DQ element Should I care? How can I deactivate these warnings? Thank you! Henrik -- Dipl.-Math. Henrik Büsing Institute for Applied Geophysics and Geothermal Energy E.ON Energy Research Center RWTH Aachen University ------------------------------------------------------ Mathieustr. 10 | Tel +49 (0)241 80 49907 52074 Aachen, Germany | Fax +49 (0)241 80 49889 ------------------------------------------------------ http://www.eonerc.rwth-aachen.de/GGE hbuesing(a)eonerc.rwth-aachen.de ------------------------------------------------------

2 6

NonLinearVariationalProblem: How to say nest=False?
by Buesing, Henrik 22 Mar '17

22 Mar '17

Dear all, In November 2015 I did problem = NonlinearVariationalProblem(F,u,nest=False,bcs=[bc0, bc1, bc2, bc3, bc4, bc5]) This is not the correct syntax anymore. How do I say nest=False today? Thank you! Henrik -- Dipl.-Math. Henrik Büsing Institute for Applied Geophysics and Geothermal Energy E.ON Energy Research Center RWTH Aachen University ------------------------------------------------------ Mathieustr. 10 | Tel +49 (0)241 80 49907 52074 Aachen, Germany | Fax +49 (0)241 80 49889 ------------------------------------------------------ http://www.eonerc.rwth-aachen.de/GGE hbuesing(a)eonerc.rwth-aachen.de ------------------------------------------------------

2 2

Re: [firedrake] NonLinearVariationalProblem: How to say nest=False?
by Andrew McRae 22 Mar '17

22 Mar '17

Per http://www.firedrakeproject.org/solving-interface.html, set "mat_type": "aij" in the solver_parameters dictionary. On 22 March 2017 at 16:05, Buesing, Henrik <HBuesing(a)eonerc.rwth-aachen.de> wrote: > Dear all, > > > > In November 2015 I did > > > > *problem = NonlinearVariationalProblem(F,u,nest=False,bcs=[bc0, bc1, bc2, > bc3, bc4, bc5])* > > > > This is not the correct syntax anymore. How do I say nest=False today? > > Thank you! > > Henrik > > > > > > > > -- > > Dipl.-Math. Henrik Büsing > > Institute for Applied Geophysics and Geothermal Energy > > E.ON Energy Research Center > > RWTH Aachen University > > ------------------------------------------------------ > > Mathieustr. 10 | Tel +49 (0)241 80 49907 > <+49%20241%208049907> > > 52074 Aachen, Germany | Fax +49 (0)241 80 49889 > <+49%20241%208049889> > > ------------------------------------------------------ > > http://www.eonerc.rwth-aachen.de/GGE > > hbuesing(a)eonerc.rwth-aachen.de > > ------------------------------------------------------ > > >

1 0

Guidelines for using/running jobs on larger HPC machines
by Justin Chang 22 Mar '17

22 Mar '17

Hi all, Generally speaking, what should the PYOP2/FIREDRAKE_TSFC_KERNEL_CACHE_DIR environment variables be set to? Is it sufficient to have something like: export PYOP2/FIREDRAKE_TSFC_KERNEL_CACHE_DIR=$HOME/fd-cache This works on my smaller university cluster but I wonder if for a system like Edison at NERSC ( http://www.nersc.gov/users/computational-systems/edison/) if there is a better directory for this. Also, from what I read, Edison's SLURM scheduler loads the executable to the allocated compute nodes from the current working directory, which supposedly can be really slow - they recommend using something like: srun --bcast=/tmp/$SLURM_JOB_ID --compress=lz4 ... if 2000 or more nodes are needed. But even on jobs that require no more than a single compute node (24 cores), the firedrake/python modules seem to load very slowly. This is the description of the file storage system on Edison ( http://www.nersc.gov/users/computational-systems/edison/file-storage-and-i-…) btw Any help or thoughts appreciated. Justin

2 1

Re: [firedrake] firedrake-install on macOS Sierra
by David Ham 21 Mar '17

21 Mar '17

I think we've seen this occasionally on new macs. I think it's caused by the virtualenv getting installed in a different python installation from the one which is being used to install Firedrake. Try the following: Open a new terminal (this is to try to avoid any bash path caching effects) Type "which python". The answer you want is /usr/local/bin/python run python and check that you can import virtualenv If all the above works, try running firedrake-install in the new shell. If anything above fails, tell us. Regards, David On Tue, 21 Mar 2017 at 10:28 Anna Kalogirou <A.Kalogirou(a)leeds.ac.uk> wrote: > Dear all, > > > I am installing firedrake on my new laptop from scratch (running macOS > Sierra). > > > The installation script runs and installs virtualenv, up to the point > where I get this: > > Virtual env installed. Please run firedrake-install again. > > > But when I run > > python firedrake-install > > again, it just does the same thing (similar to the previous step) and does > not install firedrake (and petsc etc..). What shall I do? > > > Best, Anna. > -- Dr David Ham Department of Mathematics Imperial College London

2 2

firedrake-install on macOS Sierra
by Anna Kalogirou 21 Mar '17

21 Mar '17

Dear all, I am installing firedrake on my new laptop from scratch (running macOS Sierra). The installation script runs and installs virtualenv, up to the point where I get this: Virtual env installed. Please run firedrake-install again. But when I run python firedrake-install again, it just does the same thing (similar to the previous step) and does not install firedrake (and petsc etc..). What shall I do? Best, Anna.

1 0

Re: [firedrake] circles in firedrake?
by David Ham 21 Mar '17

21 Mar '17

Hi Francis, There appears to be something called vext.vtk which claims to make system vtk accessible from inside a virtualenv. There is also this email: http://vtk.1045678.n5.nabble.com/Re-VTK-in-virtualenv-td5716725.html I have to confess that I have not really tried either. Regards, David On Mon, 20 Mar 2017 at 01:40 Francis Poulin <fpoulin(a)uwaterloo.ca> wrote: > > Thanks for suggesting PyGmsh! I bumped into it a while ago but never > really figured it out. Now I'm keen since it should help building and > modifying meshes more easily. > > I installed pygmsh easily using the instructions on the website. It > imports correctly. There is even an example that builds a circle, I think. > > When I try running this, as well as other examples, I get an error because > no module named vtk is found. See below. I get this comes from meshio? > > Traceback (most recent call last): > > File "circle.py", line 24, in <module> > meshio.write('circle.vtu', *out) > File > "/home/fpoulin/software/firedrake/local/lib/python2.7/site-packages/meshio/__init__.py", > line 187, in write > field_data=field_data > File > "/home/fpoulin/software/firedrake/local/lib/python2.7/site-packages/meshio/vtk_io.py", > line 204, in write > import vtk > ImportError: No module named vtk > > When I do a pip install it tells me that it is already installed and up to > date. > > > (firedrake) fpoulin@fpoulin-Gazelle:~/software/pygmsh/test/examples$ pip > install -U meshio > Requirement already up-to-date: meshio in > /home/fpoulin/software/firedrake/lib/python2.7/site-packages > Requirement already up-to-date: pipdated in > /home/fpoulin/software/firedrake/lib/python2.7/site-packages (from meshio) > Requirement already up-to-date: numpy in > /home/fpoulin/software/firedrake/lib/python2.7/site-packages (from meshio) > Requirement already up-to-date: requests in > /home/fpoulin/software/firedrake/lib/python2.7/site-packages (from > pipdated->meshio) > Requirement already up-to-date: appdirs in > /home/fpoulin/software/firedrake/lib/python2.7/site-packages (from > pipdated->meshio) > > > I realize this isn't necessarily a firedrake question but if anyone had an > idea on how to get this working I would be keen to try it. > > > Cheers, Francis > ------------------ > Francis Poulin > Associate Professor > Department of Applied Mathematics > University of Waterloo > > email: fpoulin(a)uwaterloo.ca > Web: https://uwaterloo.ca/poulin-research-group/ > Telephone: +1 519 888 4567 x32637 <(519)%20888-4567> > > ------------------------------ > *From:* firedrake-bounces(a)imperial.ac.uk [firedrake-bounces(a)imperial.ac.uk] > on behalf of G. D. McBain [gdmcbain(a)protonmail.com] > *Sent:* Sunday, March 19, 2017 6:16 PM > > *To:* firedrake(a)imperial.ac.uk > *Subject:* Re: [firedrake] circles in firedrake? > > I would expect a user to run gmsh separately (at the command line or > through the GUI), rather than from within the Python script. I don't think > it's possible to interact with gmsh from Python nowadays without jumping > through considerable hoops. > > > Have you tried the (free third-party) PyGmsh? It generates .geo code from > Python. From there one calls Gmsh using Python subprocess and reads in the > resulting .msh. I've found it handy. > > https://github.com/nschloe/pygmsh > > -- Dr David Ham Department of Mathematics Imperial College London

