[marco.fossati]

Hi all,

even if SU2 is compiled with parallel support (see configuration summary below). When I try:

mpirun -n 8 SU2_CFD config.cfg

I get 8 serial independent jobs instead of an actual parallel job. Is there anything I am missing?

Thanks
Marco

Here's the output of the configuration script:

checking build system type... x86_64-unknown-linux-gnu
checking host system type... x86_64-unknown-linux-gnu
checking target system type... x86_64-unknown-linux-gnu
checking for a BSD-compatible install... /usr/bin/install -c
checking whether build environment is sane... yes
checking for a thread-safe mkdir -p... /bin/mkdir -p
checking for gawk... gawk
checking whether make sets $(MAKE)... yes
checking whether make supports nested variables... yes
checking for style of include used by make... GNU
checking whether the C++ compiler works... yes
checking for C++ compiler default output file name... a.out
checking for suffix of executables...
checking whether we are cross compiling... no
checking for suffix of object files... o
checking whether we are using the GNU C++ compiler... yes
checking whether /usr/bin/mpicxx accepts -g... yes
checking dependency style of /usr/bin/mpicxx... gcc3
checking whether we are using the GNU C compiler... yes
checking whether /usr/bin/mpicc accepts -g... yes
checking for /usr/bin/mpicc option to accept ISO C89... none needed
checking dependency style of /usr/bin/mpicc... gcc3
checking whether /usr/bin/mpicc and cc understand -c and -o together... yes
checking for ranlib... ranlib
checking whether we are using the GNU C++ compiler... (cached) yes
checking whether /usr/bin/mpicxx accepts -g... (cached) yes
checking dependency style of /usr/bin/mpicxx... (cached) gcc3
checking how to run the C preprocessor... /usr/bin/mpicc -E
checking for grep that handles long lines and -e... /bin/grep
checking for egrep... /bin/grep -E
checking for ANSI C header files... yes
checking for sys/types.h... yes
checking for sys/stat.h... yes
checking for stdlib.h... yes
checking for string.h... yes
checking for memory.h... yes
checking for strings.h... yes
checking for inttypes.h... yes
checking for stdint.h... yes
checking for unistd.h... yes
checking size of short int... 2
checking size of int... 4
checking size of unsigned int... 4
checking size of long int... 8
checking size of float... 4
checking size of double... 8
checking size of void *... 8
checking X11/Intrinsic.h usability... no
checking X11/Intrinsic.h presence... no
checking for X11/Intrinsic.h... no
<<< Configuring library with Metis support >>>
<<< Configuring library with Parmetis support >>>
checking for /usr/local/lib/libcgns.a... yes
checking for /usr/local/include/cgnslib.h... yes
checking that generated files are newer than configure... done
configure: creating ./config.status
config.status: creating externals/tecio/Makefile
config.status: creating externals/metis/Makefile
config.status: creating externals/parmetis/Makefile
config.status: creating Makefile
config.status: creating externals/Makefile
config.status: creating Common/lib/Makefile
config.status: creating SU2_CFD/obj/Makefile
config.status: creating SU2_DOT/obj/Makefile
config.status: creating SU2_MSH/obj/Makefile
config.status: creating SU2_DEF/obj/Makefile
config.status: creating SU2_SOL/obj/Makefile
config.status: creating SU2_GEO/obj/Makefile
config.status: creating SU2_PY/Makefile
config.status: executing depfiles commands


