Re: libacml.a
On Friday 30 June 2006 12:01, Jo Shields wrote:
> Giacomo Mulas wrote:
> > On Fri, 30 Jun 2006, Francesco Pietra wrote:
> >> Configure (amd64 debian etch dual opteron, installed g77 and gcc)
> >> requests
> >> (among other libs that I have found)
> >>
> >> libacml.a (the amd core math lib)
> >>
> >> #apt-file search libacml*
> >>
> >> is not informative.
> >>
> >> Is that necessasy (64bit) lib available from debian or should I
> >> install a
> >> different Fortran compiler?
> >
> > ACML is an optimised, proprietary math library by AMD, and is
> > available free
> > of charge from their web site. By the way, since you are apparently doing
> > heavy linear algebra calculations, you may want to check also the GOTO
> > library (I don't remember the URL, google for GOTO blas lapack library
> > and
> > you will find it). It's a hand-tuned linear algebra library which can
> > give
> > an appreciable boost, especially on opterons.
> >
> > Have fun, and please keep me informed on your progress with mpqc: I do
> > heavy
> > quantum chemical calculations myself, although I currently mostly (ab)use
> > NWChem and Gaussian, and I'm interested in comparing that with mpqc (I
> > never
> > tried it yet).
> >
> > Bye
> > Giacomo
>
> Bear in mind, however, that the Gaussian license does not allow
> comparisons with other software, which can lead to your license being
> revoked for your entire site. See bannedbygaussian.org
Back for a moment to this observation. I have not read the license because I
am not going to turn to this software. However, if the situation is that
described above, are the scientific and ethical rules set forth by any
scientific society fully complied? I mean, can a paper be accepted for
publication when substantial evidence is based on that software?
francesco pietra
Reply to: