[Wien] ifort (ifc) 8.0 and mkl 6.1

[Veerle Vanhoof]

Dear Wien2k user,

I have some problems with compiling wien2k with ifort8.0.
We use a 3.2 GHz P4, 512kB L2 with 400MHz. Compiling with ifc7.1 and mkl6.1
works fine, but the time we reach for the benchmark is 270s, while 228s
should be possible with ifort8.0 instead of ifc7.1.

Compiling with ifort8.0, mkl6.1 and following settings:

Current settings:
     O   Compiler options:        -FR -mp -w
     L   Linker
Flags:            -L../SRC_lib -L/opt/intel/mkl61/lib/32 -Vaxlib -static
     P   Preprocessor flags       '-DParallel'
     R   R_LIB (LAPACK+BLAS):     -lmkl_lapack -lmkl_ia32 -lguide -lpthread

resulted in a segmentation fault for the benchmark case.
As suggested in previous mails in the mailinglist and by Intel I tried
compiling with "limit stacksize unlimited" and

     L   Linker
Flags:            -L../SRC_lib -L/opt/intel/mkl61/lib/32 -Vaxlib -static-lib
cxa

linking all libraries except libcxa dynamically.
And with

    L   Linker
Flags:            -L../SRC_lib -L/opt/intel/mkl61/lib/32 -Vaxlib

Linking all libraries dynamically. But the segmentation fault remained.

Compiling with

    O   Compiler options:        -FR -mp -w -O0

to disable all optimizations didn't work either.

Since we use Suse Linux, we can't use the nptl-devel package which is an
alternative for Red Hat Linux 9 and
Red Hat Enterprise Linux 3.0.

Did someone already succeed in compiling wien2k with ifort8.0 and mkl6.1 for
Suse Linux?
All suggestions are welcome!

Thanks a lot!

Veerle Vanhoof




---------------------------------------------------------------------------
Veerle Vanhoof
  Katholieke Universiteit Leuven
  Instituut voor Kern- en Stralingsfysica
  Celestijnenlaan 200 D
  B-3001 Leuven
  BELGIUM
  Phone: +32 (0)16 32 71 45
  Fax: +32 (0)16 32 79 85
  email: Veerle.Vanhoof at fys.kuleuven.ac.be
---------------------------------------------------------------------------




----- Original Message ----- 
From: "Lutz Rauscher" <rauscher at qms.cache.waseda.ac.jp>
To: <wien at zeus.theochem.tuwien.ac.at>
Sent: Wednesday, February 04, 2004 11:19 AM
Subject: [Wien] ifort (ifc) 8.0 and mkl 6.1


> Dear All,
>
> I hope has not yet been discussed; I lost track a bit... I compiled Wien2k
> with ifort (ifc) 8.0 and mkl 6.1 and there are two "known issues" that
need
> to be considered if you do so:
>
> 1. ifort troubles with a static linked pthreat library. The easiest thing
is
> just to omit the -static linker option. If you really need the static
> linking there's another workaround described in the release note involving
> the nptl-devel package, but I didn't try.
>
> 2. For larger cases you get segmentation faults anyway, because ifort
"[...]
> allocates more temporaries on the stack than previous Intel Fortran
> compilers. If a program has inadequate stack space at runtime, it will
> terminate with a Segmentation fault or Signal 11. The stack space can be
> increased with the ulimit -s unlimited command (in bash shell) or limit
> stacksize unlimited (in csh shell)".
>
> That does the trick and Wien2k seems to run smooth - at least for my
cases.
> It would be interesting whether anybody has still problems.
> Its maybe worth to try, it seems to be quite fast: I measured 228s for the
> benchmark test on a 3.2 GHz P4, 512kB L2 with 400MHz dual bus memory.
>
> Regards, Lutz
>
>
> _______________________________________________
> Wien mailing list
> Wien at zeus.theochem.tuwien.ac.at
> http://zeus.theochem.tuwien.ac.at/mailman/listinfo/wien