Problem with NWChem 6.3 and Qlogic


Just Got Here
I have successfully compiled NWChem 6.3 on our cluster with Qlogic infiniband using ARMCI_NETWORK=OPENIB. However, SCF calculations never converge, they tend towards converging, but then at some point they go haywire. Re-compiling with ARMCI_NETWORK=SOCKETS solves the problem (but of course, the calculations are now slower). My build environment is as follows:
setenv ARMCI_NETWORK OPENIB
setenv IB_INCLUDE /opt/ofed/1.5.4.1/include
setenv IB_HOME /opt/ofed/1.5.4.1
setenv IB_LIB /opt/ofed/1.5.4.1/lib64/

setenv NWCHEM_TOP /home/cperry/codes/nwcheminst/nwchem-6.3
setenv NWCHEM_MODULES "qm"
setenv LARGE_FILES TRUE
setenv USE_NOFSCHECK TRUE
setenv LIB_DEFINES -DDFLT_TOT_MEM=259913728
setenv NWCHEM_TARGET LINUX64
setenv MPI_LOC /opt/openmpi
setenv MPI_INCLUDE $MPI_LOC/include
setenv MPI_LIB $MPI_LOC/lib
setenv MPI_BIN $MPI_LOC/bin
setenv USE_MPI y
setenv LIBMPI "-lmpi_f90 -lmpi_f77 -lmpi -ldat -lrt -lnsl -lutil -lm -ldl -lm -Wl,--export-dynamic -lrt -lnsl -lutil -lm -ldl -L/opt/ofed/1.5.4.1/lib64 -libverbs "
setenv MKLROOT /home/cperry/intel/composer_xe_2013.1.117/mkl/lib/intel64
setenv BLASOPT " -Wl,--start-group  $MKLROOT/libmkl_intel_ilp64.a $MKLROOT/libmkl_sequential.a $MKLROOT/libmkl_core.a -Wl,--end-group -lpthread -lm"

rm -rf $NWCHEM_TOP
tar -zxvf Nwchem-6.3-src.2013-05-17.tar.gz
mv nwchem-6.3-src.2013-05-17 nwchem-6.3
cd $NWCHEM_TOP/src
make FC=ifort  nwchem_config
make F70=ifort F90=ifort CC=icc CXX=icpc FC=ifort


Any advice on this issue would be much appreciated.

Kind regards
Chris

Forum Vet
Chris,
Unfortunately, we have not had the chance to test the OPENIB ARMCI network on QLogic ourselves yet.
In the meantime, using ARMCI_NETWORK=MPI-TS could give you better performances.
If we manage to get access to QLogic hardware or if we get reports of successful QLogic builds, we will keep you posted.
Cheers, Edo

Just Got Here
Thanks Edo


Forum >> NWChem's corner >> Compiling NWChem