Hello Mario,

Codeaster src 16.1.8 version (seems to be latest?) (htt ps://sourceforge.net/p/codeaster/src/ci/16.1.8/tree/) from Sourceforge works with the newest salome_meca 2021.0.0-2-20211014 container with yours and ing. nicola's MPI version recipe. Just update the container and version names when using commands and update also the pkginfo.py.

If I understood correctly that older versions of Codeaster src doesn't work because this new container salome_meca 2021.0.0-2-20211014 have newer version of MUMPS.

Best regards,
VonPire

Hello,

Yep, it would be nice to test code_aster with AMD's own optimized compilers and mathlibs and see the difference to GNU and Intel. However AMD provides own EPYC specific compiler options for using AOCC, GNU and Intel compilers and Intel seems to work with sequential version of aster-full-src 14.6.

I could give the GNU a chance with different optimization flags. I have tried only with default flags.

I haven't managed to get working yet the openMP and MPI version of aster-full-src 14.6.

Best regards,
VonPire

Hello Mario,

Thank you for your reply.

I gained 30% performance gain (compared to default GNU compilers and mathlib) with "onecore/sequential" version of aster-full-src-14.6.0 with my new workstation when I compiled it with Intel compilers and Intel MKL. I guess that the performance gain would be bigger when using parallel OpenMP and MPI version that has been compiled with Intel compilers and Intel MKL. So I think that there is good motivation to investigate and try this to get working.

Best regards,
VonPire

Hello Nicolas,

Thank you for your answer.

Yes, I know that aster tries to use gfortran instead of ifort, even I set ifort correctly in setup.cfg. The same Intel compilers+ Intel MKL setup.cfg works fine with aster-full-src-14.6.0-1. I got 30% more performance with that Intel setup.cfg in one core 14.6 version.

Maybe we can forget the aster-full-src-15.2, because there is no any benefit to get it working. aster-full-src-15.2 has old mumps 5.2.1 version which isn't compatible anymore with the newest aster-src from sourceforge. Newest aster-src have MUMPS 5.4.1c.

Next question is how can I build codeaster MPI version with Intel compilers and Intel MKL to salome-meca .sif container ?

Best regards.
VonPire

Hello,

I'm still stuck with this. I even tried to compile "aster-full-src-15.2.0" std version with Intel oneApi compilers, but there was error with aster installation. Everything else (MUMPS,med, hdf5,etc.) was installed successfully. 

Here is some setup.log details:

Compiler variables for aster (set as environment variables):
export               CC='/opt/intel/oneapi/compiler/2022.0.2/linux/bin/intel64/icc'
export           CFLAGS='-O3 -traceback -fPIC'
export       CFLAGS_DBG='-g  -traceback -fPIC'
export    CFLAGS_OPENMP='-openmp'
export              CXX='/opt/intel/oneapi/compiler/2022.0.2/linux/bin/intel64/icpc'
export           CXXLIB='-L/usr/lib/gcc/x86_64-linux-gnu/9 -lstdc++'
export          DEFINED='LINUX64 _USE_INTEL_IFORT _USE_OPENMP _DISABLE_MATHLIB_FPE'
export              F90='/opt/intel/oneapi/compiler/2022.0.2/linux/bin/intel64/ifort'
export         F90FLAGS='-O3 -fpe0 -traceback -fPIC'
export     F90FLAGS_DBG='-g  -fpe0 -traceback -fPIC'
export      F90FLAGS_I8=' -i8 -r8'
export  F90FLAGS_OPENMP='-openmp'
export               LD='/opt/intel/oneapi/compiler/2022.0.2/linux/bin/intel64/ifort'
export          LDFLAGS='-nofor_main'
export   LDFLAGS_OPENMP='-openmp'
export          MATHLIB='-Wl,--start-group -L/opt/intel/oneapi/mkl/2022.0.2/lib/intel64 -lmkl_intel_lp64 -L/opt/intel/oneapi/mkl/2022.0.2/lib/intel64 -lmkl_sequential -L/opt/intel/oneapi/mkl/2022.0.2/lib/intel64 -lmkl_core -Wl,--end-group'
export         OTHERLIB='-L/usr/lib/x86_64-linux-gnu -lpthread -L/usr/lib/x86_64-linux-gnu -lz -L/usr/lib/x86_64-linux-gnu -lpthread'

Filling cache...                                                       [  OK  ]
Checking permissions...                                                [  OK  ]

>>> Extraction <<<

entering directory '/tmp/install_aster.1717957'
Extracting aster-15.2.0.tgz...                                         [  OK  ]
--- 30900 files extracted
leaving directory '/tmp/install_aster.1717957'

>>> Configuration <<<

entering directory '/tmp/install_aster.1717957/aster-15.2.0'
leaving directory '/tmp/install_aster.1717957/aster-15.2.0'

>>> Configuration <<<

entering directory '/tmp/install_aster.1717957/aster-15.2.0'
leaving directory '/tmp/install_aster.1717957/aster-15.2.0'

