Code_saturne 1.4b released!
16 years 7 months ago #2247
by Claus
Code_Aster release : STA11.4 on OpenSUSE 12.3 64 bits - EDF/Intel version
Replied by Claus on topic Re:Code_saturne 1.4b released!
Hey again - I've just run my first successful simulation! 
I've started from scratch and just followed your instructions - I think my mistake was to approach 1.4b as I did with 1.3.2 - installing the aux-libs from source. The ubuntu-repos work just fine
Now, I do have a problem with running a parallel simulation - this is the error:
/root/tmp_Saturne/ASDFASDF.CASE1.09151846/cs14.exe -mpi -param strati.xml
cs14.exe does not accept the -mpi switch - im guessing it did not find the required openmpi libs during installation.
which brings me to a request:
could it be possible for the installation script to summerize installed and found libs after installation (or before), it's kinda annoying to have to recompile the whole lot because I forgot to install cgns
Otherwise - I think I forgot to emphasize how thrilled I am with the progress and the translation into English! I'll try to help whereever I can.
I've started from scratch and just followed your instructions - I think my mistake was to approach 1.4b as I did with 1.3.2 - installing the aux-libs from source. The ubuntu-repos work just fine
Now, I do have a problem with running a parallel simulation - this is the error:
/root/tmp_Saturne/ASDFASDF.CASE1.09151846/cs14.exe -mpi -param strati.xml
cs14.exe does not accept the -mpi switch - im guessing it did not find the required openmpi libs during installation.
which brings me to a request:
could it be possible for the installation script to summerize installed and found libs after installation (or before), it's kinda annoying to have to recompile the whole lot because I forgot to install cgns
Otherwise - I think I forgot to emphasize how thrilled I am with the progress and the translation into English! I'll try to help whereever I can.
Code_Aster release : STA11.4 on OpenSUSE 12.3 64 bits - EDF/Intel version
16 years 7 months ago #2248
by David Monfort
Replied by David Monfort on topic Re:Code_saturne 1.4b released!
Great!
I think you're guessing right. The install setup may not find the local MPI distribution. I think you should write your own macros_Linux.mk (or the Linux_x86_64 one) and install the kernel "by hand". I'm joining the file I use to this post.
Furthermore, if you want to build Code_Saturne with CGNS support you should disable the shared library creation (for the kernel) by setting BUILD_SO to 0 as indicated in my macros file. Indeed, the CGNS configure machinery is rather "ill-written" (
it uses rather old version of autotools) and no shared library is created by default. You can still install a shared library for CGNS, but then you have to tweak the FVM and preprocessor installation so that they find the CGNS library... and it is a bit more tricky 
As the kernel shared library is of no use for standard usage, I suggest you to use the macros file I post.
Actually, the installation script do summarize what it found/installed! Just have a look at your setup file, it is re-written by the installer (path to libraries and yes/no columns for installation and use). What cannot do the installer is deal with MPI (or other libraries) installation where the paths are not standard (i.e. /usr/include for the headers /usr/lib for the libraries). For OpenMPI on Ubuntu you might try to use the following path in the setup file: /usr/lib/openmpi.
Thank you for the comment on the translation work! Sure, it is very time-consuming... we're trying to have the specific physics translated in a next version (and code comments in a further version but that's another story). If you find some errors or else, please give us some feedback! Any help is of course welcome. <br /><br />Post edited by: David Monfort, at: 2008/09/15 22:06
I think you're guessing right. The install setup may not find the local MPI distribution. I think you should write your own macros_Linux.mk (or the Linux_x86_64 one) and install the kernel "by hand". I'm joining the file I use to this post.
Furthermore, if you want to build Code_Saturne with CGNS support you should disable the shared library creation (for the kernel) by setting BUILD_SO to 0 as indicated in my macros file. Indeed, the CGNS configure machinery is rather "ill-written" (
it uses rather old version of autotools) and no shared library is created by default. You can still install a shared library for CGNS, but then you have to tweak the FVM and preprocessor installation so that they find the CGNS library... and it is a bit more tricky As the kernel shared library is of no use for standard usage, I suggest you to use the macros file I post.Actually, the installation script do summarize what it found/installed! Just have a look at your setup file, it is re-written by the installer (path to libraries and yes/no columns for installation and use). What cannot do the installer is deal with MPI (or other libraries) installation where the paths are not standard (i.e. /usr/include for the headers /usr/lib for the libraries). For OpenMPI on Ubuntu you might try to use the following path in the setup file: /usr/lib/openmpi.
