[mpich-discuss] Error when using MPICH2: Fatal error in PMPI_Allgather

[Rajeev Thakur]

You can't pass the same buffer as the send and receive buffer to MPI_Allgather. You need to either pass different buffers or use MPI_IN_PLACE as described in the MPI 2.2 standard document.

Rajeev

On Sep 11, 2011, at 11:06 PM, Miao Yu wrote:

> Hi MPI Developers,
>  
> I have a problem when using the MPICH2 with my research model. The system on our computer is Utunbu. When I use the MPICH2-1.2.1.1-4 with 4 CPUs to run my model, I got the error messages as follow:
> ============================================================================
> Fatal error in PMPI_Allgather: Invalid buffer pointer, error stack:
> PMPI_Allgather(958): MPI_Allgather(sbuf=0x23c625f0, scount=1, MPI_INTEGER, rbuf=0x23c625f0, rcount=1, MPI_INTEGER, MPI_COMM_WORLD)
> failed
> PMPI_Allgather(931): Buffers must not be aliased
> Fatal error in PMPI_Allgather: Internal MPI error!, error stack:
> PMPI_Allgather(958)......: MPI_Allgather(sbuf=0x10c1cf18, scount=1, MPI_INTEGER, rbuf=0x10c1cf10, rcount=1, MPI_INTEGER, MPI_COMM_
> WORLD) failed
> MPIR_Allgather_impl(805).:
> MPIR_Allgather(766)......:
> MPIR_Allgather_intra(144):
> MPIR_Localcopy(357)......: memcpy arguments alias each other, dst=0x10c1cf18 src=0x10c1cf18 len=4
> Fatal error in PMPI_Allgather: Internal MPI error!, error stack:
> PMPI_Allgather(958)......: MPI_Allgather(sbuf=0x1177ff14, scount=1, MPI_INTEGER, rbuf=0x1177ff10, rcount=1, MPI_INTEGER, MPI_COMM_
> WORLD) failed
> MPIR_Allgather_impl(805).:
> MPIR_Allgather(766)......:
> MPIR_Allgather_intra(144):
> MPIR_Localcopy(357)......: memcpy arguments alias each other, dst=0x1177ff14 src=0x1177ff14 len=4
> Fatal error in PMPI_Allgather: Internal MPI error!, error stack:
> PMPI_Allgather(958)......: MPI_Allgather(sbuf=0x11256f1c, scount=1, MPI_INTEGER, rbuf=0x11256f10, rcount=1, MPI_INTEGER, MPI_COMM_
> WORLD) failed
> MPIR_Allgather_impl(805).:
> MPIR_Allgather(766)......:
> MPIR_Allgather_intra(144):
> MPIR_Localcopy(357)......: memcpy arguments alias each other, dst=0x11256f1c src=0x11256f1c len=4
> rank 3 in job 19  ocean_53883   caused collective abort of all ranks
>   exit status of rank 3: return code 1
> ============================================================================
>  
> I’ve changed other versions of MPICH2, such as version 1.4, 1.2, 1.1, 1.1.1, found that I can’t fix this problem.
> Could anyone please give me any suggestion or clue? I will appreciate for that.
>  
> Miao
> _______________________________________________
> mpich-discuss mailing list     mpich-discuss at mcs.anl.gov
> To manage subscription options or unsubscribe:
> https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss