Hi, James,
I am trying to use a makefile to generate a hybrid MPI/OpenMP application imex.exe. But I failed to do it and please see the following linking results:
U:\\imex\\bin\\opt64_hybridMPIOpenMP>clearmake -k -V imex.exe
insert_string.exe -d -v -c -f \\imex\\imex.lib\\simver.f > simver.f
ifort.exe /Qauto /cm /nologo /w /MT /Qfpp /DIFL_IA32 /DWIN_X64 /Qopenmp -c simver.f
Hello,
I'm running FMS application (http://www.gfdl.noaa.gov/fms) and some of the runs fail with the following error -
I gave tried to run parallel program using Intel MPI Library, but when I started like mpirun -hosts node01,node02 -np 36 ./prog it finished work without start witj an error message node02-ib0:56fa:17a5a440: 705 us(705 us): open_hca: get lid ERR for mlx4_0 port=2, err=Invalid argument node01-ib0:4bd1:ceb2a440: 590 us(590 us): open_hca: get lid ERR for mlx4_0 port=2, err=Invalid argument node02-ib0:56fa:17a5a440: 522 us(522 us): open_hca: getaddr_netdev ERROR: No such device. Is ib1 configured? node02-ib0:56fa:17a5a440: 7180 us(6475 us): open_hca: device mthca0 not found
HI, James,
I am testing Intel MPI on my computer. I do not know why the following MPI sample codes failed to run. after I made a minor revision on stoping message from sending from process 0 to iteself, then everything works well. Why is process 0 not allowed to send message to itself in Intel MPI and then receive it by itslef? I look forward to hearing from you. Thanks.
Dingjun
Please the original sample codes:
program vector
include 'mpif.h'
I've got a machine with 2 processors 8 cores each which gives me a total of 16 physical cores.I want to launch an application on the second processor, cores 8-15. The application uses one mpi process and 8 omp threads.Documentation suggests using I_MPI_PIN_DOMAIN to controll threads distribution. The value omp:compact pinned all threads on the first processor. But I didn't manage to find the way to move them on the second.I have also tried launching the program without any pinning options for MPI but using numactl instead.
After installing the Intel mpi rte on a 32 bit operating system, the smpd service is not installed. The service entry is missing from the windows services dialog and attempting to invoke mpiexec hangs. On the surface the installer runs to completion. Running smpd -install from the bin directory as a post-step will install the service successfully. This is supposed to happen automatically during the install and causes us significant problems.
Can you guys reproduce this? What is the solution?
-Adam
I'm interested as to why mpiexec hangs if I have the Intel MPI RTE installed, but not the service installed (you can get here by installing the RTE and then running smpd -uninstall).
A verbose runs shows this:
mpiexec
np 2 localonly notepad verbose
...\\smpd_get_full_path_name
....fixing up exe name:
'notepad' -> '(null)'
.../smpd_get_full_path_name
...handling executable:
notepad.exe
...\\smpd_get_next_host
....\\smpd_get_host_id
..../smpd_get_host_id
.../smpd_get_next_host
...\\smpd_get_next_host
Hi, Could someone tell me where I can get some examples for the hybrid Intel MPI and OpenMP codes? I want to study them. Thanks inadvance.
Dingjun
Hi, I am trying MPI 4.0.3 and would like to know which is the best way to cleanup the environment after an mpirun execution using Hydra finishes incorreclty. In previous versions (with MPDs) I was doing mpdallexit to ensure all daemons where down. I tried usingI_MPI_HYDRA_CLEANUP environment variable to generate the input file for the mpicleanup tool, but it seems this file is never created. This is how I am running mpi:
