[Q-e-developers] Internal LAPACK and BLAS

[Ye Luo]

Some routines from ESSL have the same names as LAPACK but take different
arguments. Also ESSL has an incomplete set of LAPACK routines.
The right order of linking involves ESSL is ESSL(for part of the LAPACK
routines), LAPACK, ESSL(for BLAS).

Ye

===================
Ye Luo, Ph.D.
Leadership Computing Facility
Argonne National Laboratory

2016-09-11 8:46 GMT-05:00 Filippo Spiga <spiga.filippo at gmail.com>:

> On Sep 11, 2016, at 2:41 PM, Paolo Giannozzi <p.giannozzi at gmail.com>
> wrote:
> > I think that the original reason for separate blas and lapack treatment
> is that once upon a time there were mathematical libraries with an
> incomplete (or even nonexistent) set of lapack routines
>
> ESSL I guess (only case covered in the configure).
>
> ESSL evolved quite a lot since the "old days". I have no access to a
> machine where ESSL (ESSL linux or ESSL BlueGene?) is deployed but if
> someone remember which routine was not present we can easily check. It is
> always possible to force a behaviour of the linker where libraries are
> linked in this order {ESSL, Netlib LAPACK, Netlib BLAS}, it is a flag to
> turn in the make.inc
>
> --
> Filippo SPIGA ~ http://fspiga.github.io ~ skype: filippo.spiga
>
>
> _______________________________________________
> Q-e-developers mailing list
> Q-e-developers at qe-forge.org
> http://qe-forge.org/mailman/listinfo/q-e-developers
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.quantum-espresso.org/pipermail/developers/attachments/20160911/c1ff0549/attachment.html>