-------------------------------------------------------------------------
| ___ _ _ ___ |
| / __| | | |_ ) Release 4.2.0 'Cardinal' |
| \__ \ |_| |/ / |
| |___/\___//___| Suite |
| |
-------------------------------------------------------------------------
| SU2 Lead Dev.: Dr. Francisco Palacios, Francisco.D.Palacios@boeing.com|
| Dr. Thomas D. Economon, economon@stanford.edu |
-------------------------------------------------------------------------
| SU2 Developers: |
| - Prof. Juan J. Alonso's group at Stanford University. |
| - Prof. Piero Colonna's group at Delft University of Technology. |
| - Prof. Nicolas R. Gauger's group at Kaiserslautern U. of Technology. |
| - Prof. Alberto Guardone's group at Polytechnic University of Milan. |
| - Prof. Rafael Palacios' group at Imperial College London. |
-------------------------------------------------------------------------
| Copyright (C) 2012-2016 SU2, the open-source CFD code. |
| |
| SU2 is free software; you can redistribute it and/or |
| modify it under the terms of the GNU Lesser General Public |
| License as published by the Free Software Foundation; either |
| version 2.1 of the License, or (at your option) any later version. |
| |
| SU2 is distributed in the hope that it will be useful, |
| but WITHOUT ANY WARRANTY; without even the implied warranty of |
| MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU |
| Lesser General Public License for more details. |
| |
| You should have received a copy of the GNU Lesser General Public |
| License along with SU2. If not, see <http://www.gnu.org/licenses/>. |
-------------------------------------------------------------------------

Build Configuration Summary:

Source code location: /home/marco/Useless/solver-su2/SU2
Install location: /home/marco/Useless/solver-su2
Version: 4.2.0
C++ Compiler: /usr/bin/mpicxx
C Compiler: /usr/bin/mpicc
Preprocessor flags: -DHAVE_MPI
Compiler flags: -O3
Linker flags:
MPI support: yes
Metis support: yes
Parmetis support: yes
TecIO support: no
CGNS support: yes
HDF5 support: no
SZIP support: no
ZLIB support: no
Mutation++ support: no
Jsoncpp support: no
LAPACK support: no
Datatype support:
double yes
complex no
codi_reverse no
codi_forward no

External includes: -DHAVE_PARMETIS -I$(top_srcdir)/externals/parmetis/include -DHAVE_METIS -I$(top_srcdir)/externals/metis/include
External libs: $(top_builddir)/externals/parmetis/libparmetis.a $(top_builddir)/externals/metis/libmetis.a

Build SU2_CFD: yes
Build SU2_DOT: yes
Build SU2_MSH: yes
Build SU2_DEF: yes
Build SU2_SOL: yes
Build SU2_GEO: yes

Please be sure to add the $SU2_HOME and $SU2_RUN environment variables,
and update your $PATH (and $PYTHONPATH if applicable) with $SU2_RUN.

[asonda]

Hi Marco,

try this: parallel_computation.py -f config.cfg -n 8

Cheers,

Alberto

[marco.fossati]

Thanks Alberto, but still no dice...same behaviour as before

I am checking the log of the make and the make install, but as of now I cannot see anything major.

Cheers
Marco

[asonda]

Stupid question: you say that they are running in serial because they are all running on the same core (and when you kill one process the other ones are not killed as well)?

[marco.fossati]

What I see is that I get printed the same exact line 8 times, i.e.

101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.
101217 triangles.
66956 quadrilaterals.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.

------------------------- Geometry Preprocessing ------------------------
Setting point connectivity.
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Recomputing point connectivity.
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Renumbering points (Reverse Cuthill McKee Ordering).
Recomputing point connectivity.
Recomputing point connectivity.
Recomputing point connectivity.
Recomputing point connectivity.
Recomputing point connectivity.
Recomputing point connectivity.

I believe that this is the typical output of a code which is not running properly in parallel. Other posts indicate that this is due to an incorrect configuration (with no metis/parmetis), which I do not have ( see original post) and I get no error / warning while compiling, except for this warning:

ar: `u' modifier ignored since `D' is the default (see `U')

which seems related to a way of updating the libraries. But since I have removed everything and made a clean installation from scratch, this might not be relevant for my case.

Thanks
Marco

[talbring]

Hi Marco,

this might solve your problem : https://github.com/su2code/SU2/wiki/...ne-for-example

Tim

[marco.fossati]

Hi Tim,

this was my first thought, but unless the printout I get from the configure script and make install (see my previous posts) is faulty, I am compiling with METIS. PARMETIS and the mpicc and mpicxx, which to the best of my knowledge is saying I am doing things in parallel.

