Catastrophical Error w/ Intel Compiler and w/o libintwrapper

Frank Uhlig uhlig... at googlemail.com
Thu Nov 26 11:08:09 CET 2009


Hi Matthias,

sorry, you have been clear on that, but I must have overlooked it. My
fault.
Anyways, I am not able to reproduce the aforementioned failure due to
the compile-time error and so far I will stick to the 10.x compiler
series of Intel because they do work.
Would still be nice to get any clearance about the compiling error...

Frank

On Nov 25, 10:46 pm, Matthias Krack <matthia... at psi.ch> wrote:
> Hi Frank,
>
> may be I wasn't sufficiently clear in my last mail. Your input
> contains the keyword XC_SMOOTH_RHO which currently seems to cause a
> crash when using the Intel 11.x compiler. This is a bug in the code or
> in the compiler. For the time being, just comment out the
> XC_SMOOTH_RHO keyword which might not be crucial or use the Intel
> compiler version 10.1 with libint wrappers which seems to work fine.
>
> Matthias
>
> On Nov 25, 9:26 pm, Frank Uhlig <uhlig... at googlemail.com> wrote:
>
> > Hi Matthias,
>
> > as suggested I reduced (and disabled) the optimization. So far there
> > is no change in the before reported compile-time error. The main
> > difference between us here that could cause different results in
> > compiling is that i work on an em64t architecture. So there seem to be
> > differences in the behaviour of used compilers (although i would like
> > this not to be the case).
> > By the way, my virtual memory is set to unlimited and as I mentioned
> > before, when I use cp2k in calculations not concerning the regtests
> > (see input-file) I also do get the error. And this is the problem i
> > don't understand on a 32 GB machine with eight cores. He initializes
> > the calculation - memory demand goes up - everything fine. Just right
> > before the first SCF-Cycle it crashes due to the supposed lack of
> > memory although there is still quite some available (would have to
> > check the exact numbers again).
>
> > Frank
>
> > input file:
>
> >http://groups.google.com/group/cp2k/web/W32-minimize.inp
>
> > On Nov 25, 12:06 pm, Matthias Krack <matthia... at psi.ch> wrote:
>
> > > Hi Frank,
>
> > > I tried the compilation with Intel Version 11.1 (Build 20090827
> > > Package ID: l_cprof_p_11.1.056) on IA-64 using MKL 10.1 and
> > > libint-1.1.4:
>
> > > - the compilation works if the optimization is disabled (-O0) for the
> > > file semi_empirical_int_utils.F (rest with -O2)
>
> > > - no HFX wrapper routines are needed with this compiler version, ie.
> > > ISO_C_BINDING are indeed supported
>
> > > - all related HFX regtests passed
>
> > > - only 18 regtests failed with the message "forrtl: severe (41):
> > > insufficient virtual memory" which is fortunately quickly triggered by
> > > the "ulimit -v 2000000 safeguard" in the regtest script. Otherwise the
> > > memory blowup would continue much longer.
>
> > > - all failing regtests request density smoothing using the keyword
> > > XC_SMOOTH_RHO. The regtests work if this keyword is commented.
>
> > > Obviously, there is an issue related to the density smoothing which
> > > has to be resolved.
>
> > > Matthias
>
> > > On Nov 24, 5:51 pm, Unknown <uhlig... at googlemail.com> wrote:
>
> > > > Hi Matthias,
>
> > > > thanks for the suggestion. I am already on it (ifort v10.1.021), because
> > > > the cp2k version compiled with the 11.1 seems to have reproducable
> > > > memory issues (i.e., always getting an abort due to "insufficient
> > > > virtual memory").
> > > > Right now the regtests are running and so far they are doing quite well.
> > > > Will try the -O2.
> > > > But the Problem with the ISO_C_BINDING stays, because I can't try that
> > > > with the 10.x intel Fortran Compilers as they do not have this feature
> > > > included. So any help here still would be appreciated...
>
> > > > Thanks so far,
>
> > > > Frank
>
> > > > On Tue, 2009-11-24 at 08:40 -0800, Matthias Krack wrote:
> > > > > Hi Frank,
>
> > > > > I would suggest to try an older Intel compiler version like 11.0 or
> > > > > 10.1, since there might be problems with 11.1. For instance 10.1 is
> > > > > working for me with the today's CP2K version, ie. all regtests passed.
> > > > > Moreover, I would reduce the optimization level from -O3 to -O2.
>
> > > > > Matthias
>
> > > > > On Nov 24, 12:34 pm, Frank Uhlig <uhlig... at googlemail.com> wrote:
> > > > > > Hi all,
>
> > > > > > at first with respect to the last thread concerning the compilation
> > > > > > (for link see below) I found a setup that works for me and the
> > > > > > regtests. While trying to compile the whole thing with same setup but
> > > > > > just without the libintwrapper (meanig excluding the -
> > > > > > D__HAS_NO_ISO_C_BINDING and the corresponding libraries including
> > > > > > libstdc++) I get a "catastrophical error" (see output of make).
> > > > > > Any help or suggestions concerning this would be appreciated.
>
> > > > > > Thanks,
>
> > > > > > Frank
>
> > > > > > used compilers and libraries:
>
> > > > > > Intel Fortran/C compiler v11.1.059
> > > > > > OpenMPI v1.3.3
> > > > > > Libint v1.1.4
> > > > > > FFTW3 v3.2.2
>
> > > > > > thread concerning compilation with intel and pgi:
>
> > > > > >http://groups.google.com/group/cp2k/browse_thread/thread/45381865336e...
>
> > > > > > working arch-file:
>
> > > > > >http://cp2k.googlegroups.com/web/Linux-x86-64-intel-libintwrapper.pop...
>
> > > > > > make-output:
>
> > > > > >http://groups.google.com/group/cp2k/web/cp2k_make_error.log
>
> > > > > --
>
> > > > > You received this message because you are subscribed to the Google Groups "cp2k" group.
> > > > > To post to this group, send email to cp... at googlegroups.com.
> > > > > To unsubscribe from this group, send email to cp2k+uns... at googlegroups.com.
> > > > > For more options, visit this group athttp://groups.google.com/group/cp2k?hl=en.



More information about the CP2K-user mailing list