On Tue, Mar 16, 2010, Tyler Luchko wrote:
>
> >Better: try intel without mkl. If you didn't already do so, try adding
> >"-oldmkl" to configure. But I suspect we may still have problems with
> >using the MKL wrappers to fftw.
>
> Ok, I think I have a solution to this in the attached patch. I'm now
> following the code in the configure script that sets $mkll. From what I
> have been able to google, I think this will also work for older versions
> of MKL as well (I only have access to version 11.x).
Applied. But we may want to consider only using the MKL wrappers if the
user sets a flag to do so. That way, the majority of users (who won't be
using rism anyway) won't be susceptible to this sort of error....let's wait
to see if more problem reports come in.
On another point, I see there is now an "ifdef" in amber11/src/nab/Makefile.
I'm pretty sure this is gnu-make dependent, and not portable to other versions
of make. The usual workaround is to define a variable in config.h, then
have the variable be a target. For example, there would be a target called
"libnab_rism" and one called "libnab_norism": each would make libnab.a, just
with different objects in it. There are various ways to skin this cat, and
we can talk about it.
Also: should we(you) be handling the FLIBS_FFTW2 target the same way in
the nab Makefile as in the sander Makefile? I don't quite understand how the
sander one works...
....thx...dave
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Mar 16 2010 - 17:30:04 PDT
