Hi,
Are you able to run (successfully) scalapack and blacs tests?
Maybe the problem has nothing to do with siesta, it looks like
a communication problem.
Regards,
Eduardo
On 23/09/2008, at 4:30, M Bharat Kumar wrote:
Hello All,
When trying to run simulations using siesta on our new cluster, I am
not able to complete the simulations because of mpi errors.
I tried both siesta 2.0 and siesta 2.0.1. The compilers are icc and
ifort. I tried using scalapack and blacs routines suppplied in intel
mkl libraries as well as
self-compiled scalapack and blacs. Also I tried mvapich2-1.03,
mvapich2-1.2RC2, and intel mpi 3.1.
Whatever combination I am trying, the program is stopping in the
middle of simulation with errors like
Warning! Rndv Receiver is receiving (7168 < 32128) less than as
expected
Warning! Rndv Receiver is receiving (7168 < 32128) less than as
expected
Fatal error in MPI_Bcast:
Message truncated, error stack:
MPI_Bcast(1144)........................: MPI_Bcast(buf=0xc92c9d0,
count=1, dtype=USER<vector>, root=0, comm=0xc4000006) failed
MPIR_Bcast(228)........................:
MPIDI_CH3U_Post_data_receive_found(439): Message from rank 0 and tag
2 truncated; 32128 bytes received but buffer size is 7168
Fatal error in MPI_Bcast:
Message truncated, error stack:
MPI_Bcast(1144)........................: MPI_Bcast(buf=0x14fb740,
count=1, dtype=USER<vector>, root=0, comm=0xc4000006) failed
MPIR_Bcast(228)........................:
MPIDI_CH3U_Post_data_receive_found(439): Message from rank 0 and tag
2 truncated; 32128 bytes received but buffer size is 7168
rank 11 in job 4 master_45552 caused collective abort of all ranks
exit status of rank 11: killed by signal 9
Thinking that the error may be due to stack size, I set ulimit-s
256000 and also tried using -heap_arrays flag, but with no success.
Some one please point out the source of error.
Thanks,
Bharat