Re: [Libmesh-devel] Latest updates to FE

[Boyce G. <gri...@ci...>]

Here is a patch to make edge_reinit() behave like reinit(elem,side) in 
terms of how it sets shapes_on_quadrature.

-- Boyce

On 5/3/11 11:53 AM, Boyce Griffith wrote:
> What is going wrong with caching?
>
> Looking at line 178 of fe_boundary.C:
>
> // The shape functions correspond to the qrule
> shapes_on_quadrature = (qrule != NULL);
>
> If this should be there at all, it seems like it should be:
>
> // The shape functions correspond to the qrule
> shapes_on_quadrature = (pts == NULL);
>
> -- Boyce
>
> On 5/3/11 11:42 AM, Roy Stogner wrote:
>>
>> Cool, but be vigilant. There'll be more changes coming; some bugs in
>> how the new patch interacts with caching that I need to fix.
>> ---
>> Roy
>>
>> On Tue, 3 May 2011, John Peterson wrote:
>>
>>> I can confirm that this change fixes the segfault in ex0.
>>>
>>> Checking our other regression tests now...
>>>
>>> --
>>> John
>>>
>>>
>>> On Tue, May 3, 2011 at 9:25 AM, Boyce Griffith <gri...@ci...>
>>> wrote:
>>>> Grrrrrrrr, looks like there was a bug in the patch (which I am
>>>> surprised
>>>> didn't cause me any problems).
>>>>
>>>> Index: src/fe/fe_boundary.C
>>>> ===================================================================
>>>> --- src/fe/fe_boundary.C (revision 4447)
>>>> +++ src/fe/fe_boundary.C (working copy)
>>>> @@ -137,7 +137,7 @@
>>>> else
>>>> {
>>>> // initialize quadrature rule
>>>> - if (pts != NULL) qrule->init(side->type(), side_p_level);
>>>> + qrule->init(side->type(), side_p_level);
>>>>
>>>> // FIXME - could this break if the same FE object was used
>>>> // for both volume and face integrals? - RHS
>>>>
>>>> On 5/3/11 11:20 AM, Roy Stogner wrote:
>>>>>
>>>>> Yeah; looks like Boyce's patch missed the FEXYZ versions of those
>>>>> functions. I should have noticed those warnings earlier; I'll fix it.
>>>>> ---
>>>>> Roy
>>>>>
>>>>> On Tue, 3 May 2011, Cody Permann wrote:
>>>>>
>>>>>> These errors were on OS X but I just confirmed similar errors on
>>>>>> Linux
>>>>>> as well.
>>>>>>
>>>>>> Cody
>>>>>>
>>>>>> On May 3, 2011, at 9:06 AM, John Peterson wrote:
>>>>>>
>>>>>>> I'm getting lots of warnings after the latest patch to the FE stuff
>>>>>>> when compiling in dbg mode:
>>>>>>>
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h: In instantiation
>>>>>>> of ‘libMesh::FEXYZ<0u>’:
>>>>>>> src/fe/fe_base.C:106: instantiated from here
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h:304: warning:
>>>>>>> ‘void
>>>>>>> libMesh::FE<Dim, T>::reinit(const libMesh::Elem*,
>>>>>>> const std::vector<libMesh::Point, std::allocator<libMesh::Point>
>>>>>>>>
>>>>>>>> *, const std::vector<double, std::allocator<double> >*)
>>>>>>>
>>>>>>> [with unsigned int Dim = 0u, libMeshEnums::FEFamily T = XYZ]’ was
>>>>>>> hidden
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h:646: warning: by
>>>>>>> ‘libMesh::FEXYZ<0u>::reinit’
>>>>>>>
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h:319: warning:
>>>>>>> ‘void
>>>>>>> libMesh::FE<Dim, T>::reinit(const libMesh::Elem*,
>>>>>>> unsigned int, libMesh::Real, const std::vector<libMesh::Point,
>>>>>>> std::allocator<libMesh::Point> >*,
>>>>>>> const std::vector<double, std::allocator<double> >*)
>>>>>>> [with unsigned int Dim = 0u, libMeshEnums::FEFamily T = XYZ]’ was
>>>>>>> hidden
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h:646: warning: by
>>>>>>> ‘libMesh::FEXYZ<0u>::reinit’
>>>>>>> /Users/petejw/projects/libmesh_git/include/fe/fe.h: In instantiation
>>>>>>> of ‘libMesh::FEXYZ<1u>’:
>>>>>>>
>>>>>>> And a segfault in ex0:
>>>>>>>
>>>>>>> Starting program:
>>>>>>> /Users/petejw/projects/libmesh_git/examples/ex0/ex0-dbg
>>>>>>> Reading symbols for shared libraries .+++++++++++++++++++++++. done
>>>>>>>
>>>>>>> Program received signal EXC_BAD_ACCESS, Could not access memory.
>>>>>>> Reason: KERN_INVALID_ADDRESS at address: 0x0000000000000008
>>>>>>> 0x0000000100be71a1 in libMesh::TypeVector<double>::operator()
