Re: [Spglib-users] problem with spg_get_ir_reciprocal_mesh

[Keith R. <kr...@gm...>]

Hello Fabiano and Togo,

Interesting.  One of the things on my to do list was to compare the 
performance of SPGlib k-point generation and symmetry reduction to 
CASTEP's.  Getting this to scale reasonably in CASTEP was a significant 
algorithmic challenge; the current solution requires sorting the 
k-points into shells and testing for equivalence within a shell.  But if 
you have a faster algorithm I'd love to hear it.

Keith Refson

On 12/12/16 23:26, Atsushi Togo wrote:
> Hi,
>
> I confirmed your results in spglib. Spglib outputs is wrong.
> I implemented supposing mesh doesn't break symmetry. In this case,
> algorithm with integers can handle dense mesh, e.g. 100x100x100 mesh is
> easily handled.
>
> Anyway spglib results are clearly wrong. To avoid this and to implement
> an algorithm to obtain the similar results by CASTEP that is as fast as
> current spglib routine together with keeping the user interface, I need
> some time to implement it. I show work-around in the last part of the
> test-code below, which uses least common multiple.
>
> Togo
>