This page provides instructions on how to build Xolotl and its dependencies. CMake is required to build Xolotl, although instructions for it are not included below. We recommend that you download CMake using the package manager on your system. On Red Hat-based systems, this is as simple as running "yum install cmake" as root.
Xolotl also requires GCC 4.7.2 or greater. Instructions to upgrade your GCC are given at the bottom of this page.
All of the commands are executed in bash unless stated otherwise. You should replace any reference to /home/user to the appropriate prefix for your system. All of the packages are also installed to /opt by preference, but you may adjust that prefix as you see fit too.
First things first, let's checkout the SVN repository with
svn co https://svn.code.sf.net/p/xolotl-psi/code/trunk /home/user/xolotl-psi-code
Now, assuming you have all of the dependencies for Xolotl built and a good C++11 compiler (GCC 4.7.2 or greater), Xolotl itself is quite easy to build. If you do not have all of the dependencies, check out the next section.
Create a directory in which you want to build Xolotl (for instance /home/user/xolotl-psi-code/build) and change into it. It can be any directory except for the source directory (which is /home/user/xolotl-psi-code/xolotl). From inside the build directory run the following commands:
CXX=/opt/mpich-3.0rc1_gcc-4.7.2/bin/mpicxx PETSC_DIR=/opt/petsc-latest_mpich-3.0rc1 cmake -DCMAKE_BUILD_TYPE=Release ../xolotl
with the path to your MPI compiler (here mpicxx) in CXX and the path to an installed PETSc version in PETSC_DIR (multiple PETSc versions can coexist on the same file-system).
You can also run make in parallel (make -jN, where N is the number of processes) to build Xolotl faster. This is recommended if you have a big machine.
If CMake fails to find HDF5, add the path to your HDF5 installation directory in the CMake command:
CXX=/opt/mpich-3.0rc1_gcc-4.7.2/bin/mpicxx PETSC_DIR=/opt/petsc-latest_mpich-3.0rc1 HDF5_ROOT=/opt/hdf5-1.8.10_mpich-3.0rc1_gcc-4.7.2 cmake -DCMAKE_BUILD_TYPE=Release ../xolotl
Never use the -DCMAKE_CXX_COMPILER=<C++ compiler=""> option instead of CXX=<C++ compiler=""> environment variable. It causes CMake to ignore all of your other options (such as -DBoost_ROOT if you have a special build of Boost).
The following codes must be built and configured as described below in order for Xolotl to compile.
Xolotl uses Boost for unit tests. If you don't want to run the tests, then you don't need Boost. However, if you do want to run the tests you should have the latest version of Boost installed for your system. You can get Boost from your package manager or build it from scratch by downloading the latest. Use the following commands in the Boost directory:
./b2 install --prefix=/opt/boost-1.57_gcc-4.7.2 --with-test
This will only install the parts of Boost required for testing.
If Boost is installed to a non-standard prefix such as the above, you can point CMake to it by using the -DBOOST_ROOT option.
You can run Xolotl tests with
If you want to only run a specific one, run the corresponding executable that is created in your build folder. For instance (from your build directory)
MPICH 3.0 is required for Xolotl and its dependencies. Use the following commands:
./configure --prefix=/opt/mpich-3.0rc1_gcc-4.7.2 --enable-shared=yes 2>&1 | tee c.txt
make 2>&1 | tee m.txt
make install 2>&1 | tee mi.txt
Your path needs to be updated to point to the new MPICH3.0 install, which can be done by adding the following lines in either ~/.bashrc or ~/.bash_profile:
Your library path also needs to be updated in the same file:
PETSc is used to solve the reaction-diffusion equation in Xolotl. We build PETSc without debugging enabled and let it download Hypre, BLAS and LAPACK. Xolotl now uses the "master" version of PETSc. You can download it (for instance in /home/user) using git with the following command:
git clone https://bitbucket.org/petsc/petsc petsc
Enter the petsc directory and build it using the following commands (hint: after the configure step, PETSc will print you what to execute):
./configure PETSC_DIR=$PWD --prefix=/opt/petsc-latest_mpich-3.0rc1 --with-cc=/opt/mpich-3.0rc1_gcc-4.7.2/bin/mpicc --with-cxx=/opt/mpich-3.0rc1_gcc-4.7.2/bin/mpicxx --with-fc=mpif77 --with-debugging=no --download-fblaslapack=1 --FOPTFLAGS= --with-shared-libraries=1 --download-hypre=yes --download-superlu_dist --download-parmetis --download-metis
make PETSC_DIR=/home/user/petsc PETSC_ARCH=arch-linux2-c-opt all
make PETSC_DIR=/home/user/petsc PETSC_ARCH=arch-linux2-c-opt install
make PETSC_DIR=/opt/petsc-latest_mpich-3.0rc1 test
make PETSC_DIR=/opt/petsc-latest_mpich-3.0rc1 PETSC_ARCH=arch-linux2-c-opt streams NPMAX=<number of MPI processes you intend to use>
HDF5 is required for I/O in Xolotl and its dependencies. Use the following commands:
./configure --prefix=/opt/hdf5-1.8.10_mpich-3.0rc1_gcc-4.7.2 --enable-parallel
Your path needs to be updated to point to the new HDF5 install, which can be done by adding the following lines in either ~/.bashrc or ~/.bash_profile:
Your library path also needs to be updated in the same file:
The following are the libraries required IF AND ONLY IF you wish to monitor Xolotl's performance. For more information about the performance infrastructure of Xolotl please refer to the [Performance] section of the wiki.