Any other suggestion would be welcome.

Marco

[hlk]

Originally Posted by marco.fossati

Hi Tim,

this was my first thought, but unless the printout I get from the configure script and make install (see my previous posts) is faulty, I am compiling with METIS. PARMETIS and the mpicc and mpicxx, which to the best of my knowledge is saying I am doing things in parallel.

Any other suggestion would be welcome.

Marco

Based on what you've described (compiling from source a few times, and changing configuration options), it may be possible that you have multiple versions of the SU2 executables on your system. Try 'which SU2_CFD' to check the location of the executable you are calling.

[LaSerpe]

It could also be related to which mpi implementation you are using on your machine.
Do you have openmpi or mpich?
if you have both of them you may have to specify the right one (which is the one you used to compile SU2) in order to run SU2 on multiple cores.
Try using either "mpirun.openmpi" or "mpirun.mpich"

[RcktMan77]

I would have to agree with LaSerpe. This behavior suggests that the mpirun that your shell environment is pointing to is not from the same MPI installation that was used to compile SU2.

[PhilipA]

Hello,

Were you able to solve this problem?

Best Regards

Philip

[bigfootedrockmidget]

if you run SU2 in parallel on N cores, and you see SU2 output messages being repeated N times, then your mpi is not set up correctly. Make sure that you have mpi installed correctly, then make sure that you are using SU2 with the correct compilation option to be able to use mpi. Please read the installation carefully:


https://su2code.github.io/docs_v7/Installation/


And if you still have problems, please open a new issue and give us the details of your setup (windows/linux, mpich or openmp, etc...).

[Sakun]

Originally Posted by bigfootedrockmidget

if you run SU2 in parallel on N cores, and you see SU2 output messages being repeated N times, then your mpi is not set up correctly. Make sure that you have mpi installed correctly, then make sure that you are using SU2 with the correct compilation option to be able to use mpi. Please read the installation carefully:


https://su2code.github.io/docs_v7/Installation/


And if you still have problems, please open a new issue and give us the details of your setup (windows/linux, mpich or openmp, etc...).

Hi,


Sorry to reply on this thread.
I believe i have installed SU2 using source code (SU2 version 8.0.1 "Harrier") . For this i have followed following codes,

./meson.py build
./ninja -C build install

Then i have tried to run "QuickStart" ,then i have got an error as shown in the picture.


i have installed MPICH and specified it path as well.

Appreciate if you could guide me on this

[bigfootedrockmidget]

try this for mpi, or set it to disabled for single core:

./meson.py build --optimization=2 -Dwith-mpi=enabled --prefix=/home/Codes/su2HTML]

[Sakun]

Originally Posted by bigfootedrockmidget

try this for mpi, or set it to disabled for single core:

./meson.py build --optimization=2 -Dwith-mpi=enabled --prefix=/home/Codes/su2HTML]

Hi bigfoot,

Thank you very much for your reply and the suggestion.

I have further did some dig up related to my case, then I have found out that for the source code compiled SU_2, mpirun works (used this code;

 mpirun -n 10 SU_2 inv_NACA0012.cfg

) but single core simulation didn’t work (used this code;

SU2_CFD inv_NACA0012.cfg

).

As for the precompiled binary SU2, single core simulation worked but when I used parallel code (mpi), simulation just repeated 10 times and it looks quite chaotic.
Since I will be working on 5.5 million mesh in the HPC, I would go with the source code compiled SU2 package.

Also, is there a way to get to know that allocated cores in the mpirun code have been actually worked in the simulation ?, like I have tested with

mpirun -n 10 SU_2 inv_NACA0012.cfg & mpirun -n 6 SU_2 inv_NACA0012.cfg

, but I couldn’t tell a different.

Thank you very much for your time.

[bigfootedrockmidget]

Hi,
Well if you use 10 cores instead of 5, it should finish 2 times faster. If you print the time per iteration,

SCREEN_OUTPUT= WALL_TIME, ...

Then you can see how much time each iteration takes.