WORK IN PROGRESS
Navigate: Overview | Configuration →
Download, build and run Musubi.
We assume that you are using a UNIX-like system. If you run Windows, some commands might be different.
We use Mercurial (hg) for revision control. You need to have it
installed on your system in order to download Musubi.
Follow these instructions if you haven't, yet.
Then, create a directory for everything that happens in these tutorials
(we will call this directory apes, but you can use a different name.
Inside this directory, clone the Musubi repository from
https://geb.inf.tu-dresden.de/hg/musubi by running
hg clone https://geb.inf.tu-dresden.de/hg/musubi
in your console. If this worked, you have an up-to-date copy of the Musubi source code, which we will compile now.
If you already cloned Musubi you can update Musubi via:
hg pull
hg up
We use the waf build system, you can learn more about it
from its website.
Also, you need MPI installed on your system, see for example the
OpenMPI website for instructions.
Finally, you need to set environment variables FC and CC in order to
assign the correct Fortran and C compilers to waf.
The compilers should point to the MPI wrappers from your MPI installation.
Typically these are mpif90 (Fortran) and mpicc (C).
Set them with the bash commands
export CC=mpicc; export FC=mpif90
If you updated Musubi you need yo clean the old build and the coco preprocessor. If you cloned Musubi there is no need to do so.
./waf cleanall
Once you have done all this, navigate to your Musubi directory and use the command
./waf configure
to configure the compilation.
We are now ready to compile Musubi. Run
./waf build
to get a Musubi executable including Mus_Harvester for post-processing in the build subdirectory. To compile musubi only add the following argument:
--target="musubi"
If the compilation finishes without errors, you have Musubi ready to run your first test case!
To check your Musubi-installation, navigate to your musubi directory and
create a required tracking directory for the output with
mkdir tracking
Execute Musubi with
./build/musubi
which should result in loads of output, ending with a message similar to
Done with Musubi in [s] 1.864378E-01
indicating that everything worked fine. If you get any errors up to here, read the instructions again and follow them carefully. You need to have this running before you proceed.
Most of the tutorials will require creation of a mesh and some
post-processing of the results. The corresponding tools in our toolchain
"APES tool chain" are Seeder for mesh-generation, for post-processing
Seeder-Harvesting and Musubi-Harvesting. The installation procedure for
Seeder is very similar to Musubi. Again, navigate to your apes directory
and run
hg clone https://geb.inf.tu-dresden.de/hg/seeder
to get a fresh copy of Seeder. Compile it by running
cd seeder
./waf configure build
and fix any errors before you proceed.
You can make your life easier by adding apes/seeder/build,
apes/seeder/build/sdr_harvesting, apes/musubi/build and
apes/musubi/build/mus_harvesting to your path,
for example by editing ~/.profile (MacOS X) or ~/.bashrc (Unix with
bash) or whatever it is on your system.
In the following tutorials, we assume that you have done just that.
If you have not, you must add the correct paths to the command any time
you try to call Musubi, Musubi-Harvesting, Seeder-Harvesting or Seeder.
Once you are done with all that, we can start defining our first simulation.
Once you get errors running Musubi, it is possible that something is wrong
with your code version. In order to get more detailed information concerning the
errors you can type this command inside the musubi directory which is
apes/musubi/ as default:
./waf distclean configure debug
If you run your simulation once again, you will get more information about the files that cause errors.
Now, you have to run Musubi from a different directory which is
/build/debug/musubiinstead of/build/musubi.
Next chapter: Configuration →