The ABINIT Users Mailing List ( CLOSED )

[Xavier Gonze <gonze@pcpm.ucl.ac.be>]

Dear Masao Arai,

Thanks, your suggestion has been ported in the future v5.4.4 (also  
available in the future 5.5)

Xavier

On 13 Sep 2007, at 05:11, Masao ARAI wrote:

> Dear all,
>
> I have experienced the same problem with PGI compiler (v7.0-4).
> With following modification, abinit will compile fine with PGI  
> compiler.
>
> ***************
> --- defs_fftdata.F90.org        2007-09-12 17:06:16.000000000 +0900
> +++ defs_fftdata.F90    2007-09-12 17:06:54.000000000 +0900
> @@ -96,12 +96,12 @@
>      51840, 8, 6, 6, 6, 6, 5,   55296, 8, 8, 8, 6, 6, 3,     57600,  
> 8, 8, 6,
> 6, 5, 5,     61440, 8, 8, 8, 8, 5, 3,  &
>      62208, 8, 6, 6, 6, 6, 6,   64000, 8, 8, 8, 5, 5, 5,     65536,  
> 8, 8, 8,
> 8, 4, 4   /)
>    integer, parameter :: ifftdata(mdata,ndata)= reshape(  (/ &
> -  &  ifftdata1_40(1:280) , &
> -  &  ifftdata41_80(1:280) , &
> -  &  ifftdata81_120(1:280) , &
> -  &  ifftdata121_160(1:280) , &
> -  &  ifftdata161_200(1:280) , &
> -  &  ifftdata201_231(1:217)  &
> +  &  ifftdata1_40, &
> +  &  ifftdata41_80, &
> +  &  ifftdata81_120, &
> +  &  ifftdata121_160, &
> +  &  ifftdata161_200, &
> +  &  ifftdata201_231  &
>    &     /), (/mdata,ndata/) )
>
>    integer, parameter :: ifftsizes(ndata)= (/   &
> ****************
>
> I don't understand why such modifaction is necessary. But, it works  
> on my
> machine with PGI fortran.
>
> Best regards,
> Masao Arai
>
>
> > Dear Kris,
> >
> > v5.4.3 has been tested with a very old version (4.0 , 32 bits) of the
> > PGI compiler, and it worked.
> >
> > If v5.4.2 was fine for you, I suggest that you copy the
> > defs_fftdata.F90 from v5.4.2 to v5.4.3, and recompile.
> > Nothing else should be changed. Actually, it would help if you could
> > look at this module (it is very simple),
> > and try to find a workaround for your compiler on top of v5.4.3, that
> > might be portable, and return it to me...
> > The problem with the v5.4.2 version on the MacOSX was that the line
> > defining ifftdata was too long,
> > so one had to split it ... as done in v5.4.3 .
> >
> > If nothing based on a module with parameter array is really portable
> > accross platforms,
> > one will have to use to a routine for initializing the array  
> > ifftdata .
> >
> > Best regards,
> > Xavier
> >
> > On 04 Sep 2007, at 01:36, Kris Andersen wrote:
> > > In compiling abinit version 5.4.3 with the pgi compiler, I get the
> > > following error message for defs_fftdata.F90. Is there a simple fix?
> > >
> > >   make[1]: Entering directory `/workspace/kreander/abinit-5.4.3/pgi/
> > > src/defs'
> > >   mpif90 -DHAVE_CONFIG_H -I. -I../../../src/defs -I../.. -I../defs -
> > > I../../../src/defs/../defs     -O2 -Mextend -c -o
> > > defs_fftdata.o ../../../src/defs/defs_fftdata.F90
> > >   PGF90-S-0004-Array reference not a named constant array
> > >   (../../../src/defs/defs_fftdata.F90: 98)
> > >   PGF90-S-0004-Array reference not a named constant array
> > >   (../../../src/defs/defs_fftdata.F90: 98)
> > >   0 inform,   0 warnings,   2 severes, 0 fatal for defs_fftdata
> > >   make[1]: *** [defs_fftdata.o] Error 2
> > >
> > > This seems to be related to a recent discussion by Matthieu
> > > Verstraete and Xavier Gonze on the developer mailing list?
> > > (subject: [abinit-developer] fftdata)
> > >
> > > My version of pgif90 is,
> > >
> > >   $ mpif90 -V
> > >   pgf90 7.0-7 64-bit target on x86-64 Linux
> > >
> > > Thanks for the help.
> > >
> > > --
> > > Kristopher E. Andersen         http://www.physics.nau.edu/~andersen/
> > > Northern Arizona University                    kris.andersen@nau.edu
> > > Department of Physics and Astronomy                   (928) 523-7202
>
>
>
> --
> Masao Arai
> arai.masao@nims.go.jp
> CMSC, NIMS
> Tsukuba Japan