Subject: mpi_dec_alpha
From: cmn (cmn@senamhi.gob.pe)
Date: Fri Jan 19 2001 - 15:54:21 MST
Dear ccm-users:
WE NEED HELP...
We have are running CCM3 (version ccm.10.11.brnchT.366physics.4) on a Compaq ES40 Alpha 4 processor machine under Tru64 Unix. We have run the model successfully on 1 processor.
We are now trying to run in multi-processor mode. The MPI package was downloaded from Compaq
and successfully installed and tested. We made the appropriate changes to the run scripts and
have checked the include files (param.h, misc.h, preproc.h) in run/obj and that the mpi include files
are in the correct locations, and we have added the appropriate flags to the run command and compile
commands. The model compiles without error and begins to run. However it dies - apparently during the initialization of LSM. This appears to occur at the first call to MPIGATHER in LSMRDINIT.F when MASTERPROC is TRUE. We get the following message -
[2] Aborting program !
[1] Aborting program !
[3] Aborting program !
UMP_W_CLOSE event arrived while handling unexpected data[ 0] MPID Die - ump2chck.c:461 "ump_wait failure" (-1)
[ 0] MPID Die - ump2chck.c:104 "Found too big a header packet in channel" (4192)
[ 0] MPID Die - ump2chck.c:104 "Found too big a header packet in channel" (4192)
[ 0] MPID Die - ump2chck.c:104 "Found too big a header packet in channel" (4192)
and many, many of these lines continue to be printed ......
the size of the data we are gathering is 131x16
mpir8 is the correct value of 11 mpicom=91
We have already tried making the memory and segment sizes much larger for ipc and
for ordinary memory. the stacksize is unlimited.
Any help with this will be immensily appreciated -
Mauricio Carrillo - SENAMHI (Peruvian Meteorological Service)
This archive was generated by hypermail 2b27 : Tue Jan 22 2002 - 11:15:39 MST