We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

cs.DC

Change to browse by:

References & Citations

DBLP - CS Bibliography

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Computer Science > Distributed, Parallel, and Cluster Computing

Title: Sphynx: a parallel multi-GPU graph partitioner for distributed-memory systems

Abstract: Graph partitioning has been an important tool to partition the work among several processors to minimize the communication cost and balance the workload. While accelerator-based supercomputers are emerging to be the standard, the use of graph partitioning becomes even more important as applications are rapidly moving to these architectures. However, there is no distributed-memory parallel, multi-GPU graph partitioner available for applications. We developed a spectral graph partitioner, Sphynx, using the portable, accelerator-friendly stack of the Trilinos framework. In Sphynx, we allow using different preconditioners and exploit their unique advantages. We use Sphynx to systematically evaluate the various algorithmic choices in spectral partitioning with a focus on the GPU performance. We perform those evaluations on two distinct classes of graphs: regular (such as meshes, matrices from finite element methods) and irregular (such as social networks and web graphs), and show that different settings and preconditioners are needed for these graph classes. The experimental results on the Summit supercomputer show that Sphynx is the fastest alternative on irregular graphs in an application-friendly setting and obtains a partitioning quality close to ParMETIS on regular graphs. When compared to nvGRAPH on a single GPU, Sphynx is faster and obtains better balance and better quality partitions. Sphynx provides a good and robust partitioning method across a wide range of graphs for applications looking for a GPU-based partitioner.
Comments: To appear in Parallel Computing
Subjects: Distributed, Parallel, and Cluster Computing (cs.DC); Discrete Mathematics (cs.DM); Mathematical Software (cs.MS)
MSC classes: 68W10
Report number: SAND2021-0352-O
Cite as: arXiv:2105.00578 [cs.DC]
  (or arXiv:2105.00578v1 [cs.DC] for this version)

Submission history

From: Erik Boman [view email]
[v1] Sun, 2 May 2021 23:47:28 GMT (769kb,D)

Link back to: arXiv, form interface, contact.