Dear all,
I am running a 3D simulation with roughly 910x570x610 grid points.
When allocating 10 nodes with 8 cores each to this simulation, each node
uses ~34.5GB memory.
When allocating 20 nodes with 8 cores each to this simulation, each node
uses ~34GB memory.
The difference seems to be quite small. (It looks more like every node
stores all fields and the entire permittivity grid.)
In previous posts it was mentioned, that there was a bug (in 2006),
which was taken care of. Is there still a bug when using conductivity?
(According to my administrator we have a meep version downloaded in fall
2009.)
Furthermore, my geometry list contains ~800000 objects. When starting
the simulation, the output-file is more or less instantaneously
generated, where I can find the geometry list again. But for the next 30
hours nothing happens. The nodes are completely busy, but no further
output happens. After these 30 hours, the simulation starts and I can
find the usual lines like:
time for set_epsilon = 389.353 s
time for set_conductivity = 129.938 s
time for set_conductivity = 124.898 s
time for set_conductivity = 121.701 s
-----------
creating output file "./SF750-eztrack.h5"...
Meep progress: 0.00490000078400012/120.0 = 0.0% done in 260.8s,
6385618.7s to go
on time step 1 (time=0.0049), 260.757 s/step
Now is the question, what is meep doing during these 30 hours (note,
that eps-averaging? is already set to false), and why is it 30 hours, no
matter if I'm using 10 or 20 nodes?
Thanks already in advance for your reply,
Stephan Fahr.
And here is a shortened version of the corresponding ctl-file
(reset-meep)
(define-param fcen 1.333333333)
(define-param EpsmSiR 11.68886453)
(define-param EpsmSiI 0.07675385826)
(define-param DefMat 2.25)
(define-param sx 8.941090198) ; size of cell in X direction
(define-param sy 5.5762) ; size of cell in Y direction
(define-param sz 6.0564) ; size of cell in Z direction
(set! geometry-lattice (make lattice (size sx sy sz)))
(set! default-material (make medium (epsilon DefMat)))
(set! k-point (vector3 0 0 0))
(set! ensure-periodicity true)
(set! output-single-precision? true)
(set! force-complex-fields? true)
(set! eps-averaging? false)
(set! geometry (list
(make block (center 3.3 0 0) (size 2.3411 sy sz)
(material (make medium (epsilon EpsmSiR)
(D-conductivity EpsmSiI))))
~800 000 more objects
))
(set! pml-layers (list (make pml (thickness 0.8) (direction X))))
(set! resolution 102.0408)
(set! sources (list
(make source
(src (make continuous-src (frequency fcen) (width 1)))
(component Ez)
(center -3.621545099 0 0) (size 0 sy sz) )))
(run-until 120
(in-point ( vector3 3.6215 0.01 0.01 )
(to-appended "eztrack" (at-every 0.90 output-efield-z)))
)
_______________________________________________
meep-discuss mailing list
[email protected]
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
