[MPICH] standard input limited to 4096 bytes?

[Martin Kleinschmidt]

On Di, 17 Okt 2006, Zhengting Gan wrote:

>Hi Martin,
>
>I think you should add MPI_finalize() in your code.
>

Thanks for the idea, but it does not help:
added MPI_finalize as last command before end - program hangs after 
write(*,*) '0' 
and before
write(*,*)'0a'


>
>----- Original Message ----- 
>From: "Martin Kleinschmidt" <mk at theochem.uni-duesseldorf.de>
>To: <mpich-discuss at mcs.anl.gov>
>Sent: Tuesday, October 17, 2006 10:56 AM
>Subject: [MPICH] standard input limited to 4096 bytes?
>
>
>>
>>Hi,
>>
>>I'm having problems with standard input.
>>
>>the following program:
>>
>>     program  test
>>     
>>     include 'mpif.h'
>>     integer myid, nproc
>>
>>     write(*,*) '0'
>>     call MPI_init(MPIerr)
>>     write(*,*) '0a'
>>     call MPI_comm_rank(MPI_comm_world,myid, MPIerr)
>>     write(*,*) '0b'
>>     call MPI_comm_size(MPI_comm_world,nproc,MPIerr)
>>     write(*,*) '0c'
>>
>>     end
>>
>>runs correctly when started with
>>mpiexec -n2 ./a.out < input_file_smaller_than_4097_bytes
>>
>>but hangs in call MPI_init when called with
>>mpiexec -n2 ./a.out < input_file_larger_than_4096_bytes
>>
>>note that the program does NOT access standard input!
>>
>>I'm using mpich2-1.0.4p1, intel compilers 9.x, fedora core2.
>>
>>Is this a general problem of mpich? 
>>Is there any solution?
>>In my "real" program, I regularly need standard input from files larger
>>than 4096 bytes.
>>
>>
>>thanks!
>>
>>  ...martin
>>
>>
>>

-- 
Statistiken sind hinterhaeltige Begleiter. Jedenfalls, wenn man sie
ueberinterpretiert. Zum Beispiel Slowenien. (Er ueberlegt.) Zu Slowenien
habe ich gar keine Statistik, die gibt es erst seit ein paar Jahren.
(Ein Eurosport-Kommentator)