Learn how easy it is to sync an existing GitHub or Google Code repo to a SourceForge project! See Demo

Close

OpenFVM v1.1 parallel & petsc-3.0.0-p8

2009-10-08
2013-04-25
  • Hi @ all!

    There is a little problem compiling OpenFVM parallel version with petsc 3.0.0. Because of the new API the function MatSetOptions in source/matrix.c in line 38 and 39 needs 3 Parameters (PETSC_TRUE or PETSC_FALSE). I was able to compile OpenFVM after inserting PETSC_TRUE as third parameter, but I have not tested it yet.

    The value in line 39, MAT_IGNORE_LOWER_TRIANGULAR, is not available anymore. I don't know which version is the latest supporting this.

    Which Metis is needed? Metis or ParMetis?

    After compiling and starting a job, in the Header "OpenFVM parallel v1.0" is shown. Is this ok?

    Greetings
    Alex

     
  • Sorry, my fault, MAT\_IGNORE\_LOWER\_TRIANGULAR is still available! But other arguments cause warnings.

     
  • Billy Araujo
    Billy Araujo
    2009-10-08

    Hi,

    About the MatSetOptions I will check into that.
    The code uses Metis because the decomposition is made in serial before computations but I believe ParMetis also includes Metis.
    There are some inconsistencies in the version, I think this has been corrected in revision of SVN. I recommend using the latest revision from SVN.

     
  • Hi!

    I have downloaded the newest svn version. Some warnings has been fixed, but there are still two warnings in source/matrix.c:

    source/matrix.c: In Funktion »Q_GetVal«:
    source/matrix.c:66: Warnung: Übergabe des Arguments 6 von »MatGetValues« von inkompatiblem Zeigertyp
    source/matrix.c: In Funktion »Q_SetEntries«:
    source/matrix.c:85: Warnung: Übergabe des Arguments 6 von »MatSetValues« von inkompatiblem Zeigertyp

    http://www.mcs.anl.gov/petsc/petsc-as/snapshots/petsc-current/docs/manualpages/Mat/MatGetValues.html
    Says that MatGetValues last Parameter has to be scalar (two dimensional array), MatSetValues needs it too.

    Greetings

     
  • Addition: After executing I get an PETSC ERROR: Caught signal number 11 SEGV: Segmentation Violation, probably memory access out of range. I think these incompatible datatypes cause this error.

     
  • Billy Araujo
    Billy Araujo
    2009-10-08

    I have tested here and I can't reproduce your error. What happens if you run ./test heatp1 in the flow directory?

     
  • Allocating memory…<br>
    Memory allocated.<br>
    PETSC ERROR: ------------------------------------------------<br>
    PETSC ERROR: Caught signal number 11 SEGV: Segmentation Violation, probably memory access out of range

    Which version of petsc do you use and what configure options?

     
  • Billy Araujo
    Billy Araujo
    2009-10-09

    I used latest petsc 3.0.0p8 with options:

    ./config/configure.py -with-cc=gcc -with-fc=gfortran -download-f-blas-lapack=1 -download-mpich=1

    Are you able to run petsc tests (make test)? What system you have?

     
  • OS: Ubuntu 9.04 x86<br>
    Configure options:<br>
    ./config/configure.py -with-mpi-dir=/srv/shared -with-c++-support -with-c-support -prefix=/srv/shared/opt/petsc -with-blas-lapack-dir=/srv/shared -with-fftw=1 -with-fftw-include=/srv/shared/include -with-fftw-lib=/srv/shared/lib/libfftw3.a -with-scalar-type=complex -with-shared=1

    So mpicc and mpif90 are my compilers/linkers

    Yes I am able to run make test with no error

    Greetings

     
  • Billy Araujo
    Billy Araujo
    2009-10-09

    OK. the problem must be -with-scalar-type=complex and that is why it gives warnings about PETSc_sclalar.

     
  • Hi!

    My actual configure options:<br>
    ./config/configure.py -with-mpi-dir=/srv/shared -with-c++-support -with-c-support -prefix=/srv/shared/opt/petsc -with-blas-lapack-dir=/srv/shared

    Do I need shared libraries?

    Output of ./test heatp1 after memory allocated :<br>
    PETSC ERROR: Null argument, when expecting valid pointer!<br>
    PETSC ERROR: Null Object: Parameter # 3!<br>
    …<br>
    PETSC ERROR: VecWAXPY() line 675 in src/vec/vec/interface/rvector.c

    This error occours a few times, between the errors I get i.e. "Time = 1.000E+00"

    And I get an output, a chart

    The example "karman" now works!

    Greetings

     
  • Billy Araujo
    Billy Araujo
    2009-10-09

    Could you run the example with verbose like:
    ../OpenFVM lid fv 1
    ?
    Did you install gmsh 2.0 or above? I installed with apt-get install gmsh in ubuntu.
    It would also be nice to know the line where the error occurs.

    Thanks,

    x-flow.

     
  • Hi!

    I ran OpenFVM testlid fv 2, on two processors. I got:

    PETSC ERROR: -------------- Error Message ------------------------
    PETSC ERROR: Null argument, when expecting valid pointer!
    PETSC ERROR: Null Object: Parameter # 3!
    PETSC ERROR: ------------------------------------------------
    PETSC ERROR: Petsc Release Version 3.0.0, Patch 8, Fri Aug 21 14:02:12 CDT 2009
    PETSC ERROR: See docs/changes/index.html for recent updates.
    PETSC ERROR: See docs/faq.html for hints about trouble shooting.
    PETSC ERROR: See docs/index.html for manual pages.
    PETSC ERROR: ------------------------------------------------
    PETSC ERROR: OpenFVM on a linux-gnu named headnode1 by service Wed Oct 14 14:49:33 2009
    PETSC ERROR: Libraries linked from /srv/shared/opt/petsc/lib
    PETSC ERROR: Configure run at Fri Oct  9 15:26:14 2009
    PETSC ERROR: Configure options -with-mpi-dir=/srv/shared -with-c++-support -with-c-support -prefix=/srv/shared/opt/petsc -with-blas-lapack-dir=/srv/shared -with-shared=0
    PETSC ERROR: ------------------------------------------------
    PETSC ERROR: VecWAXPY() line 675 in src/vec/vec/interface/rvector.c

    I have installed the latest online available gmsh version, i think it's 2.4.2

     
  • Billy Araujo
    Billy Araujo
    2009-10-14

    Did you try
    mpirun -np 2 ../OpenFVM lid fv 2
    ?

     
  • I have started it with:

    mpiexec -np 2 OpenFVM testlid fv 2

    In addition to the errors I get:

    Matrix u Number of iterations: 8 Residual: +3.914364E-13 Time: +1.304400E-01

    Matrix v Number of iterations: 8 Residual: +2.254256E-13 Time: +1.308432E-01

     
  • With the example lid, I get the same errors in the same line

     
  • Billy Araujo
    Billy Araujo
    2009-10-19

    Hi,

    I couldn't reproduce these errors. Also I don't get the warnings from the compiler when building the sources complaining about incompatible parameters in MatGetValues ad MatSetValues. What compiler are you using?

     
  • Hi!

    After configure without -with-scalar-type=complex I don't get the MatGetValues and MatSetValues anymore.

    I have downloaded the latest svn version and don't get any errors at compiling or executing karman.

    Thanks!