Parallel sparse matrix-vector and matrix-transpose-vector multiplication using compressed sparse blocks

By Aydin Buluç, Jeremy T. Fineman, Matteo Frigo, John R. Gilbert, Charles E. Leiserson. 2009

457 citations recorded by the source corpus.

Subfields: Computer Networks and Communications, Hardware and Architecture

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Computer Networks and Communications

1. Aydin Buluç
2. John R. Gilbert
3. Pradeep K. Dubey
4. Jongsoo Park
5. Jimeng Sun
6. Brian Vinter
7. R. Vuduc
8. Hai Helen Li

In Hardware and Architecture

1. Aydin Buluç
2. John R. Gilbert
3. Pradeep K. Dubey
4. Jongsoo Park
5. Jimeng Sun
6. Brian Vinter
7. R. Vuduc
8. Jiajia Li

Citations in the corpus

• Sparse GPU Kernels for Deep Learning (2020, neutral): “transposed and non-transposed computation [53], [54]”
• Faster CNNs with Direct Sparse Convolutions and Guided Pruning (2016, methodology): “Our optimized implementation 1 fully takes advantage of the reuse, applying loop tiling to both input and output channels, with column blocking (Buluç et al., 2009) applied to the weight sparse matrix.”
• HiCOO: Hierarchical Storage of Sparse Tensors (2018, methodology): “4rn CSB, block sizes are typically around VI x VI for an I x I sparse matrix [33].”
• The Combinatorial BLAS: design, implementation, and applications (2011, background): “Almost all popular sparse matrix storage formats are internally composed of a number of arrays (Dongarra 2000; Saad 2003; Buluç et al. 2009), since arrays are cache friendlier than pointer-”
• CSR5: An Efficient Storage Format for Cross-Platform Sparse Matrix-Vector Multiplication (2015, methodology): “But to accelerate large-scale computation, parallel SpMV is still required to be hand-optimized with specific data storage formats and algorithms [1, 2, 3, 4, 6, 7, 10, 13, 16, 17, 19, 21, 22, 24, 29, 30, 31, 33, 34].”