[CP2K:2231] GLOBAL / BLACS_GRID and globenv%blacs_grid_layout

Teodoro Laino teodor... at gmail.com
Mon Aug 17 21:06:25 UTC 2009

Hi Noam,

blacs_grid_layout is an optional argument, which is never set in the 
call to this subroutine (therefore using always the
default value).

The idea you had  (trying to match the GLOBAL/BLACS_GRID) is perfect. 
You can set a variable in the globenv and then
use it in the place where you really need it or alternatively directly 
taking this value from the input file without need of stoing it
in an additional type. You don't need to have a globenv in the cp_blacs 
type, instead you can pass an
optional argument (which has a default) in the places where you need it.

Please let me know i f you need help or when you have this patch 
available, I will be extremely happy to work on it for having this patch 
in the CVS.

Noam Bernstein wrote:
> Hi - I'd like to patch CP2K so that it can (optionally) call the appropriate
> BLACS_SET after creating a BLACS grid to ensure repeatability
> of collectives.
> I was hoping to follow along with the usage of GLOBAL / BLACS_GRID.
> I see that cp_blacs_env_create() takes an optional variable
> blacs_grid_layout, and I figured I could just add another one
> (blacs_repeatable, or something of the sort).  However, I don't
> see any place where this optional variable is being passed in
> from globenv.
> Am I missing something?  Are the changes necessary for BLACS_GRID
> to really be used just incomplete?  Will the information be passed down
> from the top level as an integer (blacs_grid_layout) or by passing
> globenv?
> I'd be happy to produce a patch that implements blacs_grid_layout
> and adds blacs_repeatable, but I don't want to do it in a way that
> conflicts with developer desires, style, or planned changes.
>      Noam
> >

More information about the CP2K-user mailing list