[mpich-discuss] MPI_MINLOC for MPI_2DOUBLE_PRECISION

[Dave Goodell]

Can you send us your config.log?  MPICH2 does generally support  
MPI_2DOUBLE_PRECISION for MINLOC/MAXLOC, but only when the C and  
Fortran compilers agree about certain things.  If we have your  
config.log it will help us determine why they aren't agreeing in your  
environment.

-Dave

On Apr 16, 2010, at 10:53 AM, Fabricio Harter wrote:

> Dear All,
>
> I built WRFVAR3.2 using MPICH2 on a cluster Dual-Core AMD  
> Opteron(tm) Processor 2210 with Intel ifort compiler. The program  
> builds, but when I run it reaches a point where MPI_Reduce aborts as  
> the following message:
>
>
> DYNAMICS OPTION: Eulerian Mass Coordinate
>     alloc_space_field: domain            1,    105610432 bytes  
> allocated
>  WRF TILE   1 IS      1 IE     45 JS      1 JE     30
>  WRF NUMBER OF TILES =   1
> [cli_0]: aborting job:
> Fatal error in MPI_Reduce: Invalid MPI_Op, error stack:
> MPI_Reduce(850).............: MPI_Reduce(sbuf=0x37420d8,  
> rbuf=0x37420e8, count=1, MPI_2DOUBLE_PRECISION, MPI_MINLOC, root=0,  
> comm=0x84000000) failed
> MPIR_MINLOC_check_dtype(215): MPI_Op MPI_MINLOC operation not  
> defined for this datatype
>
>
> The MPI standards for both MPI 2.1 and MPI 2.2 state that MPI_Reduce  
> are supported for that (p. 165 for 2.1 and p. 168 for 2.1)
>
> Is there a version of MPICH that actually support MPI_MINLOC for  
> MPI_2DOUBLE_PREC ?
>
>
> Thanks in advance,
>
> Fabricio Harter.
>
>
>
>
>
> _______________________________________________
> mpich-discuss mailing list
> mpich-discuss at mcs.anl.gov
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss