Subpixel averaging progress loops over and over. Not finishing. 3D simulation.

[ahgibbons]

I am trying to make a simulation in 3D, however, the subpixel-averaging step never seems to finish. The percentage increases then loops back to a low percent and repeats. It's not clear to me if this process will finish or not, I left it running for 10 minutes with no apparent progress.

I removed all geometry from the simulation and the same happens.

I reinstalled pymeep with conda create -n mp -c conda-forge pymeep pymeep-extras and I have the same result.

Minimum working example is below:

import meep as mp

CELL_WIDTH = 200 # nm
CELL_HEIGHT = 200 # nm
CELL_DEPTH  = 200 # nm

LEFT = -CELL_WIDTH/2
CENTER = 0
resolution = 10
dt = 100
runtime = 400 #2000
PML_width = 50.0 # nm
LIGHT_SOURCE_POS = LEFT+PML_width + 50

wavelength = 50

cell = mp.Vector3(CELL_WIDTH, CELL_HEIGHT, CELL_DEPTH)

sources = [mp.Source(mp.ContinuousSource(wavelength=wavelength,width=2),
        component=mp.Ez,
        size=mp.Vector3(1, CELL_HEIGHT, CELL_DEPTH),
        center=mp.Vector3(LIGHT_SOURCE_POS,0,0))]

pml_layers = [mp.PML(PML_width,direction=mp.X),
                mp.PML(PML_width, direction=mp.Y),
                mp.PML(PML_width, direction=mp.Z)]

sim = mp.Simulation(cell_size=cell,
        boundary_layers=pml_layers,
        geometry=[],
        k_point = mp.Vector3(),
        sources=sources,
        resolution=resolution)

 
sim.run(mp.at_beginning(mp.output_epsilon),
    mp.to_appended("ez", mp.at_every(dt, mp.output_efield_z)),
    until=runtime)

[oskooi]

Since the geometry object is empty, the issue here is not subpixel smoothing but rather that you have a 3D cell with (200*10)^3 = 8e9 voxels which will require at least 768 Gb of memory (see this FAQ). Trying to load this much data into memory is why the field initialization stage prior to time stepping is taking so long.

[ahgibbons]

Oh..... that was a silly mistake. I got the resolution backwards.
Thank you for the help!