[CIG-SEISMO] specfem3d : simple_model with CPML

[komatitsch at lma.cnrs-mrs.fr]

Hi,

you can try ./configure --enable-debug to get more useful error  
messages (if so please let us know)

thanks
dimitri


houssen <houssen at ipgp.fr> a écrit :

> OK, I believed the right branch to be on was master ! As far as I  
> understand, you fast-forwarded master up to devel. Anyway, I on  
> devel now.
> At IPGP, we have the SEM code you may know ! :D For a start, I'd  
> like to compare SEM and specfem3d on a simple case (with and without  
> PML - then I'd like to play on NGLL).
>
> To be back on the initial problem :
> 1. As far as I understand, changes between the old master and devel  
> are mainly cosmetic (comments: #).
> 2. The problem (crash) persists on devel.
> 3. I started with what I believe to be the most simple case  
> (xmeshfem3d, simple_model) :
> 3.1. Without PML, specfem3D runs fine (mesh OK, database OK, run OK).
> 3.2. With PML, the code crashes (mesh OK, database KO).
>
> To focus on 3.2: basically, from the case without PML which works I  
> just turned ON the PML_CONDITIONS keyword in Par_file.
> 1. I have now files from devel.
> 2. The only differences (PML - no PML) between Par_files are:
> ~> diff  
> valid-run-specfem3d-nopml/meshfem3D/simple_model/DATA/Par_file  
> valid-run-specfem3d-pml/meshfem3D/simple_model/DATA/Par_file
> 90c90
> < PML_CONDITIONS = .false.
> ---
>> PML_CONDITIONS = .true.
> 105c105
> < STACEY_ABSORBING_CONDITIONS     = .true.
> ---
>> STACEY_ABSORBING_CONDITIONS = .false.
> I turned OFF STACEY_ABSORBING_CONDITIONS as I had an error message  
> on the old master when I turned on PML_CONDITIONS.
> 3. No difference between Mesh_Par_files:
> ~> diff  
> valid-run-specfem3d-nopml/meshfem3D/simple_model/DATA/meshfem3D_files/Mesh_Par_file  
> valid-run-specfem3d-pml/meshfem3D/simple_model/DATA/meshfem3D_files/Mesh_Par_file
>
> When I turn ON PML_CONDITIONS in Par_file, I get this:
>  creating PML region
>  Created a total of            0  unique CPML elements
>     (i.e.,    0.00000000     % of the mesh)
> In xmeshfem3D logs, and, I get this:
>    ...creating C-PML damping profiles
> Program received signal SIGSEGV: Segmentation fault - invalid memory  
> reference.
> Backtrace for this error:
> #0  0x7fa98f0d7d17 in ???
> #1  0x7fa98f0d6f7d in ???
> #2  0x7fa98e3da03f in ???
> #3  0x442aef in pml_set_local_dampingcoeff_
>   at src/generate_databases/pml_set_local_dampingcoeff.f90:95
> #4  0x418f0d in create_regions_mesh_
>   at src/generate_databases/create_regions_mesh.f90:238
> #5  0x45fe08 in setup_mesh_
>   at src/generate_databases/setup_mesh.f90:66
> #6  0x41c511 in generate_databases_
>   at src/generate_databases/generate_databases.f90:254
> #7  0x40269f in xgenerate_databases
>   at src/generate_databases/program_generate_databases.f90:34
> #8  0x40269f in main
>   at src/generate_databases/program_generate_databases.f90:39
> --------------------------------------------------------------------------
> mpirun noticed that process rank 0 with PID 0 on node yoda exited on  
> signal 11 (Segmentation fault).
> --------------------------------------------------------------------------
> In xgenerate_database logs.
>
> I tried to debug a bit, but, I didn't get what's wrong in  
> Mesh_Par_file and / or Par_file : I understand the PML region has no  
> element but I can't understand why and can't neither fix the issue.
>
> Franck
>
> Le 2016-10-25 19:53, Dimitri Komatitsch a écrit :
>> Dear Franck,
>>
>> Could you try with the "devel" version and let us know?
>>
>> Let me replace "master" with a copy of "devel" because it is old.
>>
>> You can also check EXAMPLES/CPML_examples.
>
> I do not have CUBIT. My understanding is that EXAMPLES/CPML_examples  
> are all meshed with CUBIT.
> ~/specfem3d> find . -name Mesh_Par_file
> ./EXAMPLES/BENCHMARK_CLAERBOUT_ADJOINT/ELASTIC/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/BENCHMARK_CLAERBOUT_ADJOINT/ACOUSTIC/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/meshfem3D_examples/cavity/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/meshfem3D_examples/simple_model/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/meshfem3D_examples/sep_bathymetry/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/meshfem3D_examples/many_interfaces/DATA/meshfem3D_files/Mesh_Par_file
> ./EXAMPLES/meshfem3D_examples/socal1D/DATA/meshfem3D_files/Mesh_Par_file
>
>>
>> Please keep us informed,
>> Thanks,
>>
>> Dimitri.
>>
>> On 25/10/2016 16:28, houssen wrote:
>>> Hello,
>>>
>>> I am new to specfem3d. Is it possible to run the test
>>> EXAMPLES/meshfem3D_examples/simple_model from specfem (git clone) with
>>> CPML ? If yes how ?
>>> I tried to play over different parameters in Mesh_Par_file and Par_file
>>> without success (in particular adding CPML_CONDITIONS keyword in
>>> Mesh_Par_file - this didn't work).
>>>
>>> I git-cloned https://github.com/geodynamics/specfem3d.git (master). I am
>>> running debian with gnu compilers on a laptop.
>>>
>>> Thanks
>>>
>>> Franck
>>>
>>>
>>> _______________________________________________
>>> CIG-SEISMO mailing list
>>> CIG-SEISMO at geodynamics.org
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>>
>
>