GitHub - rsln-s/aggregative-coarsening-for-multilevel-hypergraph-partitioning: Implementation of the algorithm described in "Aggregative Coarsening for Multilevel Hypergraph Partitioning" by Shaydulin and Safro SEA 2018

rsln-s / aggregative-coarsening-for-multilevel-hypergraph-partitioning Public

Notifications You must be signed in to change notification settings
Fork 2
Star 6

Implementation of the algorithm described in "Aggregative Coarsening for Multilevel Hypergraph Partitioning" by Shaydulin and Safro SEA 2018

View license

6 stars 2 forks Branches Tags Activity

Star

Notifications

Branches Tags

Name		Name	Last commit message	Last commit date
Latest commit History 3 Commits
cmake		cmake
commonTools		commonTools
demos/buildAgainstTrilinos		demos/buildAgainstTrilinos
doc		doc
driver_test		driver_test
packages		packages
sampleScripts		sampleScripts
.gitdist.default		.gitdist.default
.gitignore		.gitignore
.mailmap		.mailmap
CMakeLists.txt		CMakeLists.txt
CTestConfig.cmake		CTestConfig.cmake
Copyright.txt		Copyright.txt
INSTALL		INSTALL
INSTALL.rst		INSTALL.rst
LICENSE		LICENSE
PackagesList.cmake		PackagesList.cmake
ProjectName.cmake		ProjectName.cmake
README		README
RELEASE_NOTES		RELEASE_NOTES
TPLsList.cmake		TPLsList.cmake
Version.cmake		Version.cmake
checkin-test.py		checkin-test.py
clone_extra_repos.py		clone_extra_repos.py
compile_commands.json		compile_commands.json
project-checkin-test-config.py		project-checkin-test-config.py
sparse_checkout.sh		sparse_checkout.sh

Repository files navigation

This repository contains the source code implementing the algorithm described in "Aggregative Coarsening for Multilevel Hypergraph Partitioning" by Shaydulin and Safro SEA 2018

====================

Install

====================

Dependencies:

Tested with:
- gcc 4.8.1
- openmpi 1.8.4
and
- intel 16.0 and newer

Installation:

cd algebraic-distance-on-hypergraphs
mkdir build
cd build
cmake -D CMAKE_INSTALL_PREFIX:FILEPATH="/home/rshaydu/algebraic-distance-on-hypergraphs/build" -D MPI_BASE_DIR="/software/openmpi/1.8.4_gcc" -D MPI_C_COMPILER:FILEPATH="/software/openmpi/1.8.4_gcc/bin/mpicc" -D MPI_CXX_COMPILER:FILEPATH="/software/openmpi/1.8.4_gcc/bin/mpic++" -D Trilinos_ENABLE_ALL_PACKAGES:BOOL=OFF -D Trilinos_ENABLE_Zoltan:BOOL=ON  -D Zoltan_ENABLE_EXAMPLES:BOOL=ON -D TPL_ENABLE_MPI:BOOL=ON -D Trilinos_ENABLE_Fortran:BOOL=OFF -D Zoltan_ENABLE_TESTS:BOOL=ON -D CMAKE_EXPORT_COMPILE_COMMANDS=ON -D COMM=mpi ..
make

==================

Notes on algorithms:

The paper describes two versions, inner-product coarsening and stable matching coarsening. By default, when run with PHG_COARSENING_METHOD=rs_alg_dist, the inner product coarsening is used. The code for stable matching is provided only for reproducibility purposes; stable matching based coarsening doesn't produce better results and is more computationally expensive (especially in this inefficient implementation). To use it, go to `packages/zoltan/src/phg/rs_amg_match.cpp` and uncomment `#define RS_STABLE_MATCHING`. 

Note that both algorithm only work in serial (when run with one process).

===================

In general, the user is advised to refer to Trilinos documentation for troubleshooting any problems during installation: http:https://trilinos.org/about/documentation/

If you have any questions, contanct me at [email protected], I'll be happy to help.

===================

For the full results, refer to http:https://shaydul.in/hypergraph-partitioning-archive/misc/aggregative-coarsening-for-multilevel-hypergraph-partitioning/