>>>>>>> (this=0x0, i=0) at type_vector.h:356
>>>>>>> 356 return _coords[i];
>>>>>>> (gdb) bt
>>>>>>> #0 0x0000000100be71a1 in libMesh::TypeVector<double>::operator()
>>>>>>> (this=0x0, i=0) at type_vector.h:356
>>>>>>> #1 0x00000001008d801b in libMesh::FE<1u,
>>>>>>> (libMeshEnums::FEFamily)0>::shape (order=libMeshEnums::SECOND, i=0,
>>>>>>> p=@0x0) at src/fe/fe_lagrange_shape_1D.C:39
>>>>>>> #2 0x00000001008045c3 in libMesh::FE<1u,
>>>>>>> (libMeshEnums::FEFamily)0>::init_shape_functions (this=0x10281c600,
>>>>>>> qp=@0x7fff5fbfe0e0, elem=0x102003f80) at src/fe/fe.C:501
>>>>>>> #3 0x00000001007c0200 in libMesh::FE<1u,
>>>>>>> (libMeshEnums::FEFamily)0>::reinit (this=0x10281c600,
>>>>>>> elem=0x102003f80, pts=0x7fff5fbfe0e0, weights=0x0) at
>>>>>>> src/fe/fe.C:135
>>>>>>> #4 0x00000001008655c7 in libMesh::FE<1u,
>>>>>>> (libMeshEnums::FEFamily)0>::reinit (this=0x10281c600,
>>>>>>> elem=0x102003f80, s=1, tolerance=9.9999999999999995e-07, pts=0x0,
>>>>>>> weights=0x0) at src/fe/fe_boundary.C:180
>>>>>>> #5 0x0000000100790875 in libMesh::JumpErrorEstimator::reinit_sides
>>>>>>> (this=0x7fff5fbfe570) at
>>>>>>> src/error_estimation/jump_error_estimator.C:421
>>>>>>> #6 0x0000000100791696 in libMesh::JumpErrorEstimator::estimate_error
>>>>>>> (this=0x7fff5fbfe570, system=@0x102004580,
>>>>>>> error_per_cell=@0x7fff5fbfe8d0, solution_vector=0x0,
>>>>>>> estimate_parent_error=false) at
>>>>>>> src/error_estimation/jump_error_estimator.C:307
>>>>>>> #7 0x0000000100004808 in main (argc=1, argv=0x7fff5fbfe9a8) at
>>>>>>> ex0.C:125
>>>>>>>
>>>>>>> Some moose tests are failing as well, working on getting stack
>>>>>>> traces
>>>>>>> for those.
>>>>>>>
>>>>>>> In the meantime, could we revert the latest patch Roy?
>>>>>>>
>>>>>>> --
>>>>>>> John
>>>>>>>
>>>>>>>
>>>>>>> ------------------------------------------------------------------------------
>>>>>>>
>>>>>>>
>>>>>>>
>>>>>>> WhatsUp Gold - Download Free Network Management Software
>>>>>>> The most intuitive, comprehensive, and cost-effective network
>>>>>>> management toolset available today. Delivers lowest initial
>>>>>>> acquisition cost and overall TCO of any competing solution.
>>>>>>> http://p.sf.net/sfu/whatsupgold-sd
>>>>>>> _______________________________________________
>>>>>>> Libmesh-devel mailing list
>>>>>>> Lib...@li...
>>>>>>> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
>>>>>>
>>>>>>
>>>>>>
>>>>>> ------------------------------------------------------------------------------
>>>>>>
>>>>>>
>>>>>>
>>>>>> WhatsUp Gold - Download Free Network Management Software
>>>>>> The most intuitive, comprehensive, and cost-effective network
>>>>>> management toolset available today. Delivers lowest initial
>>>>>> acquisition cost and overall TCO of any competing solution.
>>>>>> http://p.sf.net/sfu/whatsupgold-sd
>>>>>> _______________________________________________
>>>>>> Libmesh-devel mailing list
>>>>>> Lib...@li...
>>>>>> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
>>>>>>
>>>>>
>>>>>
>>>>>
>>>>> ------------------------------------------------------------------------------
>>>>>
>>>>>
>>>>> WhatsUp Gold - Download Free Network Management Software
>>>>> The most intuitive, comprehensive, and cost-effective network
>>>>> management toolset available today. Delivers lowest initial
>>>>> acquisition cost and overall TCO of any competing solution.
>>>>> http://p.sf.net/sfu/whatsupgold-sd
>>>
>>> ------------------------------------------------------------------------------
>>>
>>>
>>> WhatsUp Gold - Download Free Network Management Software
>>> The most intuitive, comprehensive, and cost-effective network
>>> management toolset available today. Delivers lowest initial
>>> acquisition cost and overall TCO of any competing solution.
>>> http://p.sf.net/sfu/whatsupgold-sd
>>> _______________________________________________
>>> Libmesh-devel mailing list
>>> Lib...@li...
>>> https://lists.sourceforge.net/lists/listinfo/libmesh-devel
>>>