Re: [AMBER] problem of CUDA MPI

From: Albert <mailmd2011.gmail.com>
Date: Fri, 21 Jun 2019 17:27:52 +0200

do you have further suggestions?

thanks a lot.

On Thu, Jun 20, 2019 at 2:39 PM Albert <mailmd2011.gmail.com> wrote:

> The CPU MPI compiling is all right.
>
> I am using openmpi-4.0.1+gfortran-7
>
> > mpif90 -show
> gfortran -I/soft/openmpi-4.0.1/include -pthread
> -I/soft/openmpi-4.0.1/lib64 -Wl,-rpath -Wl,/soft/openmpi-4.0.1/lib64
> -Wl,--enable-new-dtags -L/soft/openmpi-4.0.1/lib64 -lmpi_usempif08
> -lmpi_usempi_ignore_tkr -lmpi_mpifh -lmpi
>
> thanks again.
>
>
> On Thu, Jun 20, 2019 at 7:49 AM David Case <david.case.rutgers.edu> wrote:
>
>> On Thu, Jun 20, 2019, Albert wrote:
>> >
>> >./configure -mpi -cuda gnu
>> >
>> >make[2]: Entering directory '/soft/amber18/AmberTools/src/etc'
>> >mpif90 -DBINTRAJ -DEMIL -DMPI -o numprocs numprocs.F90
>> >mpif-sizeof.h:66:41:
>> >
>> >Error: Array specification at (1) has more than 7 dimensions
>> >mpif-sizeof.h:72:41:
>> >
>> >Error: Array specification at (1) has more than 7 dimensions
>> >mpif-sizeof.h:78:41:
>> >
>> >Error: Array specification at (1) has more than 7 dimensions
>> >mpif-sizeof.h:84:41:
>> >
>> >Error: Array specification at (1) has more than 7 dimensions
>> >mpif-sizeof.h:90:41:
>> >
>> >Error: Ambiguous interfaces in generic interface 'pmpi_sizeof' for
>> >‘pmpi_sizeof_real64_r14’ at (1) and ‘pmpi_sizeof_real64_r15’ at (2)
>>
>> What version of MPI are you running? What version of gfortran?
>> And, what does 'mpif90 -show' return?
>>
>> ....thanks...dac
>>
>> _______________________________________________
>> AMBER mailing list
>> AMBER.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber
>>
>
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Jun 21 2019 - 02:30:02 PDT
Custom Search