Segmented Merge: A New Primitive for Parallel Sparse Matrix Computations - Network and Parallel Computing Access content directly
Conference Papers Year : 2021

Segmented Merge: A New Primitive for Parallel Sparse Matrix Computations

Haonan Ji
  • Function : Author
  • PersonId : 1161080
Shibo Lu
  • Function : Author
  • PersonId : 1161081
Kaixi Hou
  • Function : Author
  • PersonId : 1161082
Weifeng Liu
  • Function : Author
  • PersonId : 1160995
Brian Vinter
  • Function : Author
  • PersonId : 1161084

Abstract

Segmented operations, such as segmented sum, segmented scan and segmented sort, are important building blocks for parallel irregular algorithms. We in this work propose a new parallel primitive called segmented merge. Its function is in parallel merging q sub-segments to p segments, both of nonuniform lengths. We implement the segmented merge primitive on GPUs and demonstrate its efficiency on parallel sparse matrix transposition (SpTRANS) and sparse matrix-matrix multiplication (SpGEMM) operations.
Fichier principal
Vignette du fichier
511910_1_En_15_Chapter.pdf (2.27 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

hal-03768762 , version 1 (04-09-2022)

Licence

Attribution

Identifiers

Cite

Haonan Ji, Shibo Lu, Kaixi Hou, Hao Wang, Weifeng Liu, et al.. Segmented Merge: A New Primitive for Parallel Sparse Matrix Computations. 17th IFIP International Conference on Network and Parallel Computing (NPC), Sep 2020, Zhengzhou, China. pp.170-181, ⟨10.1007/978-3-030-79478-1_15⟩. ⟨hal-03768762⟩
24 View
9 Download

Altmetric

Share

Gmail Facebook X LinkedIn More