CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   SU2 Installation (http://www.cfd-online.com/Forums/su2-installation/)
-   -   new user: compiling problems (http://www.cfd-online.com/Forums/su2-installation/112310-new-user-compiling-problems.html)

jentink January 25, 2013 14:26

new user: compiling problems
 
I just downloaded su2 and am trying to compile on a cluster here at NASA Langley that uses the module envrionment for desired packages.

When I enable mpi, tecplot, cgns, and metis, I am getting some c++ errors on int types. When I merely try the serial version (configure, make) with nothing special added, I get "omp.h" not found message.

With the latter case, when all 4 packages are enabled, I have the intel_2011.9.293 module loaded, the tecplot module loaded, and the mpich2_intel module loaded. I am using mpicxx for the mpi compiler command.

The config.log shows an error on conftest.cc: mpicxx -c -g -O2 conftest.cc >&5
error: identifier "exit" is undefined exit (42);

When I run make, I get overloaded virtual function warnings:
-------------------------------------------------------------------

then mv -f "../src/.deps/libSU2_a-geometry_structure.Tpo" "../src/.deps/libSU2_a-geometry_structure.Po"; else rm -f "../src/.deps/libSU2_a-geometry_structure.Tpo"; exit 1; fi
../src/../include/geometry_structure.hpp(580): warning #654: overloaded virtual function "CGeometry::SetVertex" is only partially overridden in class "CPhysicalGeometry"
class CPhysicalGeometry : public CGeometry {
^
----------------------------------------------------------------------------------------

and errors on type "int":
--------------------------------------------------------------------------------------

../src/geometry_structure.cpp(4298): error: argument of type "int *" is incompatible with parameter of type "idx_t={int64_t={long}} *"
METIS_SetDefaultOptions(options);
^

jentink January 25, 2013 14:27

I forgot to add - my config line looks like this:


./configure CC="icpc -fargument-alias" --prefix=/u/tjentink/su2 --with-MPI=mpicxx --with-Tecio=/usr/local/pkgs-modules/tecplot360-2012R1_64_K2/lib --with-Metis-lib=/u/tjentink/su2/lib --with-Metis-include=/u/tjentink/su2/include --with-Metis-version=5 --with-CGNS-lib=/usr/local/pkgs-modules/cgnslib_3.1.3_64/lib --with-CGNS-include=/usr/local/pkgs-modules/cgnslib_3.1.3_64/include

fpalacios January 28, 2013 15:18

Thanks a lot for using our software.

We are aware of the warnings with the intel compilation (no warnings using gnu). But, it looks like a problem with the METIS version, which one are you using?

Best,
Francisco

jentink January 28, 2013 15:25

Right now I am using version 5 of metis (see compile line above). The cluster I am using has a Metis module, but when I load it, I only get the library file - there is no include files. Should I download and use version 4?

Any idea why I can't compile in serial? When trying that, I get a message saying omp.h can't be found.

fpalacios January 28, 2013 15:28

Quote:

Originally Posted by jentink (Post 404600)
Right now I am using version 5 of metis (see compile line above). The cluster I am using has a Metis module, but when I load it, I only get the library file - there is no include files. Should I download and use version 4?

Any idea why I can't compile in serial? When trying that, I get a message saying omp.h can't be found.

I'm gonna try with metis 5.0 but, for the time being, could you please try with metis-4.0.3?

Thanks!
Francisco

jentink January 28, 2013 16:07

I used Metis 4.0.3 and it all seemed to compile this time.

jentink January 28, 2013 16:59

When I try to run any of the pyton commands, I'm getting a syntax error when libSU2.py is imported.


File "/misc/home1/tjentink/su2/bin/libSU2.py", line 1289
except OSError as e:

jentink January 29, 2013 12:36

Quote:

Originally Posted by jentink (Post 404625)
When I try to run any of the pyton commands, I'm getting a syntax error when libSU2.py is imported.


File "/misc/home1/tjentink/su2/bin/libSU2.py", line 1289
except OSError as e:

python version issue. I switched to python 3.2 and this went away. However, the has_key attribute went away in version 3 of python. I had to replace all the d.has_key('a') instances in libSU2.py with 'a' in d.

jentink January 29, 2013 13:43

in mesh_adaptation.py, I had errors whenever the following occurs:

if save_all:
print "Saving cycle" + .....


I had to put the line in parenthesis :

if save_all:
print ("Saving cycle" + .....)

economon January 29, 2013 15:37

Just to follow up on some of these issues, since we're particularly interested in getting the code to compile on clusters...

Which version of Python was available on the cluster natively before switching to version 3.2?

We have had success in the past with Python versions 2.6/2.7 and sometimes by installing a local copy of the free Enthought Python distribution (http://www.enthought.com/products/epd_free.php) that comes prepackaged with all of the modules that SU2 needs (NumPy, SciPy, etc.).

jentink January 29, 2013 15:44

Quote:

Originally Posted by economon (Post 404880)
Which version of Python was available on the cluster natively before switching to version 3.2?

We have had success in the past with Python versions 2.6/2.7 and sometimes by installing a local copy of the free Enthought Python distribution (http://www.enthought.com/products/epd_free.php) that comes prepackaged with all of the modules that SU2 needs (NumPy, SciPy, etc.).

default version on the cluster I'm using is 2.4.2. they had a 3.2.2 module available to load. that's the one I'm currently using.


All times are GMT -4. The time now is 12:42.