do you mean that I should install a new version of fftw for lammps?
please don't get ahead of yourself and jump to conclusions without
having proof. you _are_ a scientist, are you not?
first you have to confirm my hypothesis and try out the peptide example.
if i can run your input and have a lammps binary that works, and
you cannot run an input that is known to work, only then we have proof.
Now I use the version of fftw-2.1.5.
please pay attention to what i was saying. it is _not_ that you
need a different version of fftw, you would need it to be _compiled
differently_. fftw-2 can be compiled for single and double precision
and in both cases you can get a lammps binary. on one case
(double precision) it will work throughout and the other case
(single precision) it will work except for pppm.
you also have to make sure that there is no other version of
fftw-2 installed of your system that might collide with the
double precision version. you have to make sure that your
makefile references the correct directories for including the
header files. there are many ways how people can mess
up a compilation, if they are not careful. sometimes the
tiniest details can make a big difference.
cheers,
axel.