The original idea of the -jagged command line option was to discretize the
particle exactly as without it, but replacing each dipole by a cube of JxJxJ
dipoles, where J is the argument of -jagged. This expected behavior is
exemplified by issue 21.
However, currently this is not exactly so, since function FitBox in fft.c
treats differently even and odd values of J. So '-grid 3' leads to 4 dipoles
along the x-axis, and '-grid 3 -jagged 2' leads to 6 (not 8) dipoles along the
x-axis. While this can be justified by optimizations concerns, it distorts the
original goal of -jagged.
It became even more weird after r1008. For example,
adda -grid 4 -jagged 2
leads to non-symmetric box 8x10x10. This behavior is not outright wrong (there
is some logic behind it), but is definitely not the expected one.
Original issue reported on code.google.com by yurkin on 15 Feb 2011 at 5:43
Original issue reported on code.google.com by
yurkin
on 15 Feb 2011 at 5:43