dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Tue, Oct 06, 2009 at 01:05:01PM +0200, Kristian Oelgaard wrote:
> Quoting Anders Logg <logg@xxxxxxxxx>:
>
> > On Tue, Oct 06, 2009 at 10:32:12AM +0200, Kristian Oelgaard wrote:
> > > Quoting Anders Logg <logg@xxxxxxxxx>:
> > >
> > > > On Tue, Oct 06, 2009 at 09:43:55AM +0200, Kristian Oelgaard wrote:
> > > > > Quoting Anders Logg <logg@xxxxxxxxx>:
> > > > >
> > > > > > On Mon, Oct 05, 2009 at 11:18:19AM +0200, Kristian Oelgaard wrote:
> > > > > > >
> > > > > > > Hi,
> > > > > > >
> > > > > > > I'm trying to update my plasticity solver to use the latest version
> > of
> > > > > > DOLFIN
> > > > > > > and I ran into a problem with the new design of Function/Expression
> > > > etc.
> > > > > > > I get the following error:
> > > > > > >
> > > > > > > terminate called after throwing an instance of 'std::runtime_error'
> > > > > > >   what():  evaluate_basis() is not supported for QuadratureElement
> > > > > > >
> > > > > > > Program received signal SIGABRT, Aborted.
> > > > > > > [Switching to Thread 0xb6a436d0 (LWP 21130)]
> > > > > > > 0xb7f15430 in __kernel_vsyscall ()
> > > > > > > (gdb) where
> > > > > > > #0  0xb7f15430 in __kernel_vsyscall ()
> > > > > > > #1  0xb6b726d0 in raise () from /lib/tls/i686/cmov/libc.so.6
> > > > > > > #2  0xb6b74098 in abort () from /lib/tls/i686/cmov/libc.so.6
> > > > > > > #3  0xb6d8f8f8 in __gnu_cxx::__verbose_terminate_handler () from
> > > > > > > /usr/lib/libstdc++.so.6
> > > > > > > #4  0xb6d8d7d5 in ?? () from /usr/lib/libstdc++.so.6
> > > > > > > #5  0xb6d8d812 in std::terminate () from /usr/lib/libstdc++.so.6
> > > > > > > #6  0xb6d8d94a in __cxa_throw () from /usr/lib/libstdc++.so.6
> > > > > > > #7  0x080d6b39 in strain2d_0_finite_element_0::evaluate_basis
> > > > > > (this=0x9ba9038,
> > > > > > > i=0, values=0x9bad220, coordinates=0xbfb2f3c8, c=@0xbfb2f5d8)
> > > > > > >     at
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> /home/oelgaard/fenics_projects/fenics_apps/fenics-plasticity/demo/forms/2Dlinear_QE_forms/Strain2D.cpp:495
> > > > > > > #8  0xb7cf2d55 in dolfin::Function::eval (this=0xbfb306b4,
> > > > > > values=0xbfb2f3b0,
> > > > > > > x=0xbfb2f3c8, ufc_cell=@0xbfb2f5d8, cell_index=0) at
> > > > > > > ./dolfin/fem/FiniteElement.h:53
> > > > > > > #9  0xb7cf9e77 in dolfin::Function::eval (this=0xbfb306b4,
> > > > > > values=0xbfb2f3b0,
> > > > > > > data=@0xbfb306fc) at dolfin/function/Function.cpp:275
> > > > > > > #10 0xb7d00c77 in dolfin::GenericFunction::evaluate
> > (this=0xbfb306f8,
> > > > > > > values=0xbfb2f3b0, coordinates=0xbfb2f3c8, cell=@0xbfb2f5d8)
> > > > > > >     at dolfin/function/GenericFunction.cpp:35
> > > > > > > #11 0x080bc905 in plas2d_1_finite_element_2::evaluate_dof
> > > > (this=0x9bb7558,
> > > > > > i=0,
> > > > > > > f=@0xbfb306f8, c=@0xbfb2f5d8)
> > > > > > >     at
> > > > > > >
> > > > > >
> > > > >
> > > >
> > >
> >
> /home/oelgaard/fenics_projects/fenics_apps/fenics-plasticity/demo/forms/2Dlinear_QE_forms/Plas2D.cpp:10433
> > > > > > > #12 0xb7d00f4c in dolfin::GenericFunction::restrict_as_ufc_function
> > > > > > > (this=0xbfb306f8, w=0x9bb7c38, element=@0x9bb7580,
> > > > dolfin_cell=@0xbfb2f520,
> > > > > > >     ufc_cell=@0xbfb2f5d8, local_facet=-1) at
> > > > > > ./dolfin/fem/FiniteElement.h:61
> > > > > > > #13 0xb7cf2f8e in dolfin::Function::restrict (this=0xbfb306b4,
> > > > w=0x9bb7c38,
> > > > > > > element=@0x9bb7580, dolfin_cell=@0xbfb2f520, ufc_cell=@0xbfb2f5d8,
> > > > > > > local_facet=-1)
> > > > > > >     at dolfin/function/Function.cpp:490
> > > > > > > #14 0xb7ceb0ce in dolfin::UFC::update (this=0xbfb2f5b0,
> > > > cell=@0xbfb2f520)
