[CIG-SEISMO] error with xgenerate_database (more details)

[Dimitri Komatitsch]

Hi Andrea, Hi all,

Thanks! Fixed.

Dimitri.

On 08/25/2015 07:16 PM, Andrea Colombi wrote:
> Dear all,
> thanks for the email. I have definitely seen this error more than once
> in the recent version of specfem3d cart. In my case I only had to
> slightly modify the design of the mesh to get rid of it but i do not
> remember precisely how (maybe increasing the thickness in the z
> direction?). If you are working with cubit I can have a look if you
> need. Sorry for not being much of help.
>
> I have however found the following error (which i believe is more
> related to compiler/architecture issues)
> The following line in pml_set_local_dampingcoeff.f90 cause an error in
> the arithmetic operation with intel compiler.
>      ! In INRIA research report section 6.1:
> http://hal.inria.fr/docs/00/07/32/19/PDF/RR-3471.pdf
>      ! pml_damping_profile_l = - ((NPOWER + 1) * vp * log(CPML_Rcoef) /
> (TWO * delta)) * dist**(NPOWER)
>      ! due to tests it is more accurate to use following definition in
> case NPOWER = 1 defined in constants.h.in <http://constants.h.in>
>      pml_damping_profile_l = - ((NPOWER + 1) * vp * log(CPML_Rcoef) /
> (TWO * delta)) &
>                              * dist**(1.2_CUSTOM_REAL * NPOWER)
>
> I have replaced with this as NPOWER=1 and it does not create me problems
>
> pml_damping_profile_l = - (2.0* vp * log(CPML_Rcoef) / (2.0 * delta))*
> dist**(1.2)
>
>
> Andrea
>
> 2015-08-25 0:34 GMT+01:00 Dimitri Komatitsch <komatitsch at lma.cnrs-mrs.fr
> <mailto:komatitsch at lma.cnrs-mrs.fr>>:
>
>
>     Dear Moritz, dear all,
>
>     It seems Andrea Colombi (cc'ed) has also noticed this problem and
>     has fixed it. I will send another email tomorrow with more details
>     and I will then commit Andrea's changes to the official source code
>     ("devel" version).
>
>     Thank you,
>     Best regards,
>
>     Dimitri.
>
>     On 08/25/2015 12:55 AM, Dimitri Komatitsch wrote:
>
>
>         Dear Moritz,
>
>         Yes, this seems to come from a small bug in the PML routines (in the
>         part that avoids a singularity in the PML layers). Zhinan and/or
>         René,
>         do you have an idea of why Moritz' run stops with this error
>         message?
>
>         Could you please email your input files to Zhinan (cc'ed), so
>         that he
>         can have a look and fix the bug?
>
>         Thanks,
>         Best wishes,
>
>         Dimitri.
>
>         On 07/11/2015 10:51 PM, moritzwehler at muenchen-ist-toll.de
>         <mailto:moritzwehler at muenchen-ist-toll.de> wrote:
>
>             Hey there,
>
>             after running xgenerate_database in specfem3d_cartesian I
>             received the
>             following error
>
>             STOP error in seperation of alpha_x, alpha_y, alpha_z in
>             case alpha_y ==
>             alpha_z .and. alpha_x /= alpha_y
>             STOP error in seperation of alpha_x, alpha_y, alpha_z in
>             case alpha_y ==
>             alpha_z .and. alpha_x /= alpha_y
>             STOP error in seperation of alpha_x, alpha_y, alpha_z in
>             case alpha_x ==
>             alpha_z .and. alpha_y /= alpha_z
>             STOP error in seperation of alpha_x, alpha_y, alpha_z in
>             case alpha_x ==
>             alpha_y .and. alpha_y /= alpha_z
>
>             The only reason I can think of is that I made the element
>             size of my
>             mesh (=50 meters) to small. Has anyone stumbled upon the
>             same or a
>             similar error yet?
>
>             Thanks a lot in advance and best regards
>
>             Moritz
>
>
>
>             _______________________________________________
>             CIG-SEISMO mailing list
>             CIG-SEISMO at geodynamics.org <mailto: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,
>     UPR 7051, Marseille, France http://komatitsch.free.fr
>
>

-- 
Dimitri Komatitsch
CNRS Research Director (DR CNRS), Laboratory of Mechanics and Acoustics,
UPR 7051, Marseille, France    http://komatitsch.free.fr