3 2

Mismatching function spaces error
by Anna Kalogirou 20 Mar '17

20 Mar '17

Dear all, I am trying to solve a set of PDEs and ODEs together as a mixed system. The ODE variables only depend on time, so I define them as functions from the real function space R. First question: Is this a correct approach? Second question: I am running into a “Mismatching function spaces” error; can someone please take a look at my code (found here<https://bitbucket.org/annakalog/buoy2d/src/957e88739223e11d84826c76923a9b07…>) and let me know what I’m doing wrong? The error arises when the value of one of the mixed-system variables is assigned to another variable which belongs to the same function space. [I remember having the same issue some time ago, but I thought that it was fixed.] Thank you. Best, Anna.

3 6

Re: [firedrake] Mismatching function spaces error
by Andrew McRae 20 Mar '17

20 Mar '17

Good spot, this is the split(foo) vs foo.split() "gotcha". split(foo) is supposed to give symbolic UFL objects that can be used in forms: U = FunctionSpace(...); V = FunctionSpace(...) W = U*V urho = TrialFunction(W) u, rho = split(urho) # could have done u, rho = TrialFunctions(W) a = (u*.... + div(u)*.... + rho*...)*dx Whilst foo.split() splits a Function into subfunctions for further use (assignation, output, etc) urho_out = Function(W) solve(a == L, urho_out, bcs=...) u_out, rho_out = urho_out.split() File("temp.pvd").write(rho_out) Furthermore, these somehow work interchangeably *most of the time*, but not *all of the time*. Apologies. On 20 March 2017 at 07:48, Floriane Gidel [RPG] <mmfg(a)leeds.ac.uk> wrote: > > Sorry, > Z0.assign(Z1_out) > ------------------------------ > *De :* Floriane Gidel [RPG] > *Envoyé :* lundi 20 mars 2017 07:47:28 > *À :* firedrake(a)imperial.ac.uk > *Objet :* RE: Mismatching function spaces error > > Hi Anna, > > I think you need to write something like > > Phi_out, eta_out, mu_out, Z1_out, W1_out, I_out = w.split() > Z0.assign(Z0_out) > > Best, > Floriane > ------------------------------ > *De :* firedrake-bounces(a)imperial.ac.uk <firedrake-bounces(a)imperial.ac.uk> > de la part de Anna Kalogirou <A.Kalogirou(a)leeds.ac.uk> > *Envoyé :* dimanche 19 mars 2017 22:18:38 > *À :* firedrake(a)imperial.ac.uk > *Objet :* [firedrake] Mismatching function spaces error > > Dear all, > > I am defining a mixed system as follows: > > V = FunctionSpace(mesh, "CG", 1) > V_R = FunctionSpace(mesh, "R", 0) > W = V * V * V * V_R * V_R * V_R > > and some of the functions defined are: > > Z0 = Function(V_R) > w = Function(W) > phi, eta, mu, Z1, W1, I = split(w) > > After defining and solving some solvers etc, I do the following assignment > but run into a “Mismatching function spaces” error: > Z0.assign(Z1) > > Variable Z1 is solved in a mixed system and I want to assign it to Z0, > which belongs to space V_R. > Shouldn’t the assignment line above work? Why am I getting the error? > > Thank you, > > Anna. > > > On 19 Feb 2017, at 23:09, Anna Kalogirou <a.kalogirou(a)leeds.ac.uk> wrote: > > Dear all, > > I am trying to solve a set of PDEs and ODEs together as a mixed system. > The ODE variables only depend on time, so I define them as functions from > the real function space R. > > First question: Is this a correct approach? > > Second question: I am running into a “Mismatching function spaces” error; > can someone please take a look at my code (found here > <https://bitbucket.org/annakalog/buoy2d/src/957e88739223e11d84826c76923a9b07…>) > and let me know what I’m doing wrong? The error arises when the value of > one of the mixed-system variables is assigned to another variable which > belongs to the same function space. [I remember having the same issue some > time ago, but I thought that it was fixed.] > > Thank you. > > Best, Anna. > _______________________________________________ > firedrake mailing list > firedrake(a)imperial.ac.uk > https://mailman.ic.ac.uk/mailman/listinfo/firedrake > > >