>>> Configuration <<<

entering directory '/tmp/install_aster.1717957/aster-15.2.0'
Command line : export PATH=/tmp/tmp86r0b7_q:${PATH} ; . env.d/aster_full_std.sh ; ./waf configure --install-tests --prefix=/home/USER/aster/15.2
configure aster installation...                                       
Command output :
configure aster installation...                                        [FAILED]
Exit code : 1
readlink: missing operand

In the config.log for some reason aster wants to use gfortran even ifort was set up:

/usr/bin/gfortran
----------------------------------------
Checking for C compiler version
icc 20.2.1
----------------------------------------
Checking for Fortran compiler version
gfortran 9.4.0
----------------------------------------
fortran link verbose flag
==>
        PROGRAM MAIN
        END

<==
[1/2] Compiling build/std/.conf_check_e01dd803d6cb274f73e8aa4a2b8a1f70/test.f

['/usr/bin/gfortran', '-O3', '-fpe0', '-traceback', '-fPIC', '-c', '-o/tmp/install_aster.1717957/aster-15.2.0/build/std/.conf_check_e01dd803d6cb274f73e8aa4a2b8a1f70/testbuild/release/test.f.1.o', '/tmp/install_aster.1717957/aster-15.2.0/build/std/.conf_check_e01dd803d6cb274f73e8aa4a2b8a1f70/test.f']
err: gfortran: error: unrecognized command line option ‘-fpe0’
gfortran: error: unrecognized command line option ‘-traceback’

from /tmp/install_aster.1717957/aster-15.2.0: Test does not build: Traceback (most recent call last):
  File "/tmp/install_aster.1717957/aster-15.2.0/.waf3-2.0.20-df7687050314fa98c5daa534cb234b8c/waflib/Configure.py", line 335, in run_build
    bld.compile()
  File "/tmp/install_aster.1717957/aster-15.2.0/.waf3-2.0.20-df7687050314fa98c5daa534cb234b8c/waflib/Build.py", line 176, in compile
    raise Errors.BuildError(self.producer.error)
waflib.Errors.BuildError: Build failed
-> task in 'testprog' failed with exit status 1 (run with -v to display more information)

from /tmp/install_aster.1717957/aster-15.2.0: The configuration failed
==>
        PROGRAM MAIN
        END

So I'm not able to install even "aster-full-src-15.2.0" with Intel OneApi compilers. Tips and help would be welcome.

Best regards,
VonPire

Hello,

Thank you Mario for your reply.

I investigated processors from the blue company and AMD and came to conclusion that EPYC 7443P had best performance when considering 1.5k€ to 2k€ processor. So I ordered yesterday EPYC 7443P and ROMED8-2T motherboard. I was lucky and found them with a good price from Poland. I searched over day for them, because they were out of stock nearly every online shop at nordic countries and Germany.

Now I'm thinking about what RAM memory I should choose. EPYC 7443P supports Octa channel memory and DDR4-3200MHz. Capacity would be 258GB so I need 8 memory stick with 32Gb memory. I have understood that the memory for EPYC should be also ECC and registered. So I'm searching DDR4-3200 ECC RDIMM 32Gb memory sticks.

Candidates could be:
Kingston KSM32RD4/32MEI
MICRON MTA18ASF4G72PDZ-3G2E1

However the question arises that should I use Single Rank (1RX4 or 1RX8) or Double Rank (2RX4 or 2RX8) memory sticks to achieve maximum memory bandwidth and latencies or speed ? Do you guys have any knowledge are there any benefit with single rank vs double rank when it comes to RAM with FEA simulations ?

Best Regards,
VonPire

Second COMM file

MED file

Hello Micke,

Have you tried does the Mortar algorithm work with friction ? At least In DEFI_CONTACT it is possible to use FORMULATION=CONTINUE, ALGO_CONT=LAC, APPARIEMENT=MORTAR, FROTTEMENT=COULOMB, COULOMB=0.1

At least Mortar should work without friction FROTTEMENT=SANS...

Mortar is SEGMENT-SEGMENT or FACE-FACE type contact and should capture penetration and contact pressures very precisely and there is no need to adjust penalization parameters.

Remember that you have to precondition the slave surface with CREA_MAILLAGE, DECOUPE_LAC when using Mortar algorithm.

For more info refer to U4.44.11 (DEFI_CONTACT), U2.04.04 (General info and tips for contacts), R5.03.52 (Continue method), R5.03.55 (Mortar method) and code-aster.org/V2/UPLOAD/DOC/Formations/13-contact-friction.pdf (Contact and friction presentation)

Best regards,
VonPire

Hello Ioannis,

Thank you for answering my question.

Would it be possible with other post-processing program then ? GMSH ?

If I remember correctly even CalculiX CGX can do the "3d" visualization for CalculiX shells and beams results ?

Regards,
VonPire