[Wien] kgen and lapw1 on intel machines

Peter Blaha pblaha at zeus.theochem.tuwien.ac.at
Thu Jul 22 08:08:48 CEST 2004


>    Current settings:
>      O   Compiler options:        -O -FR -mp -w
>      L   Linker Flags:            -L../SRC_lib -Vaxlib
>      P   Preprocessor flags       '-DParallel'
>      R   R_LIB (LAPACK+BLAS):     -llapack_lapw -lblas_lapw
>
>        --> I found
>            this does not work. Some of errors are

You may have to specify these libraries twice:
       -llapack_lapw -lblas_lapw -llapack_lapw -lblas_lapw
in addition you may have to add xerbla.o, which can be created in
SRC_lib from xerbla.f  (ifc -c xerbla.f -mp -w)


>      R   R_LIB (LAPACK+BLAS):     -llapack_lapw -lf77blas -latlas -lguide
> -lpthread
>
> Then, both compilation and actual run are fine.  Here I use lapw's
> lapack and atlas' blas and others. Here are my questions.
>
>   (Q.a) Are these setting completely right without any caveat?

Yes.
>   (Q.b) Is it true that in principle, I could get rid of -L../SRC_lib
> and -llapack_lapw, instead simply add -llapack? (But as I mentioned
> previously, when I do that, the codes do not run properly).

Then you have some incompatibilities in llapack?
>
>   (Q.c) Once I use -llapack_lapw, is that true that I lose some
> optimization (lapack) associated with atlas? I remember that Peter
> said something before like lapack's optimization is not a big deal?

Cpu time does NOT go into lapack, this is just a driver, which calls BLAS.
So optimization is not crucial.


                                      P.Blaha
--------------------------------------------------------------------------
Peter BLAHA, Inst.f. Materials Chemistry, TU Vienna, A-1060 Vienna
Phone: +43-1-58801-15671             FAX: +43-1-58801-15698
Email: blaha at theochem.tuwien.ac.at    WWW: http://info.tuwien.ac.at/theochem/
--------------------------------------------------------------------------




More information about the Wien mailing list