[Pw_forum] C_POINTER on ibm power4

Axel Kohlmeyer akohlmey at cmm.chem.upenn.edu
Sat Dec 16 14:44:22 CET 2006

On 12/16/06, Sergey Lisenkov <proffess at yandex.ru> wrote:
>  Dear users,
>  In new version of pwscf (3.2) the parameter C_POINTER in include/fft_defs.h
> is defined by configure and can be integer*8 (64-bit case) and integer*4
> (32-bit case), if I am not totally mistaken. On Power4+ architecture (with
> environment variable OBJECT_MODE=64, which enables 64-bit version) configure
> defines compilation option "-q64" in both FFLAGS and LDFLAGS. However,
> C_POINTER is defined as integer*4. I guessed that it was error and replaced
> it by integer*8. When my calculations started I saw in the output file:
> G cutoff =  316.5977  (******* G-vectors)     FFT grid: (198,198,***)
> G cutoff =    9.0456  (  93393 G-vectors)  smooth grid: ( 36, 36,168)
> The system is not too big to have z component of FFT of "***". It should be
> around 392 (info from the another machine). So, does it mean that on IBM
> Power4 I have to redefine C_POINTER as integer*4 (as it was defined by

dear sergey,

if you have trouble with the size of pointers, you would get segmentation
faults and NaNs but not '****' outputs. those seems actually possible,
198/36 is 5.5. 168*5.5 is 924. if 36 is much larger than the actually
required xy-grid then you'd have a factor of ~6 which would give you a
z-grid of > 1000.

if you grep through the sources, you'll see that C_POINTER is used
only to interface to the intel mkl and fftw. if you use neither, having
C_POINTER defined to the wrong value should not affect you in case
you are using essl.


> configure)?
>  Thanks,
>    Sergey
> _______________________________________________
> Pw_forum mailing list
> Pw_forum at pwscf.org
> http://www.democritos.it/mailman/listinfo/pw_forum

Axel Kohlmeyer   akohlmey at cmm.chem.upenn.edu   http://www.cmm.upenn.edu
  Center for Molecular Modeling   --   University of Pennsylvania
Department of Chemistry, 231 S.34th Street, Philadelphia, PA 19104-6323
tel: 1-215-898-1582,  fax: 1-215-573-6233,  office-tel: 1-215-898-5425
If you make something idiot-proof, the universe creates a better idiot.

More information about the users mailing list