1 0

Re: [firedrake] circles in firedrake?
by Andrew McRae 20 Mar '17

20 Mar '17

As in, the gmsh input code that we used to create the circle mesh is being deleted in that commit, at lines 456-463. You could copy those removed lines into a .geo file and run it through gmsh to get the mesh that was being created before (but it has the cell area -> 0 problem). I would expect a user to run gmsh separately (at the command line or through the GUI), rather than from within the Python script. I don't think it's possible to interact with gmsh from Python nowadays without jumping through considerable hoops. Not sure why you're getting that error. It's been a long time since I last used gmsh. Taking a guess from https://github.com/FluidityProject/fluidity/wiki/Gmsh-tutorial, try adding one more line to the .geo: Physical Surface(7) = {6}; On 19 March 2017 at 16:04, Francis Poulin <fpoulin(a)uwaterloo.ca> wrote: > Thank you Andrew for your help. I am a bit sorry to see CircleMesh go > away but if it only gave a structured mesh, then I can see the advantage to > using gmsh. Also, I should really learn how to use gmsh better, so this > will force me to do just that. > > I looked around for circle.geo but couldn't find it in the repo. > > I found another version of circle.geo that seems to build a circle. It is > copied here > > L = 10; > dx = 1; > > Point(1) = { 0, 0, 0, dx}; > Point(2) = { L, 0, 0, dx}; > Point(3) = {-L, 0, 0, dx}; > Point(4) = { 0, L, 0, dx}; > Point(5) = { 0,-L, 0, dx}; > > Circle(1) = {3,1,5}; > Circle(2) = {5,1,2}; > Circle(3) = {2,1,4}; > Circle(4) = {4,1,3}; > > Line Loop(5) = {1,2,3,4}; > > Plane Surface(6) = {5}; > > I am not sure why the origin needs to be included but it does give the > correct shape at least. > > Unfortunately, when I try importing this in firedrake using > > from firedrake import * > > #gmsh -2 circle.geo > > mesh = Mesh("circle.msh") > > > I get the error > > Traceback (most recent call last): > File "test.py", line 7, in <module> > mesh = Mesh("circle.msh") > File "<decorator-gen-261>", line 2, in Mesh > File "/home/fpoulin/software/firedrake/src/PyOP2/pyop2/profiling.py", > line 61, in wrapper > return f(*args, **kwargs) > File "/home/fpoulin/software/firedrake/src/firedrake/firedrake/mesh.py", > line 1259, in Mesh > topology = MeshTopology(plex, name=name, reorder=reorder, > distribute=distribute) > File "<decorator-gen-259>", line 2, in __init__ > File "/home/fpoulin/software/firedrake/src/PyOP2/pyop2/profiling.py", > line 61, in wrapper > return f(*args, **kwargs) > File "/home/fpoulin/software/firedrake/src/firedrake/firedrake/mesh.py", > line 348, in __init__ > dmplex.validate_mesh(plex) > File "firedrake/dmplex.pyx", line 1009, in > firedrake.dmplex.validate_mesh (firedrake/dmplex.c:14063) > ValueError: Provided mesh has some entities not reachable by traversing > cells (maybe rogue vertices?) > > > Can you point out what I'm doing wrong? Or where to find the circle.geo? > > Also, I tried use gmsh in firedrake as commented above and that didn't > work. Should gmsh work in the latest version? > > Cheers, Francis > > ------------------ > Francis Poulin > Associate Professor > Department of Applied Mathematics > University of Waterloo > > email: fpoulin(a)uwaterloo.ca > Web: https://uwaterloo.ca/poulin-research-group/ > Telephone: +1 519 888 4567 x32637 <(519)%20888-4567> > > ------------------------------ > *From:* firedrake-bounces(a)imperial.ac.uk [firedrake-bounces(a)imperial.ac.uk] > on behalf of Andrew McRae [A.T.T.McRae(a)bath.ac.uk] > *Sent:* Sunday, March 19, 2017 8:07 AM > *To:* firedrake(a)imperial.ac.uk > *Subject:* Re: [firedrake] circles in firedrake? > > We used to have a CircleMesh in Firedrake, but it was removed in this > commit > <https://github.com/firedrakeproject/firedrake/commit/54bff6b70e046106ddd4e7…>. > If I recall correctly, this is because we outsourced the creation of that > mesh to gmsh, but the gmsh-from-Python packages are no longer available > with newer versions of Ubuntu. > > That commit has a small gmsh snippet (.geo). You could manually run this > through gmsh to build a mesh (.msh), and load this in to your firedrake > script: mesh = Mesh("circle.msh"). Note that this CircleMesh (and, IIRC, > FEniCS's equivalent) give a "structured" mesh with radial lines, and > correspondingly small cells in the middle. If this is undesirable, you > might be better off making a simple unstructured circle mesh in gmsh and > loading it in. > > Regarding the code you gave in your email, I think this has also worked > just fine. However, the built-in Firedrake plotting functionality only > seems to support 2D triangle and "unstructured" quadrilateral cells, rather > than the "interval x interval" cells created by an extruded mesh. (I > suspect this was an accidental omission) > > If you instead do File("temp.pvd").write(U) and open the result in > Paraview, I'm sure it would look fine. Of course, if you use this to get a > circular mesh, it will also have small cells at the centre. > > Best, > Andrew > > On 18 March 2017 at 17:29, Francis Poulin <fpoulin(a)uwaterloo.ca> wrote: > >> Hello, >> >> I want to define my mesh to be a circle. In Fenics I know there is a >> command called CircleMesh that makes this very easy. >> >> Q1) Is there a simple way of doing this in Firedrake? >> >> I looked around and see how circlemanifoldmesh can do this with help for >> ExtrudedMesh. I found some code that can make a ring and I tried running >> it myself and defining a function on that mesh. Unfortunately, I get an >> error, see below. >> >> Q2) If one needs to use CircleManifoldMesh is there something else one >> needs to do to use a circular geometry as your mesh? >> >> Below are the lines that I tried and the error >> >>> from firedrake import * >> >>> m = CircleManifoldMesh(20, radius=2) >> >>> mesh = ExtrudedMesh(m, 5, extrusion_type='radial') >> >>> x = SpatialCoordinate(mesh) >> >>> V = FunctionSpace(mesh,'CG',1) >> >>> U = Function(V).interpolate(-(pow(x[0]-0.5,2) + pow(x[1]-0.5,2))) >> >>> plot(U); >> Traceback (most recent call last): >> File "<stdin>", line 1, in <module> >> File "/home/fpoulin/software/firedrake/src/firedrake/firedrake/plot.py", >> line 308, in plot >> **kwargs) >> File "/home/fpoulin/software/firedrake/src/firedrake/firedrake/plot.py", >> line 502, in two_dimension_plot >> num_sample_points) >> File "/home/fpoulin/software/firedrake/src/firedrake/firedrake/plot.py", >> line 457, in _two_dimension_triangle_func_val >> raise RuntimeError("Unsupported Functionality") >> RuntimeError: Unsupported Functionality >> >> >> ------------------ >> Francis Poulin >> Associate Professor >> Department of Applied Mathematics >> University of Waterloo >> >> email: fpoulin(a)uwaterloo.ca >> Web: https://uwaterloo.ca/poulin-research-group/ >> Telephone: +1 519 888 4567 x32637 <(519)%20888-4567> >> >> >

3 2