[CIG-SEISMO] Problems with Scotch in Specfem2D

[Dimitri Komatitsch]

Dear Philip,

Thank you for your feedback. Are you using the current SVN version ? 
which you can get using:

svn checkout http://geodynamics.org/svn/cig/seismo/2D/SPECFEM2D/trunk 
SPECFEM2D

The error messages below are from Scotch, not from Specfem2D, and 
indicate that no default partitioning strategy is defined, which is 
weird because in Scotch version 5.1.11 (and I guess in 5.1.10 as well) 
the best strategy is automatically selected by Scotch (and thus we 
removed the "strategy selection" option from SPECFEM2D). I see three 
options:

- you are using an older version of SCOTCH (?)

- you are not using the SVN version of SPECFEM2D

- there is a bug somewhere in Scotch when using 64 bits

Please try with the latest SVN version above, and if you still get the 
bug please email me your input file and a tar file of the SPECFEM and 
SCOTCH source codes and I will try to locate and fix the bug.

Thank you,
Best regards,

Dimitri.


On 02/07/2011 08:03 PM, Philip Knaute wrote:
> Dear all,
>
> After some time using Specfem 2D 6.1.1 with the internal mesh
> partitioning routines and without Scotch, I now tried to get the code
> with the scotch package to run.
>
> Sadly was not very successful, as of yet. Compiling Scotch and the
> Specfem2D code with 64-bit version of the Intel compiler package 11.1
> works fine, but when I try to run xmeshfem2D I get the error message below:
>
> -------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> Parameter file successfully read...
>
> The mesh contains4elements
>
> Control elements have4nodes
>
> Reading interface data from file DATA/Lower_Fars_Test.dat
>
> Min and max value of X in the grid =0.000000000000000E+000
>
> 2650.00000000000
>
> Min and max value of Z in the grid =0.000000000000000E+000
>
> 598.000000000000
>
> Saving the grid in Gnuplot format...
>
> drawing horizontal lines of the grid
>
> drawing vertical lines of the grid
>
> Grid saved in Gnuplot format...
>
> ERROR: stratParserParse: invalid strategy string, before
> "t,map=t,poli=S,sep=(m{type=h,vert=80,low=h{pass=10}f{bal=0.003322,move=80},asc=b{bnd=(d{dif=1,rem=1,pass=40}|)f{bal=0.003322,move=80},org=f{bal=0.003322,move=80}}}|m{type=h,vert=80,low=h{pass=10}f{bal=0.003322,move=80},asc=b{bnd=(d{dif=1,rem=1,pass=40}|)f{bal=0.003322,move=80},org=f{bal=0.003322,move=80}}})}"
>
> ERROR: SCOTCH_stratGraphMap: error in mapping strategy
>
> ERROR: SCOTCH_stratGraphMapBuild: error in sequential mapping strategy
>
> forrtl: severe (174): SIGSEGV, segmentation fault occurred
>
> ImagePCRoutineLineSource
>
> xmeshfem2D000000000044D0BEUnknownUnknownUnknown
>
> xmeshfem2D000000000044B82CUnknownUnknownUnknown
>
> xmeshfem2D000000000042B4E4part_unstruct_mp_1273part_unstruct.F90
>
> xmeshfem2D000000000043B024MAIN__1497meshfem2D.F90
>
> xmeshfem2D000000000040CF8CUnknownUnknownUnknown
>
> libc.so.60000003F2DC1D994UnknownUnknownUnknown
>
> xmeshfem2D000000000040CE99UnknownUnknownUnknown
>
> -------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
> If I play around with the ‘–DINTSIZE flag’ and ‘-integer-size xx’ for
> scotch or Specfem respectively, the code only stops working earlier.
>
> Does anypne experienced something similar or knows a solution for my
> problem.
>
> I would be very grateful.
>
> Thank you very much in advance.
>
> Cheers,
>
> Philip
>
>
>
> _______________________________________________
> CIG-SEISMO mailing list
> CIG-SEISMO at geodynamics.org
> http://geodynamics.org/cgi-bin/mailman/listinfo/cig-seismo

-- 
Dimitri Komatitsch - dimitri.komatitsch aT get.obs-mip.fr
Professor, University of Toulouse and Institut universitaire de France,
CNRS UMR 5563 GET, France   http://www.univ-pau.fr/~dkomati1