A parallel graph algorithms research project, covering GPU, multicore, and distributed memory implementations of graph coloring, partitioning, and traversal.
In particular, this codebase implements our Compressed Column Index (CCI) scheme (see technical paper or presentation slides), which for well-compressed matrices can beat CUSPARSE's tuned sparse matrix dense vector multiply.
To get our code, and run performance tests:
git clone http://projects.cs.uaf.edu/parallelgraph.git cd parallelgraph/src/test make make cuda
- Permanent Project: Yes
Main Git RepositoryThis URL has access.
Faculty: Orion Lawlor