[CP2K-user] [CP2K:14361] CP2K-7.1 Toolchain crash at ELPA instalation

Krzysztof Kolman krzyszt... at gmail.com
Fri Dec 11 20:15:01 UTC 2020
Previous message (by thread): [CP2K-user] [CP2K:14361] CP2K-7.1 Toolchain crash at ELPA instalation
Next message (by thread): [CP2K-user] cp2k not stopping
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
Hello Tiziano,

Thank you again for your help! After compiling psmp version of cp2k 
(openMPI based ), both of them, SIRIUS and ELPA are included. After the 
compilation I have benchmarked both of them and I got some results that I 
do not understand (ssmp and psmp were compiled with CUDA enabled). For 
testing, I was using my desktop pc, Ubuntu 20.04, Intel Core i7-4790, (4 
cores, 2 threads), Gtx 1660 super and 16 gb of RAM. As benchmark I was 
using H2O-64.inp. Please find below what I got:

1) time cp2k.ssmp -o H2O-64.out H2O-64.inp

real    8m7,011s
user    27m34,937s
sys    1m2,519s

2) time mpirun -np 4 cp2k.psmp -o H2O-64.out H2O-64.inp

no load on GPU, only video memory used
real    49m20,818s
user    361m47,512s
sys    2m30,998s

Why do I observe such huge difference? I run it twice and I got the same 
result. After that I started to play little bit.

3) time mpirun -np 4 -x OMP_NUM_THREADS=2 cp2k.psmp -o H2O-64.out H2O-64.inp

real    7m59,329s
user    41m36,361s
sys    0m47,529s

4) time mpirun -np 4 --map-by slot:PE=1 -x OMP_PLACES=threads -x 
OMP_PROC_BIND=SPREAD -x OMP_NUM_THREADS=2 cp2k.psmp -o H2O-64.out 
H2O-64.inp 

real    7m50,873s
user    41m25,381s
sys    0m46,038s

Tiziano, do you know why such huge discrepancies are observed? Thank you in 
advance for you comments!

Best regards,
Krzysztof
On Wednesday, December 9, 2020 at 10:25:01 PM UTC+1 tiz... at chem.uzh.ch 
wrote:

> Hi Krzysztof,
>
> I assume you used the `ssmp` (or the deprecated `sopt`) VERSION?
> SIRIUS and ELPA are only available when built with MPI support.
> So, you have to run:
>
> make ARCH=local VERSION=psmp
>
> Best,
> Tiziano
>
> On 12/9/20 7:19 PM, Krzysztof Kolman wrote:
> > Hi Tiziano,
> > 
> > Thank you for your help! This time it worked :) I have one more 
> > question. After the compilation I performed the regression testing. In 
> > the beginning I got following message:
> > 
> > CP2K supports:  cp2kflags: omp libint fftw3 libxc acc pw_cuda xsmm 
> > dbcsr_acc spglib libvori libbqb grid_cuda
> > Skipping QS/regtest-cdft-hirshfeld-2 : missing required feature : 
> parallel
> > Skipping QS/regtest-cdft-hirshfeld-2 : missing required feature : 
> mpiranks>1
> > Skipping SIRIUS/regtest-1 : missing required feature : sirius
> > Skipping QS/regtest-elpa-2 : missing required feature : elpa
> > Skipping QMMM/QS/regtest-cdft : missing required feature : parallel
> > Skipping QMMM/QS/regtest-cdft : missing required feature : 
> > mpiranks==2||mpiranks==4||mpiranks==6||mpiranks==16||mpiranks==24
> > Skipping QS/regtest-rma : missing required feature : mpi3
> > Skipping QS/regtest-rma : missing required feature : mpiranks==4
> > Skipping QS/regtest-cdft-3 : missing required feature : parallel
> > Skipping QS/regtest-cdft-3 : missing required feature : 
> > mpiranks==2||mpiranks==4||mpiranks==6||mpiranks==16||mpiranks==24
> > Skipping QS/regtest-pexsi : missing required feature : pexsi
> > Skipping QS/regtest-rma-3D : missing required feature : mpi3
> > Skipping QS/regtest-rma-3D : missing required feature : mpiranks==8
> > Skipping TMC/regtest_ana_on_the_fly : missing required feature : parallel
> > Skipping TMC/regtest_ana_on_the_fly : missing required feature : 
> mpiranks>2
> > Skipping TMC/regtest : missing required feature : parallel
> > Skipping TMC/regtest : missing required feature : mpiranks>1
> > Skipping QS/regtest-elpa-qr : missing required feature : elpa
> > Skipping TMC/regtest_ana_post_proc : missing required feature : parallel
> > Skipping TMC/regtest_ana_post_proc : missing required feature : 
> mpiranks>1
> > Skipping QS/regtest-elpa : missing required feature : elpa
> > Skipping Fist/regtest-quip : missing required feature : quip
> > Skipping Fist/regtest-plumed2 : missing required feature : plumed2
> > 
> > Here it claims that ELPA and SIRIUS are missing. Do you know maybe why?
> > 
> > Best regards,
> > Krzysztof
> > 
> > On Wednesday, December 9, 2020 at 4:14:07 PM UTC+1 
> > tiz... at chem.uzh.ch wrote:
> > 
> > Hi Krzysztof,
> > 
> > please use the latest git-master of CP2K if you want to try out a
> > CUDA-enabled ELPA. In that version you should not only be able to build
> > ELPA with support for a CUDA-kernel, but CP2K should actually be
> > able to
> > use it.
> > 
> > Best regards,
> > Tiziano
> > 
> > 
> > On 12/9/20 12:15 PM, Krzysztof Kolman wrote:
> > > Hello,
> > >
> > > I have encounter a problem while trying to compile cp2k at Ubuntu
> > 20.04
> > > with CUDA support. When I use following toolchain command:
> > > ./install_cp2k_toolchain.sh --with-libxsmm=install
> > > --with-openblas=install --with-fftw=install --with-reflapack=no
> > > --enable-cuda --enable-omp --gpu-ver=V100 --with-cmake=system
> > > --with-openmpi=install
> > >
> > > I get such error:
> > > ==================== Installing ELPA ====================
> > > elpa-2019.05.001.tar.gz is found
> > > Installing from scratch into
> > > /cp2k-7.1/tools/toolchain/install/elpa-2019.05.001
> > > ERROR: (./scripts/install_elpa.sh, line 103) Non-zero exit code
> > detected.
> > >
> > > From the make.log located at /build/elpa-2019.05.00/obj_no_thread
> > I can
> > > see that nvcc crashes:
> > > nvcc fatal   : Unknown option 'Wl,-rpath'
> > >
> > > Please find attached the make. log file as well as config.log and
> > > configure.log located at /build/elpa-2019.05.00/obj_no_threa.
> > >
> > > I think it is also worth to mention that I manage to compile cp2k
> > with
> > > flag --with-elpa=no.
> > >
> > > Thank you in advance for your help!
> > >
> > > Best regards,
> > > Krzysztof
> > >
> > > --
> > > 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 cp... at googlegroups.com
> > > <mailto:cp... at googlegroups.com>.
> > > To view this discussion on the web visit
> > >
> > 
> https://groups.google.com/d/msgid/cp2k/60525719-7909-44a5-b729-4314898ca143n%40googlegroups.com
> > <
> https://groups.google.com/d/msgid/cp2k/60525719-7909-44a5-b729-4314898ca143n%40googlegroups.com
> >
> > 
> > >
> > <
> https://groups.google.com/d/msgid/cp2k/60525719-7909-44a5-b729-4314898ca143n%40googlegroups.com?utm_medium=email&utm_source=footer
> > <
> https://groups.google.com/d/msgid/cp2k/60525719-7909-44a5-b729-4314898ca143n%40googlegroups.com?utm_medium=email&utm_source=footer
> >>.
> > 
> > 
> > -- 
> > Tiziano Müller
> > University of Zurich
> > Department of Chemistry
> > Winterthurerstrasse 190
> > CH-8057 Zürich
> > 
> > Tel: +41 44 63 54234
> > www.chem.uzh.ch <http://www.chem.uzh.ch>
> > tiz... at chem.uzh.ch
> > 
> > -- 
> > 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 cp... at googlegroups.com 
> > <mailto:cp... at googlegroups.com>.
> > To view this discussion on the web visit 
> > 
> https://groups.google.com/d/msgid/cp2k/c8f0d53a-bef6-44ae-ba86-e8752d0bc79dn%40googlegroups.com 
> > <
> https://groups.google.com/d/msgid/cp2k/c8f0d53a-bef6-44ae-ba86-e8752d0bc79dn%40googlegroups.com?utm_medium=email&utm_source=footer
> >.
>
> -- 
> Tiziano Müller
> University of Zurich
> Department of Chemistry
> Winterthurerstrasse 190
> CH-8057 Zürich
>
> Tel: +41 44 63 54234
> www.chem.uzh.ch
> tiz... at chem.uzh.ch
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.cp2k.org/archives/cp2k-user/attachments/20201211/4a8e7b67/attachment.htm>
Previous message (by thread): [CP2K-user] [CP2K:14361] CP2K-7.1 Toolchain crash at ELPA instalation
Next message (by thread): [CP2K-user] cp2k not stopping
Messages sorted by: [ date ] [ thread ] [ subject ] [ author ]
More information about the CP2K-user mailing list