- operations with fractional translations were discarded if not commensurate
with FFT grid (unless input variable "use_all_frac" was set to true)
Now:
- FFT grids are made commensurate with fractional translations
(unless use_all_frac is set to true)
Advantages:
- the number of k-points and of symmetry operations no longer depends upon the
cutoff via the FFT grid (could happen, and did happen, before)
- cleaner code, allowing easier automatic parallelization over k-points
Disadvantage:
- the FFT grid may become slightly larger
Changes are minimal because the machinery was already in place (see variable
"fft_fact" that contains factors for FFT dimensions). Cleanup: useless calls
to fft_type_allocate removed (it is called by fft_init_type anyway).
Documentation and examples updated. The usage of "use_all_frac" is a little
bit weird, though (should be the other way round).
truncated to a small number of significant digits. This is completely
irrelevant in terms of resuts but stll sufficient to yield very small but
visible discrepancies with respect to other XC implementations. I have
converted to full precision all such constans I have spotted. There might
be more cases like these.
PW tests updated: a number of small changes, fixes and corrections
affecting the numerical results had accumulated.
git-svn-id: http://qeforge.qe-forge.org/svn/q-e/trunk/espresso@13592 c92efa57-630b-4861-b058-cf58834340f0