The ABINIT Users Mailing List ( CLOSED )

["Anglade Pierre-Matthieu" <anglade@gmail.com>]

Hi all,

Following XG suggestions I've tryed to compile a sequential version
with the followings flags (used for 4.x.x versions of abinit on
IBM/pwr4):
" -O4  -qarch=auto -qtune=auto  -qstrict -qsuppress=1520-003"
I've not tested if this gives more speed. Yet most of the self test
are OK on our machine.

regards

PMA



On 9/2/06, Deyu Lu <dylu@ucdavis.edu> wrote:
> Pierre-Mathieu and Yann:
>        Thank you for helping me with compiling. For your reference, I
> found that abinis compiled with optimization level higher than -O2 on
> 64-bit aix/pw4 machine is not stable. Even using "-O3 -qstrict" won't
> help. To run the code safely, I recommend the following:
>
> ./configure --enable-fortran-wrapper FC=mpxlf90_r CC=/usr/bin/cc --with-
> mpi-prefix=/usr/lpp/ppe.poe --enable-mpi --with-mpi-
> run=/usr/lpp/ppe.poe/bin/mpiexec --with-mpi-fc=mpxlf90_r TRUE_CPPFLAGS="
> " AR="/usr/bin/ar -X 64"  --enable-64bit-flags --with-fc-optflags="-g -
> O2 -qnoipa -q64 -qarch=pwr4 -qtune=pwr4"
>
>         One needs also modify wrap-fc as the way Pierre-Mathieu
> suggested. Maybe you could make proper changes in the wrapper and
> configure file in the future release so that people won't have trouble
> compiling abinit on this kind of machine.
>
>        Best
>        Deyu


--
Pierre-Matthieu Anglade