[CIG-SEISMO] SPECFEM3D / PS

[Dimitri Komatitsch]

Hi James, Hi all,

Very good idea. Tomorrow I will rename it and will move it to the 
Par_file, and will also remove the stop statement (more precisely, I 
will convert it to a stop statement in case UNDO_ATT is not set but PML 
is set).

Vadim is cleaning his local implementation of UNDO_ATT, we will soon 
commit it to the development branch (see 
https://github.com/geodynamics/specfem3d/issues/14 ). For now only 
SPECFEM3D_GLOBE and SPECFEM2D have it, regarding SPECFEM3D (Cartesian) 
we are almost done.

Best,
Dimitri.

On 10/30/2017 09:14 PM, James A. Smith wrote:
> Hi Dimitri,
> 
> Shouldn't we also remove the code that stops the simulation if you attempt to run with PML for adjoint run?
> 
> Also, the UNDO_ATTENUATION flag is in setup/constants.h. Does it make more sense for it to be in Par_file if users will change it occasionally?
> 
> Best,
> James
> 
> ________________________________________
> From: Dimitri Komatitsch [komatitsch at lma.cnrs-mrs.fr]
> Sent: Monday, October 30, 2017 3:07 PM
> To: cig-seismo at geodynamics.org; James A. Smith; Dylan Mikesell
> Cc: Vadim Monteiller
> Subject: Re: [CIG-SEISMO] SPECFEM3D / PS
> 
> Hi all,
> 
> PS: we could thus consider renaming the UNDO_ATTENUATION flag to
> UNDO_ATTENUATION_AND_OR_PML. I know people do not like when we rename
> flags because Par_files then need to be updated, but this one has become
> a bit confusing.
> 
> Best,
> Dimitri.
> 
> On 10/30/2017 08:59 PM, Dimitri Komatitsch wrote:
>>
>> Hi Dylan,
>>
>> Yes, just use UNDO_ATTENUATION, which can undo any dissipative medium
>> exactly, including PML. It is described in
>> http://komatitsch.free.fr/preprints/GJI_undo_attenuation_2016.pdf .
>> We do that routinely here for FWI with PML, works beautifully.
>> Please cite that paper if you use it.
>>
>> Thanks,
>> Best,
>> Dimitri.
>>
>> On 10/30/2017 06:44 PM, James A. Smith wrote:
>>> Hi Dylan,
>>>
>>> I have tested PML for the adjoint simulation. It seems to work,
>>> although I am still playing with getting higher quality inversion
>>> results (as compared to Stacey). Please see this issue:
>>> https://github.com/geodynamics/specfem3d/issues/312
>>>
>>> You will need to enable UNDO_ATTENUATION for PML to work in this case
>>> (in setup/constanst.h). You will also have to comment out the lines in
>>> the source code that stop the code if PML is turned on for an adjoint
>>> simulation and recompile. I hope it works for you.
>>>
>>> Best,
>>> James
>>>
>>> ------------------------------------------------------------------------
>>> *From:* CIG-SEISMO [cig-seismo-bounces at geodynamics.org] on behalf of
>>> Dylan Mikesell [dylanmikesell at boisestate.edu]
>>> *Sent:* Friday, October 27, 2017 2:17 PM
>>> *To:* cig-seismo at geodynamics.org
>>> *Subject:* [CIG-SEISMO] SPECFEM3D
>>>
>>> Hi SPECFEM developers,
>>>
>>> Is anyone currently working on the PML implementation for the adjoint.
>>> A student and I are working on an adjoint tomography and planning to
>>> use SPECFEM3D. We tried to save the forward simulations today and got
>>> the following.
>>>
>>> STOP PML_CONDITIONS is still under test for adjoint simulation
>>>
>>> I checked the devel branch and this is still there. Is anyone actively
>>> working on this topic? It would be nice to get the PML going with the
>>> adjoint.
>>>
>>> Best wishes,
>>>
>>> Dylan
>>>
>>>
>>> _______________________________________________
>>> CIG-SEISMO mailing list
>>> CIG-SEISMO at geodynamics.org
>>> http://lists.geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo
>>>
>>
> 
> --
> Dimitri Komatitsch, CNRS Research Director (DR CNRS)
> Laboratory of Mechanics and Acoustics, Marseille, France
> http://komatitsch.free.fr
> 

-- 
Dimitri Komatitsch, CNRS Research Director (DR CNRS)
Laboratory of Mechanics and Acoustics, Marseille, France
http://komatitsch.free.fr