How to engage more then one processor?
Just Got Here
10:02:13 PM PDT - Thu, May 9th 2013
Hi,

I am new to NWChem.
I installed NWChem via "Ubuntu Software Center" on my laptop with i7 processor.
It works fine, but "System Monitor" is showing just one processor working during calculations.

Is it possible to engage other processors to work in parallel with this type of installation and how?

Thanks

Forum Vet
8:40:15 AM PDT - Fri, May 10th 2013
Ubuntu version
Mehcwn
What Ubuntu version are you using?
"cat /etc/lsb-release" shows this kind of information

Thanks, Edo

Forum Vet
9:52:07 AM PDT - Fri, May 10th 2013
ubuntu package not build with MPI
Mehcwn
I have just installed the Ubuntu 12.04 package and I have noticed that it's a serial nwchem binary,
not linked against MPI.

If you wish to avoid dealing with installation with the source code, you might want to install the Debian pkgs,
that 1) are linked against MPI and 2) are compatible (just tested) with the Ubuntu 12.04 software stack (MPI included).

Cheers, Edo

Just Got Here
12:29:01 PM PDT - Fri, May 10th 2013
Hi Edo,

Thank you for your reply.
I am running Ubuntu 12.04.2 LTS. I've removed previous version of NWChem and I've installed NWChem 6.1.1 from the source.I had no problem installing it (no errors appeared etc.) but now, silly problem occurred!


Below you can find environment variables and code compilation procedure that I've set-up/performed

Thanks.

export NWCHEM_TOP=/home/ant/programs/nwchem/nwchem-6.1.1-src
export NWCHEM_TARGET=LINUX64
export NWCHEM_MODULES=all
export USE_MPI=y
export USE_MPIF=y
export USE_MPIF4=y
export MPI_LOC=/usr
export MPI_LIB=$MPI_LOC/lib
export MPI_INCLUDE=$MPI_LOC/include
export LIBMPI="-lmpi_f90 -lmpi_f77 -lmpi -ldl -Wl,--export-dynamic -lnsl -lutil"
export LARGE_FILES=TRUE
export USE_NOFSCHECK=TRUE
export LIB_DEFINES=-DDFLT_TOT_MEM=16777216
export FC=gfortran
export CC=gcc
cd $NWCHEM_TOP/src
make nwchem_config
make FC=gfortran >& make.log

Forum Vet
5:43:59 PM PDT - Fri, May 10th 2013
Quote:Mehcwn May 10th 11:29 am
Hi Edo,




Please have a look at the link below for how to run NWChem with MPI

http://nwchemgit.github.io/index.php/Release61:Running#Parallel_execution_on_UNIX-based_para...

As far as comparison with baselines results are concerned,
there is a QA suite residing in

$NWCHEM_TOP/QA

A script named

doqmtests.mpi [no. of processes]

can be used to test the installation

Just Got Here
11:31:50 PM PDT - Fri, May 10th 2013
Hi,

Thanks!
I've noticed that when all 8 cores are running wall time is 55.5s and cpu time is 20.7s .
I've also noticed (looking at "System Monitor") delay in calculations for ~15sec in case of 8 processors.
Processors either do nothing or they work with random power/percentage.

Is there a way to decrease this time difference?

Forum Vet
11:57:59 AM PDT - Mon, May 13th 2013
Could you be more specific on the cases showing a large difference between wall and cpu time?
If the case is too small, there is not enough workload to be parallelized and with 8 processes you are are already running out of steam. Using fewer processes (4 or even 2) is probably a more efficient way to exploit your computational resources.

Just Got Here
3:35:18 PM PDT - Tue, May 14th 2013
Hi,

I re-run input file ("Water Molecule Sample Input File") from the page"Getting Started":
http://nwchemgit.github.io/index.php/Release61:Getting_Started

Command ~$ which nwchem gives: /usr/bin/nwchem , so to run input file I was using command:
mpirun -np 8 /usr/bin/nwchem h2o.nw each time changing np to 1,2,4 and 8

Results:
np=1 cpu: 15.2s wall:19.1s
np=2 cpu: 12.5s wall:24.4s
np=4 cpu: 12.5s wall:25.9s
np=8 cpu: 21.7s wall:80.9s

Since I am new to NWChem, how (and where) can I learn how to optimally run the program?

Do I have to provide any particular additional data/files?

Thank you

P.S: Bellow is copy/paste of "GA Statistics" for the last run (np=8):



                               NWChem Input Module

                               -------------------




Summary of allocated global arrays



 No active global arrays





                        GA Statistics for process    0

                        ------------------------------



      create   destroy   get      put      acc     scatter   gather  read&inc
calls: 1.77e+04 1.77e+04 1.03e+05 3.82e+04 4.54e+04 2586 0 0
number of processes/call 1.54e+00 1.70e+00 1.83e+00 3.59e+00 0.00e+00
bytes total: 5.29e+07 2.23e+07 1.47e+07 8.03e+04 0.00e+00 0.00e+00
bytes remote: 1.95e+07 8.01e+06 9.97e+06 -3.06e+05 0.00e+00 0.00e+00
Max memory consumed for GA by this process: 55952 bytes

MA_summarize_allocated_blocks: starting scan ...
MA_summarize_allocated_blocks: scan completed: 0 heap blocks, 0 stack blocks
MA usage statistics:

allocation statistics:
heap stack
---- -----
current number of blocks 0 0
maximum number of blocks 26 51
current total bytes 0 0
maximum total bytes 1062168 22510080
maximum total K-bytes 1063 22511
maximum total M-bytes 2 23


Total times  cpu:       21.7s     wall:       80.9s


Forum >> NWChem's corner >> Running NWChem