Scotch: a software package for graph and mesh/hypergraph partitioning, graph clustering, and sparse matrix ordering

The Scotch distribution is a set of programs and libraries which implement the static mapping and sparse matrix reordering algorithms developed within the Scotch project.

Scotch has many interesting features:

• Its capabilities can be used through a set of stand-alone programs as well as through the Scotch library, which offers both C and Fortran interfaces.

• It provides algorithms to partition graph structures, as well as mesh structures defined as node-element bipartite graphs and which can also represent hypergraphs.

• To speed-up its computations, the Scotch library dynamically takes advantage of either POSIX threads or native Windows threads. The PT-Scotch library, used to manage very large graphs distributed across the nodes of a parallel computer, uses the MPI interface, possibly in combination with multi-threading when the MPI implementation allows for it.

• It can map any weighted source graph onto any weighted target graph. The source and target graphs may have any topology, and their vertices and edges may be weighted. Moreover, both source and target graphs may be disconnected. This feature allows for the mapping of programs onto disconnected subparts of a parallel architecture made up of heterogeneous processors and communication links.

• It computes amalgamated block orderings of sparse matrices, for efficient solving using BLAS routines.

• Its running time is linear in the number of edges of the source graph, and logarithmic in the number of vertices of the target graph for mapping computations.

• It can handle indifferently graph and mesh data structures created within C or Fortran programs, with array indices starting from 0 or 1.

• It offers extended support for adaptive graphs and meshes through the handling of disjoint edge arrays.

• It is dynamically parametrizable thanks to strategy strings that are interpreted at run-time.

• It uses system memory efficiently, to process large graphs and meshes without incurring out-of-memory faults.

• It is highly modular and documented. Since it is available under a free/libre software license, it can be used as a testbed for the easy and quick development and testing of new partitioning and ordering methods.

• It can be easily interfaced to other programs. The programs comprising the Scotch project have been designed to run in command-line mode without any interactive prompting, so that they can be called easily from other programs by means of system() or popen() calls, or piped together on a single command line. Moreover, vertex labeling capabilities allow for easy renumbering of vertices.

• It provides many tools to build, check, and display graphs, meshes and matrix patterns.

• It is written in C and uses the POSIX interface, which makes it highly portable. Additionally, PT-Scotch uses the MPI interface.

Obtaining Scotch

Scotch is publicly available under the CeCILL-C free software license, as described here. The license itself is available here.

To use the lastest version of Scotch, please clone the master branch:

git clone [email protected]:scotch/scotch.git

Tarballs of the Scotch releases are available here.

Documentation

The most recent user and maintenance manuals are available here.

The principles and internals of Scotch are also described in several papers:

• Distillating knowledge about SCOTCH, in Combinatorial Scientific Computing, U. Naumann, O. Schenk, H. D. Simon and S. Toledo, eds., Dagstuhl Seminar Proceedings vol. 9061, Schloss Dagstuhl - Leibniz-Zentrum für Informatik, Dagstuhl, Germany, 2009, ISSN 1862-4405

• Scotch and PT-Scotch Graph Partitioning Software: An Overview, in Combinatorial Scientific Computing, U. Naumann, O. Schenk eds., Chapman and Hall/CRC, 2012, ISBN 978-0-429-15217-7

• Contributions to multilevel parallel graph partitioning, habilitation thesis, 2009

Installation

Two flavors of installation are available:

• With CMake, using version 3.10 or higher:

mkdir build && cd build && cmake .. && make -j5

Many options can be provided from the command line, using the CMmake flag -D.

Linux and MacOS-X are fully supported. MacOS-X users must use recent versions of Flex and Bison that are available from Brew; older versions from Xcode will fail. To use them, run, e.g.:

cmake -DBUILD_SHARED_LIBS=ON -DBISON_EXECUTABLE=/usr/local/Cellar/bison/3.8.2/bin/bison -DFLEX_EXECUTABLE=/usr/local/Cellar/flex/2.6.4_2/bin/flex

You may also use alternate compilers, by overloading the CMAKE_*_COMPILER variables, e.g.:

cmake -DCMAKE_Fortran_COMPILER=ifx -DCMAKE_C_COMPILER=icx -DMPI_HOME=/path/to/oneAPI/mpi/latest/

Windows plaftorms are also supported, featuring native multi-threading. Processor affinity is not yet implemented. If a Unix-like version of Make is not available, CMake can generate Microsoft NMAKE Makefiles:

cmake -G "NMake Makefiles" -DCMAKE_BUILD_TYPE=Release ..
nmake

When creating dynamic libraries, the -DLIBSCOTCHERR flag allows one to select at compile time which error library to link against. This flag is set to "" by default, but can be set to any predefined or user-defined error handling library, e.g.,

cmake -DLIBSCOTCHERR=scotcherr

• With a traditional Makefile:

CMake installation is easy and straightforward. It allows one to compile and install Scotch and PT-Scotch, depending on flags such as the use of multi-threading and/or MPI. The traditional Makefile installation gives additional freedom to perform (cross-)compilation for non-standard systems and configurations.

Contributing to Scotch

To report a bug or discuss potential improvements, You can contact directly the PI at [email protected]. However, the GitLab environment provides features that are worth taking advantage of, so we recommend you to take the time to use them. Before reporting a bug or submitting a patch in the Inria GitLab environment, you will need an account on the server. Please dot not hesitate to send an e-mail to [email protected] so that we create an account for you on the Inria Gitlab repository. You will then be able to open issues in the bug tracker, request features, or propose patches using the "merge requests" feature.

Past and current contributors

The following people contribute(d) to the development of Scotch:

• Clément BARTHÉLEMY

• Cédric CHEVALIER

• Sébastien FOURESTIER

• Marc FUENTES

• Jun-Ho HER

• Amaury JACQUES

• Cédric LACHAT

• Selmane LEBDAOUI

• Tetsuya MISHIMA

• Xavier MULLER

• François PELLEGRINI (PI)

• Florent PRUVOST

• Luca SCARANO

Citing Scotch

Feel free to use the following publications to reference Scotch:

• "Scotch and PT-Scotch Graph Partitioning Software: An Overview" https://hal.inria.fr/hal-00770422

• "PT-Scotch: A tool for efficient parallel graph ordering" https://hal.inria.fr/hal-00402893

Licence

https://gitlab.inria.fr/scotch/scotch/blob/master/LICENSE_en.txt