[Trilinos-Users] Ctest Pliris error

Joe Kotulski jdkotul at sandia.gov
Mon Aug 31 12:46:57 MDT 2009



                Jason,

                The variables MPI_ANY_SOURCE and MPI_ANY_TAG used in 
this receive
                are MPI variables defined in mpi.h and are not set in 
the code.  The only possibility
                I see for this happening is for some reason these are 
being overwritten in memory and are no longer
                valid message tags.  This code has been checked for 
memory issues using valgrind for this
                size of problem.  So, unfortunately, I can't give a 
definitive reason fro this failure.


                                                                         
                                       Joe Kotulski
                                                                         
                                       Sandia National Labs

Jason Slemons wrote:
>
> I’ve just tested my new trilinos build with Ctest –R ‘Pliris_’ and 
> found one error, a very large tag:
>
>  
>
> Fatal error in MPI_Irecv: Invalid tag, error stack:
>
> MPI_Irecv(144): MPI_Irecv(buf=0x850720, count=4000, MPI_CHAR, 
> src=MPI_ANY_SOURC\
>
> E, tag=134218727, MPI_COMM_WORLD, request=0x7fffffffbea8) failed
>
> MPI_Irecv(97).: Invalid tag, value is 134218727
>
> aborting job:
>
>  
>
> I have looked for this tag, indeed any MPI_Irecv in any Pliris 
> directory but haven’t been able to find it. I  think the tag is too 
> big, but where is the source for this test?
>
>  
>
> --Jason Slemons
>
>  
>
> I’ve put the whole run below:
>
>  
>
> 460/835 Testing Pliris_vector_random_MPI_4
>
> Test command: /usr/bin/aprun -n 4 
> /lus/nid00008/slemons/trilinos-090603a-Source\
>
> /6.BUILD/packages/pliris/test/vector_random/Pliris_vector_random.exe 
> 1000 2
>
> Test timeout computed to be: 600
>
> proc 1 is alive of   4 Processors
>
> proc 3 is alive of   4 Processors
>
> proc 0 is alive of   4 Processors
>
> proc 2 is alive of   4 Processors
>
>  Matrix Size 1000
>
>  Processors in a row  2
>
>  ---- Building Pliris solver ----
>
>  ------ PARALLEL Distribution Info for : ---------
>
>  ------ PARALLEL Distribution Info for : ---------
>
>  ------ PARALLEL Distribution Info for : ---------
>
>  ------ PARALLEL Distribution Info for : ---------
>
>    Processor  2
>
>    Processor  3
>
>    Processor  1
>
>     my rows  500
>
>     my rows  500
>
>     my rows  500
>
>     my cols  500
>
>     my cols  500
>
>     my cols  500
>
>     my rhs  1
>
>     my rhs  0
>
>     my rhs  0
>
>     my first col  1
>
>     my first col  501
>
>     my first row  1
>
>     my first row  501
>
>     my first col  501
>
>     my_row  0
>
>     my_row  1
>
>     my first row  501
>
>     num procs row   2
>
>    Processor  0
>
>     num procs row   2
>
>     my_col  1
>
>     my rows  500
>
>     my_col  0
>
>     my cols  500
>
>     my rhs  1
>
>     my first col  1
>
>     my first row  1
>
>     my_row  0
>
>     num procs row   2
>
>     my_col  0
>
>     my_row  1
>
>     num procs row   2
>
>     my_col  1
>
>  ****   Setting Random Matrix    ****
>
>  ****   Creating RHS   ****
>
>  ****   Packing RHS in Matrix   ****
>
>  ****   Beginning Matrix Solve   ****
>
> Application 4212915 exit codes: 1
>
> Application 4212915 raborting job:
>
> Fatal error in MPI_Irecv: Invalid tag, error stack:
>
> MPI_Irecv(144): MPI_Irecv(buf=0x850720, count=4000, MPI_CHAR, 
> src=MPI_ANY_SOURC\
>
> E, tag=134218727, MPI_COMM_WORLD, request=0x7fffffffbea8) failed
>
> MPI_Irecv(97).: Invalid tag, value is 134218727
>
> aborting job:
>
> ------------------------------------------------------------------------
>
> _______________________________________________
> Trilinos-Users mailing list
> Trilinos-Users at software.sandia.gov
> http://software.sandia.gov/mailman/listinfo/trilinos-users
>   






More information about the Trilinos-Users mailing list