Here is more information about the problem:
The error is generated by fields::add_volume_source with the lines
321 double w = user_volume.boundary_location(High, d)
> 322 - user_volume.boundary_location(Low, d);
> 323 if (where.in_direction(d) > w + gv.inva)
> 324 abort("Source width > cell width in %s direction!\n",
> direction_name(d));
I was able to output user_volume.boundary_location using the meep SWIG
wrappers. at resolutions for which the simulation worked, these are as
expected: (xmin, ymin) = (-80.5, -205.5), (xmax, ymax) = (80.5, 205.5). At
resolution 14, though these were (-80.5, -205.5) and (-9.79, 85.6). Thus it
looks like the user_volume is being truncated at some point. Is there a way
to fix this?
Also, it might be worth mentioning that removing symmetry generates the
error
Cannot split -1558399420 grid points into 16 parts
>
Which I guess has to do with using ints instead of longs (
http://comments.gmane.org/gmane.comp.science.electromagnetism.meep.general/3862
)
Just wanted to add some detail in case others end up with a similar problem.
Alex Ball
On Fri, Aug 9, 2013 at 12:01 PM, Alex Ball <ab...@caltech.edu> wrote:
> Hi All,
>
> I am simulating a large device using MEEP 1.1.1 with OpenMPI and parallel
> hdf5 support on the Amazon cloud (7 nodes, 1708 GB RAM, 28 cores used
> [using more cores appears to decrease performance]). My computational cell
> is 161x411x15.05 (a=1um). The device consists of 150,000+ cylinders with
> radii ranging from .145 um to .33 um. I have run the simulation
> successfully at resolution 5. At resolution 14, all is well until sub-pixel
> averaging finishes. below is the relevant part of the error message.
>
> ...
>> subpixel-averaging is 78.5038% done, 1.09568 s remaining
>> subpixel-averaging is 85.7806% done, 0.663234 s remaining
>> subpixel-averaging is 93.2841% done, 0.288219 s remaining
>> time for set_epsilon = 311.675 s
>> -----------
>> meep: Source width > cell width in x direction!
>> --------------------------------------------------------------------------
>> MPI_ABORT was invoked on rank 0 in communicator MPI_COMM_WORLD
>> with errorcode 1.
>>
>> NOTE: invoking MPI_ABORT causes Open MPI to kill all MPI processes.
>> You may or may not see output from other processes, depending on
>> exactly when Open MPI kills them.
>> --------------------------------------------------------------------------
>> ...
>>
>
> I have no idea why I get the " meep: Source width > cell width in x
> direction!" error only high resolution. I have checked in interactive
> mode that the source width is in fact less than the cell width (source is
> 150 um in x direction). I am using an amplitude function to define a
> gaussian source with correct phase fronts perpendicular to the propagation
> direction.
>
> Is there anything I can do to fix this? Let me know what other information
> would be helpful. Thanks!
>
> Alex Ball
>
_______________________________________________
meep-discuss mailing list
meep-discuss@ab-initio.mit.edu
http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
