[Meep-discuss] bug in output-png+h5 with meep-mpi

[matt]

Hello,

I noticed some strange behavior in output-png+h5 with meep-mpi 0.10.1





mpirun -np 4 meep-mpi silverrod1.ctl

The code below will (hopefully) reproduce the error...



Best,
Matt




(set! eps-averaging? false)
(define-param s 1600) ; the size of the computational cell, not including PML
(define-param dpml 400) ; thickness of PML layers

(define sxy (+ s (* 2 dpml))) ; cell size, including PML
(set! geometry-lattice (make lattice (size sxy sxy no-size)))

(set! geometry '())

(set! pml-layers (list (make pml (thickness dpml))))
(set-param! resolution (/ 1 10)) ;; pixels per distance unit

(define ((pw-amp k x0) x)
  (exp (* 0+1i (vector3-dot k (vector3+ x x0)))))

(define-param lcen 213)
(define-param fcen (/ 1 lcen)) ; pulse center frequency
(define-param df (/ fcen 100)) ; turn-on bandwidth

(define k (vector3-scale (* 2 pi fcen)
                         (unit-vector3 kdir))) ; k with correct length
(define kxcos (vector3-x (unit-vector3 k))) ; direction cosine of k in x
(define kycos (vector3-y (unit-vector3 k))) ; direction cosine of k in x

(set! sources
      (list

       ; left
       (make source
         (src (make continuous-src (frequency fcen) (fwidth df)))
         (component Hz) (center (* -0.5 s) 0) (size 0 s)
         (amp-func (pw-amp k (vector3 (* -0.5 s) 0))))

       ; bottom
       (make source
         (src (make continuous-src (frequency fcen) (fwidth df)))
         (component Hz) (center 0 (* -0.5 s)) (size s 0)
         (amp-func (pw-amp k (vector3 0 (* -0.5 s)))))

       ))

(define-param T 100000) ; run time
(run-until T (at-beginning output-epsilon)
           (at-end (output-png+h5 Hz "-Zc dkbluered"))
           (at-end output-tot-pwr))

The png output is incorrect has white patches blocking the fields. If I run the h5topng -Zc dkbluered on the h5 ouput, the resuling png appears correct without the white patches. Non-parallel meep does not exhibit this behavior. I attached a sample png which illustrates what happens. Depending on the number of processors I run, the white patches can change their layout. For the attached image, I ran with four processors As a side question, I never get perfect plane waves with this approach, even if I make the pml thick and let the simulation run for many time steps. Is there something else I should consider? (define-param kdir (vector3 1 1)) ; direction of k (length is irrelevant)

new.png
Description: Binary data

_______________________________________________
meep-discuss mailing list
meep-discuss@ab-initio.mit.edu
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss