A note (mainly for other MPICH devs): I have created a trac account for Harald and sent him his account information off-list. -Dave On Dec 19, 2012, at 9:23 AM CST, Harald Klimach wrote:

Hi there,

I installed MPICH2 3.0rc1 using GCC 4.7.2 and found a problem with MPI_Type_create_f90_real. It is very similar to the problem I reported in OpenMPI under: https://svn.open-mpi.org/trac/ompi/ticket/3432

Running the following program:

program test_mpi use mpi implicit none

integer, parameter :: rk_prec = 15 integer, parameter :: rk = selected_real_kind(rk_prec)

integer :: rk_mpi integer :: iError real(kind=rk) :: a_real real(kind=rk) :: res

call MPI_Init(iError) call mpi_type_create_f90_real(rk_prec, MPI_UNDEFINED, rk_mpi, iError) write(*,*) 'MPI_REAL8:', MPI_REAL8 write(*,*) 'MPI_DOUBLE_PRECISION:', MPI_DOUBLE_PRECISION write(*,*) 'type_create_f90_real:', rk_mpi a_real = 1.0 call MPI_Reduce(a_real, res, 1, rk_mpi, MPI_MIN, 0, MPI_COMM_WORLD, iError) call MPI_Finalize(iError)

end program test_mpi

Results in the following output and error:

MPI_REAL8: 1275070505 MPI_DOUBLE_PRECISION: 1275070495 type_create_f90_real: -1946157049 Fatal error in PMPI_Reduce: Invalid MPI_Op, error stack: PMPI_Reduce(1217)........: MPI_Reduce(sbuf=0x7fff43637108, rbuf=0x7fff436370f8, count=1, dtype=USER<f90_real>, MPI_MIN, root=0, MPI_COMM_WORLD) failed MPIR_MINF_check_dtype(71): MPI_Op MPI_MIN operation not defined for this datatype

Further the quadruple precision data type is not properly supported: MPI_Type_create_f90_real will complain about wrong exponent, regardless of its value, while MPI_REAL16 is available and the Reduction operation does not complain, it produces wrong results. Again there is a similar problem with OpenMPI, as reported under: https://svn.open-mpi.org/trac/ompi/ticket/3433

Thanks a lot, Harald Klimach _______________________________________________ discuss mailing list [email protected] To manage subscription options or unsubscribe: https://lists.mpich.org/mailman/listinfo/discuss