Hi, Tim!
Thanks for your reply.
I changed MPI_REAL by MPI_DOUBLE_PRECISION, but it still shows the same warnings. I want understand why these warnings appears in pairs. Look, the variables called local_YG and YG are defined as follows:
integer, allocatable :: local_YG, YG
and at line 1109 I used MPI_INTEGER to MPI_REDUCE datatype. Is it right?
The others variables, local_AccNa and AccNa are defined as follows:
real(8), allocatable :: local_AccNa(:), AccNa(:)
and at line 1068 I used MPI_REAL to MPI_REDUCE datatype. It seems obvious, right?
> "In file pmixpoliseg_v15.f90:1109
>
> call MPI_REDUCE(local_YG, YG, NG, MPI_INTEGER, MPI_SUM, mestre, MPI_COMM_
> 1
> In file pmixpoliseg_v15.f90:1068
>
> call MPI_REDUCE(local_AccNa, AccNa, ncomp*2, MPI_REAL, MPI_SUM, mestre, M
> 2
> Warning (155): Inconsistent types (INTEGER(4)/REAL(8)) in actual argument lists at (1) and (2)"
>
Well, they names are different, they types are different, so, why the warnings? This happens only with MPI calls. The same algorithm compiled without MPI routines (sequential version) works fine!
Thanks for any help.
Jose Luiz
----- Mensagem original -----
De: "Tim Prince" <TimothyPrince_at_[hidden]>
Para: "General LAM/MPI mailing list" <lam_at_[hidden]>
Enviadas: Segunda-feira, 23 de Junho de 2008 06h58min25s (GMT-0300) Auto-Detected
Assunto: Re: LAM: Problems with Fortran 90
Jose Luiz de Souza Gomes wrote:
> Hi!
>
> I have a Fortran 90 algorithm that runs sequentially and I trying to run it in parallel mode. I'm using LAM 7.0.4 and g95 Fortran compiler. When I compile it, the messages are (a lot of them):
>
> "In file pmixpoliseg_v15.f90:1109
>
> call MPI_REDUCE(local_YG, YG, NG, MPI_INTEGER, MPI_SUM, mestre, MPI_COMM_
> 1
> In file pmixpoliseg_v15.f90:1068
>
> call MPI_REDUCE(local_AccNa, AccNa, ncomp*2, MPI_REAL, MPI_SUM, mestre, M
> 2
> Warning (155): Inconsistent types (INTEGER(4)/REAL(8)) in actual argument lists at (1) and (2)"
>
If you engage your browser and look up the subject, you will see that MPI
uses MPI_DOUBLE_PRECISION rather than attempting to overlay various sizes
of MPI_REAL. You really should find a working example parallel to whatever
you are trying to do.
_______________________________________________
This list is archived at http://www.lam-mpi.org/MailArchives/lam/
--
Esta mensagem foi verificada pelo sistema de antiv�rus e
acredita-se estar livre de perigo.
--
Esta mensagem foi verificada pelo sistema de antivírus e
acredita-se estar livre de perigo.
|
