[CP2K:2359] Re: neither PGI nor Intel compiler working?

[Jörg Saßmannshausen]

Hi all,

unless I am mistaken, libint is c++ code, so the fortran compiler does not 
matter. Besides, I have build (albeit an older version) of cp2k with gfortran 
4.3.2. 
Just wanted to mention that. I have not tried the latest version of cp2k, so 
things may have changed here.
Personally, I would not recommend PGI for various reasons. If there are 
problems with the build, try and use the g95 compiler (for cp2k only) to 
check all the other libraries/stuff and see if the problem persists.

All the best

Jörg

On Monday 26 October 2009 15:56:41 Ondrej Marsalek wrote:
> Hi,
>
> On Mon, Oct 26, 2009 at 16:46, NUCP2K <chenh... at gmail.com> wrote:
> > Dear cp2k community,
> >
> >  I just downloaded the latest CVS snapshot and tried to compile the
> > codes using my once working ARCH files for PGI and Intel compilers on
> > Linux-x86-64 MPI machines. Unfortunately, none of the two compilers is
> > working though both can get the codes compiled successfully. It seems
> > to me that the "cp_fm_cholesky_decompose" is the cause of the runtime
> > trouble since my testing job always exit abnormally when the
> > subroutine is called.
>
> Please note that this can also mean (and in my experience usually
> does) that the data were corrupted sometime before this point, but the
> problem is only exposed because the matrix entering Cholesky
> decomposition is not well formed.
>
> >  Anyone has the experience on successfully compiling and running the
> > latest cp2k code on Linux PC MPI clusters? The sharing of your ARCH
> > files is highly appreciated.
>
> I had trouble with the Intel compiler as well, see the Google group
> archive. Other people have reported the same, I don't think anyone has
> claimed to have a working build of current CP2K with the Intel
> compiler, any version. I have a working (well, passed the tests) build
> with gfortran 4.3.3 and Intel MKL, I attach the arch file. It seems
> that at least gfortran 4.3.3 is needed if you want to build with
> libint, I am not sure what is the limitation without libint (if anyone
> does know, please tell me). I use current OpenMPI (now 1.3.3).
>
> Good luck,
> Ondrej
>
> >  Thanks.
> >
> > Hanning
>
> 


-- 
*************************************************************
Jörg Saßmannshausen
Research Fellow
University of Strathclyde
Department of Pure and Applied Chemistry
295 Cathedral St.
Glasgow
G1 1XL

email: jorg.sassm... at strath.ac.uk
web: http://sassy.formativ.net

Please avoid sending me Word or PowerPoint attachments.
See http://www.gnu.org/philosophy/no-word-attachments.html