November 2014 - firedrake - imperial.ac.uk

Re: [firedrake] Error - NonlinearVariationalSolver
by Lawrence Mitchell 10 Nov '14

10 Nov '14

On 10 Nov 2014, at 18:37, Lawrence Mitchell <lawrence.mitchell(a)imperial.ac.uk> wrote: > Dear Anna, > > On 10 Nov 2014, at 18:13, Anna Kalogirou <a.kalogirou(a)leeds.ac.uk> wrote: > >> Dear all, >> >> I wrote the attached simple code to solve a shallow water problem with a nonlinear solver, but I get the following error (note that the linear problem works fine). Can someone help me and see what I am doing wrong? Thanks you very much! > > > Your residual Fphi is written in terms of phi, however in your nonlinear problem, you ask to linearise around phi1. > > Fphi = (gamma*phi + 0.5*mu*inner(grad(gamma),grad(phi)) > - (gamma*phi0 + 0.5*mu*inner(grad(gamma),grad(phi0)) > + dt*gamma*eta0))*dx > phi_problem = NonlinearVariationalProblem(Fphi,phi1) > > I think you probably want: > > phi_problem = NonlinearVariationalProblem(Fphi, phi) > > Note that, compared to the LinearVariationalProblem, where the solution is placed in a separate output Function, for the nonlinear problem, the solution variable must be the Function you want to solve for in the residual. > > I think there are similar changes required for the eta and laplace solvers. > > Cheers, > > Lawrence

1 0

Error - NonlinearVariationalSolver
by Anna Kalogirou 10 Nov '14

10 Nov '14

Dear all, I wrote the attached simple code to solve a shallow water problem with a nonlinear solver, but I get the following error (note that the linear problem works fine). Can someone help me and see what I am doing wrong? Thanks you very much! Kind regards, Anna Kalogirou. =================================== Traceback (most recent call last): File "nonlinear_swe.py", line 43, in <module> phi_solver = NonlinearVariationalSolver(phi_problem) File "/opt/firedrake/2.0/lib/python2.7/site-packages/firedrake/solving.py", line 123, in __init__ self._jac_tensor = assemble(self._problem.J_ufl, bcs=self._problem.bcs) File "/opt/firedrake/2.0/lib/python2.7/site-packages/firedrake/solving.py", line 420, in assemble return _assemble(f, tensor=tensor, bcs=_extract_bcs(bcs)) File "/opt/firedrake/2.0/lib/python2.7/site-packages/firedrake/solving.py", line 439, in _assemble kernels = ffc_interface.compile_form(f, "form") File "/opt/firedrake/2.0/lib/python2.7/site-packages/firedrake/ffc_interface.py", line 235, in compile_form FFCKernel(form, name).kernels)] File "/opt/firedrake/2.0/lib/python2.7/site-packages/PyOP2-0.11.0_117_gcfe592d-py2.7-linux-x86_64.egg/pyop2/caching.py", line 202, in __new__ obj = make_obj() File "/opt/firedrake/2.0/lib/python2.7/site-packages/PyOP2-0.11.0_117_gcfe592d-py2.7-linux-x86_64.egg/pyop2/caching.py", line 192, in make_obj obj.__init__(*args, **kwargs) File "/opt/firedrake/2.0/lib/python2.7/site-packages/firedrake/ffc_interface.py", line 181, in __init__ ffc_tree = ffc_compile_form(form, prefix=name, parameters=parameters["form_compiler"]) File "/opt/firedrake/2.0/lib/python2.7/site-packages/ffc/compiler.py", line 151, in compile_form analysis = analyze_forms(forms, parameters) File "/opt/firedrake/2.0/lib/python2.7/site-packages/ffc/analysis.py", line 57, in analyze_forms parameters) for form in forms) File "/opt/firedrake/2.0/lib/python2.7/site-packages/ffc/analysis.py", line 57, in <genexpr> parameters) for form in forms) File "/opt/firedrake/2.0/lib/python2.7/site-packages/ffc/analysis.py", line 136, in _analyze_form _attach_integral_metadata(form_data, parameters) File "/opt/firedrake/2.0/lib/python2.7/site-packages/ffc/analysis.py", line 266, in _attach_integral_metadata scheme = quad_schemes[0] IndexError: list index out of range =================================== -- Dr Anna Kalogirou Research Fellow School of Mathematics University of Leeds

1 0

Re: [firedrake] printing stuff
by David Ham 08 Nov '14

08 Nov '14

You can use the firedrake.op2.MPI.rank_zero decorator to make a function which will only execute on rank 0. On 8 November 2014 18:53, Cotter, Colin J <colin.cotter(a)imperial.ac.uk> wrote: > Dear Firedrakers, > What's the best way to modify a print statement so that only one MPI > process does the print? > > cheers > --cjc > -- Dr David Ham Departments of Mathematics and Computing Imperial College London http://www.imperial.ac.uk/people/david.ham

2 1

printing stuff
by Cotter, Colin J 08 Nov '14

08 Nov '14

Dear Firedrakers, What's the best way to modify a print statement so that only one MPI process does the print? cheers --cjc

1 0

Iteration space in local mass matrix assembly
by Eike Mueller 06 Nov '14

06 Nov '14

Dear firedrakers, when I store the local DG mass matrix in a PyOP2 dat of 3x3 matrices by hand, why do I need to specify the op2.i[0] iteration space as in here V = FunctionSpace(mesh,'DG',1) V_DG0 = FunctionSpace(mesh,'DG',0) u = TestFunction(V) v = TrialFunction(V) mass = u*v*dx mass_kernel = compile_form(mass, 'mass')[0][6] mass_matrix = Function(V_DG0, val=op2.Dat(V_DG0.node_set**(3*3))) op2.par_loop(mass_kernel, mass_matrix.cell_set, mass_matrix.dat(op2.INC, mass_matrix.cell_node_map()[op2.i[0]]), mesh.coordinates.dat(op2.READ,mesh.coordinates.cell_node_map(),flatten=True), mesh.coordinates.dat(op2.READ,mesh.coordinates.cell_node_map(),flatten=True)) When I print out mass_kernel.code I get static inline void mass_cell_integral_0_otherwise (double A[3][3] , double **vertex_coordinates , double **w0 ) { […] } i.e. on each cell the kernel gets passed a 3x3 matrix (= array with 9 elements). If I leave out the "[op2.i[0]]” I only get zeros in mass_matrix. Also, why do the coordinates have to be passed with flatten=True keyword? Thanks a lot, Eike -- Dr Eike Hermann Mueller Research Associate (PostDoc) Department of Mathematical Sciences University of Bath Bath BA2 7AY, United Kingdom +44 1225 38 5803 e.mueller(a)bath.ac.uk http://people.bath.ac.uk/em459/

2 1

petition: "software is fundamental to research"
by Eike Mueller 01 Nov '14

01 Nov '14

Dear firedrakers, apologies if you have already received this, but this was sent round on our Bath HPC list and I thought it might be of interest: I just signed the petition "Everyone in the research community: We must accept that software is fundamental to research, or we will lose our ability to make groundbreaking discoveries." on Change.org <http://change.org/>. It's important. Will you sign it too? Here's the link: http://www.change.org/p/everyone-in-the-research-community-we-must-accept-t… <http://www.change.org/p/everyone-in-the-research-community-we-must-accept-t…> Thanks! Eike -- Dr Eike Hermann Mueller Research Associate (PostDoc) Department of Mathematical Sciences University of Bath Bath BA2 7AY, United Kingdom +44 1225 38 5803 e.mueller(a)bath.ac.uk http://people.bath.ac.uk/em459/

1 0