FORCESOLUTION may have a problem

[komatits]

From: Dimitri Komatitsch to @bottero and @bencesolymosi

It is true that for now this option is mostly unused and thus untested (in geophysics most people use a CMTSOLUTION, and in ocean acoustics most of the time the source is in the fluid rather than in the solid).

Bence will soon use it though.

Let me thus open a new "Git issue" about it and cut and paste your emails below in it, so that we remember to do it in a few months.

Thanks, Dimitri.

On 11/09/2015 10:29 AM, bottero wrote:

Hello Florian, You are right FORCESOLUTION option is a mess (not through the fault of Bence nor me)! For the moment I don't use it either so I won't fix it. When I'll need it (probably within the next 6 monthes) I'll fix it and commit my changes if it has not been done before.

Thank you for your feed back

Alexis

On 09/11/2015 09:55, Florian Schumacher wrote:

Hallo Alexis and Solymosi,

referring to Dimitris answer below, I'd like to point out, that I am not sure whether in ALL calls to comp_source_time_function_gauss() the hard-coded value of 5.d0*DT should be replaced by hdur_gaussian(isource) and I prefer you guys to have a look at it. I will not commit any changes to the repository.

I've also noticed that in compute_add_sources_acoustic.f90 and compute_add_sources_poroelastic.f90 (in devel branch), there are already some calls to comp_source_time_function_gauss() where 5.d0_DT was replaced by hdur_gaussian(isource) , and other calls where still 5.d0_DT remains. Is this mixture intended?

Thanks! Florian

-------- Ursprüngliche Nachricht -------- Von: Dimitri Komatitsch Gesendet: 8. November 2015 23:30:00 MEZ An: Florian Schumacher Betreff: Re: [CIG-SEISMO] SPECFEM3d_Cartesian , using point force with gaussian stf

Hi Florian,

Yes, I committed it without checking it, I thought the issue was fully fixed already. Let me thus cc Alexis Bottero and Bence Solymosi, who are currently both using the FORCESOLUTION option, to see how this could be fixed.

(or even better, if you know how to fix it yourself please do not hesitate to do it and then either email me the file or follow https://github.com/geodynamics/specfem3d/wiki/Using-Git-for-SPECFEM to commit it to Git directly)

Thanks again! Cheers,

Dimitri.

On 11/08/2015 06:47 PM, Florian Schumacher wrote:

Hallo Dimitri,

preparing the next ASKI release, I noticed that your commit 4bc3dd6 to the SPECFEM3D_Cartesian devel branch on 3 Sep 2015 based on my suggestions only makes partially sense to me: https://github.com/geodynamics/specfem3d/commit/4bc3dd673407b950e0372116a7d6b2029d03ea2c#diff-e0c13ae549ca95a62c20939960769fdf

It seems that you committed a 1-to-1 copy of the file compute_add_sources_viscoelastic.f90 that I attached to my email. As stated below, I was not sure about whether to replace 5.d0_DT by hdur_gaussian(isource) in ALL calls of comp_source_time_function_gauss() . That is why on 3 lines I left the original value of 5.d0_DT in my attached file (lines 507, 687, 853 in commit 4bc3dd6, in the current revision we are talking about lines 510, 691, 858). On these lines, I find that you only committed my comment "! FS FS does it also here make sense to replace 5.d0*DT by hdur_gaussian(isource) ? looks like it" but did not modify the code in any way. This seems odd to me, unless you intended to leave my comment there since you were unsure yourself whether to apply the change.

Did you leave the value of 5.d0*DT by intention for adjoint and GPU simulations? In case you did not notice my asking to have a look at it whether it makes sense to change it, I am sorry, I should have pointed out it out more clearly.

Sorry to bother you again on this account.

Best regards, Florian

[specfem3d-zhang-ksu]

how can I setup FORCESOLUTION and Par_file, so that I convolve the caculated seismogram with a source wavelet to obtain the resulting synthetic seismogram in post-processing after the run?

[komatits]

Fixed by @vmont and @bencesolymosi ; @casarotti will confirm that the problem is fixed.