ARMCI infiniband issues x86 intel
Clicked A Few Times
10:50:51 PM PDT - Tue, Aug 2nd 2011
hi
I trying to install NWchem in dynamo.I trying to run using Infinitband OPENIB.
The run fails at test for OPENIB.

Error Message:
But during my compilation of NWChem modules, i am getting the following error message:

checking for TARGET 64bit-ness... LINUX64

checking whether we are cross compiling... no

configure: searching for OPENIB...

checking infiniband/verbs.h usability... yes

checking infiniband/verbs.h presence... yes
checking for infiniband/verbs.h... yes
checking for library containing ibv_open_device... no
configure: error: test for ARMCI_NETWORK=OPENIB failed
make[1]: *** [build/config.status] Error 1

I feel the error is due to OPENIB?.

my enviroment variables
setenv LARGE_FILES TRUE
setenv LIB_DEFINES -DDFLT_TOT_MEM=16777216
setenv TCGRSH /usr/bin/ssh
setenv NWCHEM_TOP /home/ssk87/nwchem-src-2011-Jun-20
setenv NWCHEM_TARGET LINUX64
setenv IFORTHOME /opt/intel/fce/10.1.018
setenv ICCHOME /opt/intel/cce/10.1.018
source $IFORTHOME/bin/ifortvars.csh

setenv USE_MPI y
setenv MPI_LOC /usr/mpi/intel/mvapich-1.0.1 #location of mpich
setenv MPI_LIB $MPI_LOC/lib
setenv MPI_INCLUDE $MPI_LOC/include
setenv LIBMPI '-libfmpich -libmpich -libpmpich'

setenv IB_HOME /usr
setenv IB_INCLUDE /usr/include/infiniband
setenv IB_LIB /usr/lib64

setenv IB_LIB_NAME '-libpthread -librdmacm -libibverbs -libibumad -librt'
setenv ARMCI_NETWORK MELLANOX


I am getting the same error even for mellanox

can you have idea to fix it?

Forum Vet
1:55:23 PM PDT - Wed, Aug 3rd 2011
Could you make sure the library file libibverbs is located in /usr/lib64. This is the library that actually has the ibv_open_device function in it.

Bert



[QUOTE=Ssk87 Aug 3rd 5:50 am]hi
I trying to install NWchem in dynamo.I trying to run using Infinitband OPENIB.
The run fails at test for OPENIB.

Error Message:
But during my compilation of NWChem modules, i am getting the following error message:

checking for TARGET 64bit-ness... LINUX64

checking whether we are cross compiling... no

configure: searching for OPENIB...

checking infiniband/verbs.h usability... yes

checking infiniband/verbs.h presence... yes
checking for infiniband/verbs.h... yes
checking for library containing ibv_open_device... no
configure: error: test for ARMCI_NETWORK=OPENIB failed
make[1]: *** [build/config.status] Error 1

I feel the error is due to OPENIB?.

my enviroment variables
setenv LARGE_FILES TRUE
setenv LIB_DEFINES -DDFLT_TOT_MEM=16777216
setenv TCGRSH /usr/bin/ssh
setenv NWCHEM_TOP /home/ssk87/nwchem-src-2011-Jun-20
setenv NWCHEM_TARGET LINUX64
setenv IFORTHOME /opt/intel/fce/10.1.018
setenv ICCHOME /opt/intel/cce/10.1.018
source $IFORTHOME/bin/ifortvars.csh

setenv USE_MPI y
setenv MPI_LOC /usr/mpi/intel/mvapich-1.0.1 #location of mpich
setenv MPI_LIB $MPI_LOC/lib
setenv MPI_INCLUDE $MPI_LOC/include
setenv LIBMPI '-libfmpich -libmpich -libpmpich'

setenv IB_HOME /usr
setenv IB_INCLUDE /usr/include/infiniband
setenv IB_LIB /usr/lib64

setenv IB_LIB_NAME '-libpthread -librdmacm -libibverbs -libibumad -librt'
setenv ARMCI_NETWORK MELLANOX


I am getting the same error even for mellanox

can you have idea to fix it?

Clicked A Few Times
1:17:15 PM PDT - Fri, Aug 5th 2011
ARMCI DASSERT fail error
Even I added -libverbs, I am still getting the following running error. Could anyone help on this.

Last System Error Message from Task 0:: Inappropriate ioctl for device
0:0::: 9999
(rank:0 hostname:dynamo.iprt.iastate.edu pid:30714):ARMCI DASSERT fail. ../../ga-5-0/armci/src/armci.c:ARMCI_Error():276 cond:0
 0: ARMCI aborting 9999 (0x270f).

 0: ARMCI aborting 9999 (0x270f).
system error message: Invalid argument

Forum Vet
5:30:15 PM PDT - Mon, Aug 8th 2011
Lets take this step by step and start from a clean build.

1. Do a make realclean in nwchem-src/2011-Jun-20/src/

Set the following environment variables:

setenv LARGE_FILES TRUE
setenv NWCHEM_TOP /home/ssk87/nwchem-src-2011-Jun-20
setenv NWCHEM_TARGET LINUX64
setenv IFORTHOME /opt/intel/fce/10.1.018
setenv ICCHOME /opt/intel/cce/10.1.018
source $IFORTHOME/bin/ifortvars.csh
setenv USE_MPI y
setenv MPI_LOC /usr/mpi/intel/mvapich-1.0.1 #location of mpich
setenv MPI_LIB $MPI_LOC/lib
setenv MPI_INCLUDE $MPI_LOC/include
setenv LIBMPI '-libfmpich -libmpich -libpmpich'
setenv IB_HOME /usr
setenv IB_INCLUDE /usr/include/infiniband
setenv IB_LIB
setenv ARMCI_NETWORK OPENIB

Unset any other ones you set before, such as IB_LIB_NAME.

2. make nwchem_config NWCHEM_MODULES=all

3. make

If you get the OPENIB error you had before, locate the libibverbs.so library. Point IB_LIB to the directory in which this library resides and do a make again.

If it builds, run NWChem without an input deck (does a dummy run).

Bert

Clicked A Few Times
12:27:53 PM PDT - Tue, Aug 9th 2011
ARMCI error
Thanks Bert.
I used only libibverbs.so
I unset all the other libraries

The nwchem lib builds

But I am getting the following error by running without giving the input deck

[proxy:0:0@exalted.iprt.iastate.edu] HYDU_create_process (./utils/launch/launch.c:68): execvp error on file np (No such file or directory)
=========================================================================
= BAD TERMINATION OF ONE OF YOUR APPLICATION PROCESSES
= EXIT CODE: 65280
= CLEANING UP REMAINING PROCESSES
= YOU CAN IGNORE THE BELOW CLEANUP MESSAGES
=====================================================================================Italic text
With Input deck:

**********

*   0: pre_mkfrg failed 9999

**********
''''0:0::: 9999
 (rank:0 hostname:exalted.iprt.iastate.edu pid:12091):ARMCI DASSERT fail. ../../ga-5-0/armci/src/armci.c:ARMCI_Error():276 cond:0
Last System Error Message from Task 0:: Inappropriate ioctl for device
 0: ARMCI aborting 9999 (0x270f).

 0: ARMCI aborting 9999 (0x270f).
system error message: Illegal seekItalic text
 Italic text''Italic text
 Ie My previous error message gets listed again.

can you suggest me any other methods to fix this error.

Regards
Selva


Forum >> NWChem's corner >> Running NWChem