Hi,
On Thu, Sep 17, 2009 at 04:22:49PM -0400, case wrote:
> On Thu, Sep 17, 2009, Scott Brozell wrote:
> >
> > No, they have the same functionality, but different API's.
> > The difference is in the names:
>
> This is generally handled by a blaswrap.h header file with a bunch of cpp
> #defines. (See, e.g. amber11/src/cblas/blaswrap.h.) But Amber has used
> the convention where none of this name mangling is used, so cblas and blas
> are actually the same API (for us).
Ok, without looking I assumed that you were using
the CBLAS API as defined here:
http://www.netlib.org/blas/blast-forum/cinterface.pdf
Have you ever used that CBLAS API, e.g., blast-forum/cblas.tgz ?
> > I see now that sff uses a cblas where
> > each function's name is the BLAS name with _ postpended.
> > This looks like a hack. Why add a trailing underscore if u have
> > converted via f2c. Isn't this asking for trouble with fortran
> > mangled names which can add a trailing underscore among other mangles.
>
> The "industry standard" fortran name mangling for non-modules is to convert to
> lowercase and add an underscore. That's why the netlib cblas
> "works" as a drop-in replacement for the fortran version.
I'm skeptical that there are no longer platforms that use
just the name in lowercase or the name in uppercase.
IBM and Mac come to mind.
I think we still need CLINK_CAPS and CLINK_PLAIN.
> > And Antechamber and mopac are distributed with Chimera.
>
> OK...so this means the dock itself doesn't need to have mopac, and hence,
> doesn't need a fortran compiler.
>
> > After more use with dock, we clearly got the default wrong in amber score
> > prep.
>
> ?? I can't parse the sentence above.
prepare amber should have used the existing charges in the ligand mol2
file as its default charge treatment instead of the actual default
of always calculating am1bcc.
This has been a source of confusion (since people dont read or digest
the documentation) and a source of wasted computation (since all real
DOCK calculations start with ligand prep that calculates charges) and
worse for some movable regions calculating the charges is more
expensive than the score.
Scott
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Thu Sep 17 2009 - 15:00:02 PDT