Thank you for the comment on the translation work! Sure, it is very time-consuming... we're trying to have the specific physics translated in a next version (and code comments in a further version but that's another story
). If you find some errors or else, please give us some feedback! Any help is of course welcome. <br /><br />Post edited by: David Monfort, at: 2008/09/15 22:06
16 years 7 months ago #2250
by Claus
Code_Aster release : STA11.4 on OpenSUSE 12.3 64 bits - EDF/Intel version
Replied by Claus on topic Re:Code_saturne 1.4b released!
well I forgot about the setup file - how about that 
My problem is that the generated (2xprocessor) script from SaturneGUI, launches with the -mpi switch. cs24.exe does not recognize this switch:
[code:1]Usage: ./cs14.exe [options]
Command line options:
-ec, --echo-comm print SYRTHES communications;
-1 : only error (default)
0 : print headers of messages only
n : print headers of messages and
the n first and last elements as well
-solcom Stand-alone kernel with "geomet" mesh in
SolCom format (obsolete);
-syrthes SYRTHES coupling at selected faces
-2d : the associated SYRTHES mesh is 2D
-X : prefered projection axis
for a 2D SYRTHES mesh
-Y : prefered projection axis
for a 2D SYRTHES mesh
-Z : prefered projection axis
for a 2D SYRTHES mesh (default)
-socket: communication with IP socket
(named pipes by default)
-color <number(s)>: color numbers of faces
to select
-group <name(s)>: group names of faces
to select
-invsel: invert selection
-q, --quality verifications
-1: no tests activated (default)
0: only the initialization is tested
1 &#65533; 5: elementary tests are activated
-cwf <criterion> cut warped faces
-post: activate the post-processing related
to the cutting of warped faces
--benchmark elementary operations performance
operations done only once
for light MPI traces
--log output redirection for rank -1 or 0:
0: standard output
1: output in "listing" (default)
--logp output redirection for rank > 0:
-1: remove output (default)
0: no redirection (if independant
terminals, debugger type)
1: output in "listing_n<rang>"
-param [file_name] parameter file
-h, --help this help message
[/code:1]
So far, the executables compile fine, but the '-mpi' switch kills the calculation. I can't remember the 1.3.2 syntax...
My problem is that the generated (2xprocessor) script from SaturneGUI, launches with the -mpi switch. cs24.exe does not recognize this switch:
[code:1]Usage: ./cs14.exe [options]
Command line options:
-ec, --echo-comm print SYRTHES communications;
-1 : only error (default)
0 : print headers of messages only
n : print headers of messages and
the n first and last elements as well
-solcom Stand-alone kernel with "geomet" mesh in
SolCom format (obsolete);
-syrthes SYRTHES coupling at selected faces
-2d : the associated SYRTHES mesh is 2D
-X : prefered projection axis
for a 2D SYRTHES mesh
-Y : prefered projection axis
for a 2D SYRTHES mesh
-Z : prefered projection axis
for a 2D SYRTHES mesh (default)
-socket: communication with IP socket
(named pipes by default)
-color <number(s)>: color numbers of faces
to select
-group <name(s)>: group names of faces
to select
-invsel: invert selection
-q, --quality verifications
-1: no tests activated (default)
0: only the initialization is tested
1 &#65533; 5: elementary tests are activated
-cwf <criterion> cut warped faces
-post: activate the post-processing related
to the cutting of warped faces
--benchmark elementary operations performance
operations done only once
for light MPI traces
--log output redirection for rank -1 or 0:
0: standard output
1: output in "listing" (default)
--logp output redirection for rank > 0:
-1: remove output (default)
0: no redirection (if independant
terminals, debugger type)
1: output in "listing_n<rang>"
-param [file_name] parameter file
-h, --help this help message
[/code:1]
So far, the executables compile fine, but the '-mpi' switch kills the calculation. I can't remember the 1.3.2 syntax...
Code_Aster release : STA11.4 on OpenSUSE 12.3 64 bits - EDF/Intel version
Moderators: catux
Time to create page: 0.152 seconds