I do not see this particular error; probably because I use
older ifort (9.1) + gcc; apparently I do have the i_dnnt
defined in both f2c.a and the fortran libraries too
(nm /opt/intel/fce/9.1.036/lib/libifcoremt.a | grep i_dnnt),
[which is needed by lapack.a nm ../../lib/lapack.a | grep 'i_dnnt'],
yet it does not produce an error. I can think of multitude of
reasons, but do not know how to pin down the true one.
One possibility to fix it would be to rename i_dnnt in the f2c
(and clapack, etc) into, e.g., i_dnnt_C [it is only used in
clapack/dlaqps.c].
Best wishes,
Volodymyr
On Tue, September 15, 2009 11:05, steinbrt.rci.rutgers.edu wrote:
> Hi,
>
> thanks for the info, I havent seen this particular error before. Any
ideas
> on why it occurs in parallel but not serial compile? What would be a
good
> way to prevent this from happening to users that want to compile without
making changes to the various Makefiles?
>
> Kind Regards,
>
> Thomas
>
> On Tue, September 15, 2009 10:52 am, Volodymyr Babin wrote:
>> Hi Thomas,
>> the parallel sander failure does not look unexpected to me: f2c.a
contains intrinsic fortran routines; same routines may reside in some
sort of library that it is a part of a fortran compiler (to link with
in case they are not inlined by the fortran compiler); it is thus very
normal to expect all sorts of conflicts if one passes f2c.a to the
fortran linker (depending on the number of underscores used in f2c.a vs
the fortran library, etc).
>> Have a nice day,
>> Volodymyr
>> On Tue, September 15, 2009 10:24, steinbrt.rci.rutgers.edu wrote:
>>> Regarding the sander failure,
>>> taking i_dnnt out of the f2c Makefile allows a successful compile of
sander.MPI, which passes the test.parallel.
>>> Is this some problem with my peculiar system? Has anybody else
experienced
>>> this trouble with newish Intel compilers?
>>> Kind Regards,
>>> Thomas
>>> Dr. Thomas Steinbrecher
>>> BioMaps Institute
>>> Rutgers University
>>> 610 Taylor Rd.
>>> Piscataway, NJ 08854
>>> _______________________________________________
>>> AMBER-Developers mailing list
>>> AMBER-Developers.ambermd.org
>>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>> _______________________________________________
>> AMBER-Developers mailing list
>> AMBER-Developers.ambermd.org
>> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
>
> Dr. Thomas Steinbrecher
> BioMaps Institute
> Rutgers University
> 610 Taylor Rd.
> Piscataway, NJ 08854
>
> _______________________________________________
> AMBER-Developers mailing list
> AMBER-Developers.ambermd.org
> http://lists.ambermd.org/mailman/listinfo/amber-developers
>
_______________________________________________
AMBER-Developers mailing list
AMBER-Developers.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber-developers
Received on Tue Sep 15 2009 - 09:30:01 PDT
