That's an interesting puzzle; I can reproduce your results on my machine. Clearly, something is being discretized differently in the two cases, but off the top of my head I can't see what would cause this.
Presumably, both results converge to one another as you increase the resolution, so at the end of the day it won't matter, but it is a bit annoying... On Apr 14, 2014, at 12:45 AM, Andrey Panov <pa...@canopus.iacp.dvo.ru> wrote: > After simulating a trivial example with MEEP, I have found that the results > of > the computation of the electric (magnetic, electric+magnetic) field energy in > a > cube obtained using libctl and C++ interfaces slightly (~0.1-1%) differ. This > examle is a 16x16x16 cube with PML boundaries and isotropic constant epsilon; > a > continuous-wave point dipole as a source. The energy is calculated in the > 4x4x4 > cube at the center. Epsilon averaging is not used. The libctl and C++ files > are > attached here. I think that all the comutations are done by the same libmeep > and should be identical or have discrepancies on the order of 1e-16. > > My system is Slackware Linux 14.1 x86_64 with gcc-4.8.2, I tried this on AMD > and > Intel core 5 CPU. I have installed the latest relesad versions of Meep, > Guile, > Libctl. Everything is compiled with "-O2 -fPIC" options, Meep has not MPI > support. _______________________________________________ meep-discuss mailing list meep-discuss@ab-initio.mit.edu http://ab-initio.mit.edu/cgi-bin/mailman/listinfo/meep-discuss