compilation issue with cp2k-2024.2 (both gnu and Intel)

124 views
Skip to first unread message

giacomo buccella

unread,
Sep 12, 2024, 6:04:17 AM9/12/24
to cp2k
Hello everybody,
I'm trying to compile the code on a cluster computer, but I'm finding several problems during compilation. I tried both gnu compiler and Inter compiler.
For the Intel version, I wanted to install some packages in local and tell the code to use those versions. In many cases I was forced to do so, since the toolchain failed to install the package itsself.
Below are the examples of toolchain commands I used during the (succesfull) configurations:

****INTEL****
>>>./install_cp2k_toolchain.sh \
--with-cosma=/home/TTD/giacomo/cosma-target \
--mpi-mode=intelmpi \
--math-mode=mkl \
--with-gcc=system \
--with-intel=system \
--with-cmake=/home/TTD/giacomo/cmake \
--with-ninja=no \
--with-intelmpi=/opt/intel/oneapi/mpi/2021.10.0 \
--with-libxc \
--with-libint \
--with-libgrpp \
--with-fftw \
--with-acml=no \
--with-mkl=/opt/intel/oneapi/mkl/2023.2.0 \
--with-libxsmm \
--with-elpa \
--with-dftd4 \
--with-scalapack=/opt/intel/oneapi/mkl/2023.2.0/lib/intel64 \
--with-openblas=no \
--with-spla=/home/TTD/giacomo/spla-target \
--with-sirius=no \
--with-pexsi=no
...
>>>make -j 4 ARCH=local VERSION="psmp ssmp popt sopt"

****GNU****
>>>./install_cp2k_toolchain.sh \
--with-cosma=no \
--mpi-mode=openmpi \
--math-mode=openblas \
--with-gcc=install \
--with-intel=no \
--with-cmake=install \
--with-ninja=no \
--with-openmpi=install \
--with-libxc \
--with-libint \
--with-libgrpp \
--with-fftw \
--with-acml=no \
--with-mkl=no \
--with-libxsmm \
--with-dftd4 \
--with-scalapack \
--with-openblas \
--with-spla=no \
--with-sirius=no \
--with-pexsi=no

>>>make -j 4 ARCH=local VERSION="psmp ssmp popt sopt"

INTEL: apparently the compilation went fine, but when I tried to submit a test calculation, it gave segmentation fault. I also tried the trick of setting OMP_STACKSIZE, but there was no effect.

GNU: the compilation was killed due to several errors, the first of which was the following:

make[3]: *** [/home/TTD/giacomo/cp2k-gnu/cp2k-2024.2/Makefile:495: /home/TTD/giacomo/cp2k-gnu/cp2k-2024.2/obj/git-ref] Error 1
make[2]: *** [/home/TTD/giacomo/cp2k-gnu/cp2k-2024.2/Makefile:146: all] Error 2
make[1]: *** [/home/TTD/giacomo/cp2k-gnu/cp2k-2024.2/Makefile:128: sdbg] Error 2
make[1]: *** Waiting for unfinished jobs....

Can anyone help me out with one of the two versions? I really cannot figure out what could be the cause.

Many thanks in advance

Giacomo

Sagnik Mukhopadhyay

unread,
Sep 12, 2024, 6:25:09 AM9/12/24
to cp...@googlegroups.com

Hi,

After ./install.... You have to copy the arch files from /cp2k/tools/toolchain/install/arch to /cp2k/arch. Then you need
source /cp2k/tools/toolchain/install/setup
After that you can compile your make file.

Regards
Sagnik


--
You received this message because you are subscribed to the Google Groups "cp2k" group.
To unsubscribe from this group and stop receiving emails from it, send an email to cp2k+uns...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/cp2k/0c48a657-bc56-4956-9a7c-0da02fa8357fn%40googlegroups.com.

giacomo buccella

unread,
Sep 12, 2024, 7:19:02 AM9/12/24
to cp2k
Dear Sagnik,
thank you for your reply.
Yes, I have done that each time, the result is described.

Johann Pototschnig

unread,
Sep 13, 2024, 5:15:05 AM9/13/24
to cp2k
For the GNU installation you use:
--with-gcc=install
--with-openmpi=install
which installs a compiler and MPI distribution. 
If you are on a cluster and want efficient computations you should use one of the already installed MPI distributions. Then it is also a good idea to use the corresponding compiler.

Additionally you can check https://dashboard.cp2k.org/index.html for which compilers are supported and some known problems are listed in https://www.cp2k.org/dev:compiler_support
Which INTEL OneAPI version are you using?

Johann Pototschnig

unread,
Sep 13, 2024, 5:16:35 AM9/13/24
to cp2k
More detailed error messages would also be useful.
Reply all
Reply to author
Forward
0 new messages