[Pw_forum] Grid differences between machines - why?

Guido Roma roma at cea.fr
Fri Mar 12 08:51:25 CET 2004


Hello,

it happens that, due to different algorithms, some FFT routines
like some prime factors and not others in the grid; either they are not
allowed or they are very badly performing, so they are to be avoided.
This is what does the function "allowed" in Modules/fft_scalar.f90: it
choses a good fft grid for the machine (i.e. FFT routine) you're using,
according to your energy cutoff.
I suppose on IBM you use ESSL fft routines and on Intel FFTW.
Of course the differences in energy should be negligible as you get
closer to convergence (in both the smooth and fine grids).

Guido

Konstantin Kudin wrote:
> 
>  Here is the situation. I am running PWSCF 2.0.0 on an
> Intel box and on an IBM SP3 (?). The input files are
> identical. Intel's code is serial, IBM's is mpi,
> 32-bit executable.
> 
> On the IBM SP box I get:
>      G cutoff = 1309.2944  (  99231 G-vectors)     FFT
> grid: ( 80, 80, 80)
>      G cutoff = 1047.4356  (  71005 G-vectors)  smooth
> grid: ( 66, 66, 66)
> 
> On the Intel:
>      G cutoff = 1309.2944  (  99231 G-vectors)     FFT
> grid: ( 75, 75, 75)
>      G cutoff = 1047.4356  (  71005 G-vectors)  smooth
> grid: ( 72, 72, 72)
> 
> 
>  Why do I get different grids? They also lead to some
> differences in the energy ...
> 
>  Kostya
> 
> __________________________________
> Do you Yahoo!?
> Yahoo! Search - Find what you?re looking for faster
> http://search.yahoo.com
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

-- 
Guido Roma <roma${at}cea.fr> -- CEA-Saclay - DEN/DMN/SRMP Bat.520/130 
Phone: [+33]-1-69085738 -- Fax: ...6867 -- Mobile: [+33]-6-20069085



More information about the users mailing list