> > > > > > at
> > > > > > > dolfin/fem/UFC.cpp:212
> > > > > > > #15 0xb7cb2e85 in dolfin::Assembler::assemble_cells (A=@0x9bbb838,
> > > > > > > a=@0xbfb30b18, ufc=@0xbfb2f5b0, domains=0x0, values=0x0) at
> > > > > > > dolfin/fem/Assembler.cpp:159
> > > > > > > #16 0xb7cb3475 in dolfin::Assembler::assemble (A=@0x9bbb838,
> > > > a=@0xbfb30b18,
> > > > > > > cell_domains=0x0, exterior_facet_domains=0x0,
> > > > interior_facet_domains=0x0,
> > > > > > >     reset_sparsity=<value optimized out>, add_values=<value
> > optimized
> > > > out>)
> > > > > > at
> > > > > > > dolfin/fem/Assembler.cpp:115
> > > > > > > #17 0xb7cb35e2 in dolfin::Assembler::assemble (A=@0x9bbb838,
> > > > a=@0xbfb30b18,
> > > > > > > reset_sparsity=<value optimized out>, add_values=<value optimized
> > out>)
> > > > > > >     at dolfin/fem/Assembler.cpp:50
> > > > > > > #18 0xb7eef1f8 in dolfin::PlasticityProblem::F (this=0xbfb2f814,
> > > > > > b=@0x9bbb838,
> > > > > > > x=@0x9bb0e48) at src/PlasticityProblem.cpp:92
> > > > > > > #19 0xb7df9baf in dolfin::NewtonSolver::solve (this=0xbfb308d4,
> > > > > > > nonlinear_problem=@0xbfb2f814, x=@0x9bb0e48) at
> > > > > > dolfin/nls/NewtonSolver.cpp:76
> > > > > > > #20 0x08096370 in main () at main.cpp:184
> > > > > > >
> > > > > > > It's pretty obvious what the error is, but before you say that we
> > > > should
> > > > > > just
> > > > > > > implement evaluate_basis() for QuadratureElement (which is alway 1,
> > if
> > > > it's
> > > > > > > evaluate exactly at the quadrature point, otherwise it's undefined)
> > I
> > > > would
> > > > > > > like to know why it is necessary to call evaluate on a function
> > that
> > > > > > contains a
> > > > > > > computed result. It should be possible to just pass this function
> > to
> > > > the
> > > > > > Form
> > > > > > > which will handle the interpolation when integrating the form.
> > > > > > >
> > > > > > > Kristian
> > > > > >
> > > > > > I don't understand why evaluate_basis is called. The assembler
> > > > > > restricts all coefficients to the local element which results in a
> > > > > > call to evaluate_dof which results in a call to evaluate. Why is
> > > > > > Strain2D.cpp calling evaluate_basis on line 495? It should just
> > return
> > > > > > the value.
> > > > >
> > > > > evaluate() ends up with a call to Function::eval() which calls
> > > > evaluate_basis()
> > > > >
> > > > > The attached forms will reproduce the error.
> > > > >
> > > > > Kristian
> > > >
> > > > It seems like something goes wrong in Function::gather. Can you verify
> > > > that you end up in the else clause of that function?
> > >
> > > The else clause is in the Function::restrict() right?
> > > When solving the projection problem of the quadrature function onto
> > linears, I
> > > end up in the restrict() function twice. The first time I'm in the else
> > clause,
> > > the second time in the if clause.
> > >
> > > Kristian
> >
> > Can you create a simple code that reproduces this error? ;-)
>
> Attached is the simplest code that will crash DOLFIN.
>
> > Even better, check why FunctionSpace::has_element returns false.
> > It should only return false when something is interpolated onto a
> > different mesh or used as a coefficient in a form for a function space
> > it was not intended for.
>
> On the first run, it returns false because the elements are different. The
> element signatures are the same but:
>
> &element == &*_element
>
> returns false.
>
> On the second call to FunctionSpace::has_element(), the two elements are equal.
>
> Kristian

This should work now.

--
Anders

Attachment: signature.asc
Description: Digital signature