Dear All users,
When I performed my run on a supercomputer(16 node, ubuntu 12.04) I got these warnings while
I can not understand them and where they come from. Are they serious problems?
Dear All users,
When I performed my run on a supercomputer(16 node, ubuntu 12.04) I got these warnings while
I can not understand them and where they come from. Are they serious problems?
Dear All users,
When I performed my run on a supercomputer(16 node, ubuntu 12.04) I got
these warnings while
a machine with 16 nodes would rarely count as a supercomputer these days.
http://en.wikipedia.org/wiki/Supercomputer
I can not understand them and where they come from. Are they serious
problems?---------------
LAMMPS (12 Apr 2013)
WARNING: OMP_NUM_THREADS environment is not set. (../comm.cpp:100)
using 1 OpenMP thread(s) per MPI task
Scanning data file ...
4 = max bonds/atom
6 = max angles/atom
9 = max dihedrals/atom
Reading data file ...
orthogonal box = (-36.1936 -23.2266 -26.1039) to (26.2344 23.0024 25.4961)
4 by 2 by 2 MPI processor grid
4340 atoms
2944 bonds
3744 angles
1800 dihedrals
Finding 1-2 1-3 1-4 neighbors ...
4 = max # of 1-2 neighbors
6 = max # of 1-3 neighbors
9 = max # of 1-4 neighbors
10 = max # of special neighbors
2592 atoms in group CAR
1100 atoms in group PROP
648 atoms in group CAC
3240 atoms in group A
1100 atoms in group mobile
1 rigid bodies with 3240 atoms
Ewald initialization ...
G vector (1/distance) = 0.312147
estimated absolute RMS force accuracy = 1.71834e-05
estimated relative force accuracy = 1.19332e-06
KSpace vectors: actual max1d max3d = 13670 22 45562
Setting up run ...
Memory usage per processor = 14.0292 Mbytes
Step Temp PotEng TotEng Press Volume
0 0 -34427.839 -34427.839 -36454.517 148916.77
[hpcrc:06799] 15 more processes have sent help message help-mpi-btl-base.txt
/ btl:no-nics
[hpcrc:06799] Set MCA parameter "orte_base_help_aggregate" to 0 to see all
help / error messages
there has to be more output here that would give more insight.
in any case, have you talked to the user support of the center? that
would always be the first point of reference. these messages look like
they originate from OpenMPI, ie. are not a LAMMPS issue.
axel
Dear Axel,
Thank you very much for your reply.
Yes you are right. These 16 nodes having 16 processors while the total nodes are 256 nodes and each node has 16 processors.
I called it supercomputer that’s enough for its users. You probably are greedy for processors:).
Dear Axel,
Thank you very much for your reply.
Yes you are right. These 16 nodes having 16 processors while the total nodes
are 256 nodes and each node has 16 processors.
home come a 16 node "supercomputer" suddenly has 256 nodes?
also i seriously doubt the 16 "processors" number, that are likely
*two* processors with 8 CPU cores each.
I called it supercomputer that's enough for its users.
huh?
my laptop is a "super" computer (because it does all i want it to do
and then some), but that doesn't make it a "supercomputer". using the
term supercomputer implies an exceptionally large machine and those
machines come with an elaborate support infrastructure, which is not
true for a cluster with a handful of nodes.
You probably are greedy for processors:).
what has one to do with the other? you have no idea what i do in my
day to day work, how can you draw this conclusion? in fact, the
contrary is true. i couldn't care less about really big machines. i
care about using *any* machine adequately.
we are in the business of science and the first and most important
task for a scientist is to be accurate in his observations. if your
observations/descriptions are incomplete, imprecise or inaccurate,
that will lead to incorrect conclusions, as you nicely demonstrated
above.
axel.
I called it supercomputer that’s enough for its users. You probably are greedy for processors:).
The greed for processors must of the time comes out of the complexity of the problem one is trying to simulate. Even within the Lammps framework you could find yourself in need of bigger computing power if you were to model a solvated protein in explicit water via all-atom force fields or a reaxff system with more than a handful of atoms. Move onto the ab initio field with all electron calculations and you’ll see why a super computer or cloud may quickly become the only choice. Your statement is just limited to your experience and that of your local colleagues who do not seem to have crossed to any of the two fields mentioned above or other with similar characteristics.
Carlos
Yes, you are right. My comment was superficial.
Did not aim to make you feel bad but just to remind you that there are problems for which the supercomputer way is the only way. “Greed for processors” on its own would have never accomplished the hardware/software development supercomputers require.
Carlos
PS: Most of my work never required more than ~16 CPUs thus you are not alone there
Dear Carlos,
Thank you for your kindness.
Yes you are right I had a bad feeling.