Highly Parallel Sparse Matrix-Matrix Multiplication - Computer Science > Distributed, Parallel, and Cluster ComputingReport as inadecuate




Highly Parallel Sparse Matrix-Matrix Multiplication - Computer Science > Distributed, Parallel, and Cluster Computing - Download this document for free, or read online. Document in PDF available to download.

Abstract: Generalized sparse matrix-matrix multiplication is a key primitive for manyhigh performance graph algorithms as well as some linear solvers such asmultigrid. We present the first parallel algorithms that achieve increasingspeedups for an unbounded number of processors. Our algorithms are based ontwo-dimensional block distribution of sparse matrices where serial sections usea novel hypersparse kernel for scalability. We give a state-of-the-art MPIimplementation of one of our algorithms. Our experiments show scaling up tothousands of processors on a variety of test scenarios.



Author: Aydın Buluç, John R. Gilbert

Source: https://arxiv.org/







Related documents