PAPI, Performance Application Programming Interface, provides access to hardware counter information. In Xolotl, this library is used via GPTL to access hardware performance counters.
FIRST: Please refer to the INSTALL.txt file in the root directory to find the installation instructions for your specific operating system.
The general installation instructions are as follows (configure and make files are in the src directory):
make fulltest (optional step to run all tests)
Your path may need to be updated to set the variable PAPI_PREFIX to point to the root installation of the PAPI library and include files, which can be done by adding the following lines in either ~/.bashrc or ~/.bash_profile:
Note that versions other than PAPI 5.3.0 have been known to cause problems, so it is advised to use this version.
The General Purpose Timing Library, or GPTL, is a lightweight timing library with support for timing arbitrary regions of code, MPI operations, and collecting hardware performance counter information.
See the GPTL web page and INSTALL file in the root of the GPTL source distribution for instructions on building, testing, and installing GPTL.
To be used for building Xolotl, GPTL's code must be compiled so that the resulting static library can be linked into other shared libraries (such as Xolotl's performance library). For many systems, this means compiling the library using a flag like '-fPIC'. To configure GPTL to be built this way, add the -fPIC flag to the first definitions of CFLAGS and FFLAGS within the GPTL macros.make configuration file before building.
If you are NOT using version 5.3 of GPTL two other modifications to GPTL might be required, depending on the distribution and version of MPI you are using. GPTL supports MPICH and its variants out of the box, but requires a slight code modification for use with OpenMPI. If you are using OpenMPI, add -DMPI_STATUS_SIZE=6 to the definition of CFLAGS in the macros.make file. Also, if you are using a newer release of OpenMPI, you may need to modify the function signatures in pmpi.c so that read-only arguments are declared with 'const'. For instance, with OpenMPI 1.7.4, the signature for MPI_Send is
int MPI_Send( const void* buf, int count, MPI_Datatype datatype, int dest, int tag, MPI_Comm comm );
and GPTL's MPI wrappers will not compile without the const qualifier. Workarounds are to either add the missing const keywords in the GPTL pmpi.c file, or use an older version of OpenMPI such as version 1.6.4 that does not declare MPI functions using const qualifiers.
NOTE: Refer to the Configure Requirements section of [Performance] for further instructions on building Xolotl with GPTL (if GPTL is being built with OPENMP=yes in macros.make).
After installing GPTL, your path needs to be updated to set the variable GPTL_PREFIX to point to the new root installation of the GPTL library and include files, which can be done by adding the following lines in either ~/.bashrc or ~/.bash_profile:
The following libraries are needed in order to plot Xolotl's outputs. A wiki page on the visualization will be added soon.
Mesa is an open-source implementation of the OpenGL specification - a system for rendering interactive 3D graphics. A light version of Mesa-7.5.2 is available here. Once downloaded, you just have to untar and make it in the repository where you want it (for instance /opt)
tar -xf MesaLib-7.5.2-OSMesa.tar.bz2
After that you simply have to add MESA_PREFIX to your ~/.bashrc or ~/.bash_profile
git clone https://github.com/jsmeredith/EAVL EAVL
(note: an update this week to EAVL has yet to be reflected in Xolotl's code; this will happen shortly, but in the meantime, after cloning the repository, please revert to an older version with:
git checkout 263414fc4aa416e2376c1af9150c6934146644eb)
to install it and then add EAVL_PREFIX to your ~/.bashrc or ~/.bash_profile (if the repository was cloned to your /opt directory)
You should now be able to run Xolotl and its preprocessor. Check out our page on [Running Xolotl].
The default compiler on Red Hat Enterprise Linux (RHEL) 6 is GCC 4.4.7, but Xolotl requires GCC 4.7.2 or greater. RHEL supports GCC 4.7.2 if you have the "Linux Developer" version and you can install it following the instructions on the Developer Toolset page.
Most RHEL 6 customers do not have access to the "Linux Developer" version of RHEL 6. The Developer Toolset is open source and can be installed through the CentOS repositories. First, as root, create a file in /etc/yum.repos.d/ called testing-devtools.repo and add the following contents
name=testing devtools for CentOS $releasever
Next, also as root, install the Developer Toolset version 1.1 with yum by issuing the following command:
yum --enablerepo=testing-devtools install devtoolset-1.1
Answer yes at the prompts to install the toolset.
The Developer Toolset and GCC 4.7.2 will be installed to /opt/centos/devtoolset-1.1 and the executables will be available in /opt/centos/devtoolset-1.1/root/usr/bin. This installation will not automatically override the default install of GCC 4.4.7. It will still be on your path and take precedence over 4.7.2, so configure your path or direct CMake to build Xolotl accordingly.
The simplest way to use the new compiler is to rebuild MPI with GCC 4.7.2 (which you kind of have to do anyway) and then rebuild Xolotl with MPI.
The default version of Boost on RHEL6 will not work with GCC 4.7.2 because of changes in both packages for the C++11 standard implementation. Download the latest version of Boost from the website and build it with the instructions above. Only the testing libraries are required.
If you build Boost this way, then you will need to add the -DBoost_ROOT=
CXX=mpicxx cmake ../xolotl -DCMAKE_BUILD_TYPE=Release -DBoost_ROOT=/opt/boost -DBoost_NO_BOOST_CMAKE=ON
If your machine is Ubuntu 12.10 you're already using GNU C++ 4.7 and above.
Here is a link that will tell you how to upgrade GCC if you're on Ubuntu 12.04.