[mpich-discuss] MPI-IO of multidimensional arrays - issue reported by valgrind
Turlough Downes
turlough.downes at dcu.ie
Tue Feb 23 05:54:24 CST 2010
Hi All,
I'm using MPI-IO to let multiple processes write their own subarrays to a single file which will then contain a global array - this is for an astrophysical fluid dynamics simulation code which is parallelised using domain decomposition.
To do this I'm using MPI_Type_create_subarray(), MPI_File_open(), MPI_File_set_view() and MPI_File_write_all(). My problem is that my code is occasionally hanging on output on a Blue Gene/P system. To study this a bit more I tried running the code on my workstation. When I run this code on my local system using mpich2 (version 1.2.1) on a CentOS 8 core system using valgrind I find that the following error being reported:
==10315== Conditional jump or move depends on uninitialised value(s)
==10315== at 0x4C86DE8: (within /usr/lib64/mpich2/lib/libmpich.so.1.2)
==10315== by 0x4C88F0E: ADIOI_GEN_WriteStridedColl (in /usr/lib64/mpich2/lib/libmpich.so.1.2)
==10315== by 0x4D678C6: MPIOI_File_write_all (in /usr/lib64/mpich2/lib/libmpich.so.1.2)
==10315== by 0x4D679B8: PMPI_File_write_all (in /usr/lib64/mpich2/lib/libmpich.so.1.2)
==10315== by 0x41B6B4: hydra_write_single_binary_file (out_binary_single_file.c:101)
where the line number 101 in my code is the line which calls MPI_File_write_all(). My question is this: is this issue being reported by valgrind something to worry about? It is happening somewhere inside the MPICH2 library itself so this isn't something I myself have delved into too much.
I know that any error like this may not have an analogue in the version of MPI used on Blue Gene/P systems, but I thought it might be worth chasing down a little.
Thanks for any help.
Turlough
--
Dr Turlough Downes
Senior Lecturer,
School of Mathematical Sciences,
Dublin City University,
Glasnevin,
Dublin 9,
Ireland.
Tel: +353 1 700 5270 (DCU)
+353 1 662 1333 Ext 338 (DIAS)
FAX: +353 1 700 5786 (DCU)
+353 1 524 2302 (DIAS)
URL: http://www.dcu.ie/~downest/
More information about the mpich-discuss
mailing list