single precision compilation of cp2k

[Axel]

On Dec 3, 5:56 pm, Axel <akoh... at gmail.com> wrote:
> On Dec 3, 5:44 pm, Teodoro Laino <teodor... at gmail.com> wrote:
>
> > Thanks Axel,
>
> > I'm applying the patches right now.. should be soon in the CVS..
> > did you try a specific regtest/input file (that gives segfault)?
>
> i tried H2O-geoopt.inp
>
> there is more to come, i just noticed, that the f77_blas_poison
> does not seem to be poisonous eneough. i found plenty of
> calls to DSCAL. stay tuned. (too tempting, i should be hacking
> LAMMPS instead. ouch).

oooops. it gets even better... it looks to me
that f77_blas and friends are not even used!!!!

so before is put much time into changing all DSCAL,
DAXPY, DGEMM etc. into the corresponding BL_XXX calls,
can somebody please comment on the use (and the future
of this). it may be very convenient to use that kind
of overloading, but when it is not used anywhere, should
i go to the effort of converting everything???

ciao,
   axel.


>
> axel.
>
>
>
> > teo
>
> > On 3 Dec 2007, at 23:38, Axel wrote:
>
> > > thanks a lot. it is much appreciated.
>
> > > as a token of my appreciation, i've uploaded a minimal patch
> > > containing a few more changes needed to complete the compilation
> > > with -D__SGL. it can actually run pretty far until it dies with
> > > a segfault. ;-)
>
> > > cheers,
> > >    axel.
>
> > > On Dec 3, 2:09 pm, Teodoro Laino <teodor... at gmail.com> wrote:
> > >>> haven't changed anything yet. only looked at
> > >>> what caused errors during a compile attempt
> > >>> with "make -i". just wanted to make people aware
> > >>> of this. will send patches when i have time to work
> > >>> on it for real (may be after christmas).
>
> > >> Take it easy man.. there are better ways to use the Xmas holidays ;-)
>
> > >> In the meanwhile there's a python+teo (Xmas) present for you under
> > >> the CVS(tree) ;-) ..
>
> > >> ciao,
> > >> teo
>
> > >>> axel.