I have probably found out what is causing the error! Harminv uses integers instead of floats at some place. When all the constants (resolution, dimensions and time) are scaled properly so that the simulation output looks the same, Harminv does behave differently.
When the simulation resolution is scaled down, one comes to a zero "data capture time": runWithHarminv starting : sources will stop at time = 0 and will capture data for Harminv analysis during an additional time of 0. This causes Harminv to fail at one or another place. Unfortunately I can not fix the error in the harminv C++ code. A quick workaround is to get along without SI units and rescale all constants to use high enough numbers. Filip On Sun, 4 Mar 2012 10:37:06 +0100 Filip Dominec <[email protected]> wrote: > Hi all, I downloaded the 2D Harminv example [1] and modified it for 3D > simulation of THz resonance in a tiny dielectric sphere. The simulation works > well (there are two modes at 800 and 1500 GHz), but Harminv fails with: > > PHASE 2 : running the Harminv analysis on the probing point data. > harminv: failure on line 503 of harminv.c: invalid argument to ZGEEV > > My code can be downloaded from [2] and its full output is here: [3] > Does anybody have similar experience? May it be a bug in harminv? Can you > point me to where I may be doing anything wrong, please? > > Thanks in advance, > Filip > __ > [1]: http://www.mail-archive.com/[email protected]/msg00012.html > [2]: http://fzu.cz/~dominecf/sphere_harminv.py > [3]: http://fzu.cz/~dominecf/output_error.txt _______________________________________________ meep-discuss mailing list [email protected] http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss
