What question did this study set out to answer?

To develop an efficient sparse-matrix representation and GPU parallelization approach specifically optimized for sparse matrix-multivector product computations.

June 11, 2018

Efficient sparse-matrix multi-vector product on GPUs

View Full Paper

Ask AI

Bookmark

To develop an efficient sparse-matrix representation and GPU parallelization approach specifically optimized for sparse matrix-multivector product computations.
Contrasted computational profiles and memory access patterns between sparse matrix-vector and sparse matrix-multivector operations.
Formulated a specialized matrix representation and parallel execution scheme to minimize data-movement overhead on GPUs.
Benchmarked performance against standard vendor libraries across all test matrices in the SuiteSparse matrix collection.
Demonstrated marked execution speedups over existing vendor GPU implementations across diverse matrix topologies.
Achieved higher data-movement efficiency and improved GPU thread utilization across the entire SuiteSparse benchmark suite.

Discussion

No takes yet. Share an insight, caveat, or question.

Implication

Experimental study demonstrates superior data movement and throughput for sparse matrix-multivector multiplication on graphics processors, highlighting viable alternatives to vendor libraries.

Cite This Study

Hong et al. (2018) studied this question.

synapsesocial.com/papers/6a192d10427684435388b8a2 https://doi.org/10.1145/3208040.3208062

View Full Paper

Ask AI

Bookmark

Implication

Experimental study demonstrates superior data movement and throughput for sparse matrix-multivector multiplication on graphics processors, highlighting viable alternatives to vendor libraries.

Efficient sparse-matrix multi-vector product on GPUs

Key Points

Discussion

Implication

Cite This Study

Efficient sparse-matrix multi-vector product on GPUs

Key Points

Discussion

Implication

Cite This Study

Also Consider

Also Consider