Efficient and Effective Sparse Tensor Reordering
Abstract
This paper proposed two reordering schemes for sparse tensors: BFSMCS and LEXIORDER. BFSMCS is a Breadth First Search (BFS)like heuristic approach based on the maximum cardinality search family; LEXIORDER is an extension of doubly lexical ordering of matrices to tensors. CANDECOMP/PARAFAC decomposition (CPD) is taken as an example to show their effect on existing three sparse tensor formats for CPUs: coordinate (COO), compressed sparse fiber (CSF), and hierarchical coordinate (HICOO). LEXIORDER obtains up to 4.14× speedup on sequential HICOOMTTKRP and 11.88× speedup on its parallel case. COO and CSFMTTKRP also achieves performance improvement in different degree. Our two reordering methods are more efficient and effective than other stateoftheart reordering methods used in SPLATT.
 Authors:

 BATTELLE (PACIFIC NW LAB)
 UMR5668, CNRS and ENS Lyon, France
 Georgia Institute of Technology
 Publication Date:
 Research Org.:
 Pacific Northwest National Lab. (PNNL), Richland, WA (United States)
 Sponsoring Org.:
 USDOE
 OSTI Identifier:
 1574893
 Report Number(s):
 PNNLSA138751
 DOE Contract Number:
 AC0576RL01830
 Resource Type:
 Conference
 Resource Relation:
 Conference: Proceedings of the ACM International Conference on Supercomputing (ICS 2019), June 2628, 2019, Phoenix, AZ
 Country of Publication:
 United States
 Language:
 English
 Subject:
 HPC
Citation Formats
Li, Jiajia, Ucar, Bora, Catalyurek, Umit, Sun, Jimeng, Barker, Kevin J., and Vuduc, Richard. Efficient and Effective Sparse Tensor Reordering. United States: N. p., 2019.
Web. doi:10.1145/3330345.3330366.
Li, Jiajia, Ucar, Bora, Catalyurek, Umit, Sun, Jimeng, Barker, Kevin J., & Vuduc, Richard. Efficient and Effective Sparse Tensor Reordering. United States. doi:10.1145/3330345.3330366.
Li, Jiajia, Ucar, Bora, Catalyurek, Umit, Sun, Jimeng, Barker, Kevin J., and Vuduc, Richard. Wed .
"Efficient and Effective Sparse Tensor Reordering". United States. doi:10.1145/3330345.3330366.
@article{osti_1574893,
title = {Efficient and Effective Sparse Tensor Reordering},
author = {Li, Jiajia and Ucar, Bora and Catalyurek, Umit and Sun, Jimeng and Barker, Kevin J. and Vuduc, Richard},
abstractNote = {This paper proposed two reordering schemes for sparse tensors: BFSMCS and LEXIORDER. BFSMCS is a Breadth First Search (BFS)like heuristic approach based on the maximum cardinality search family; LEXIORDER is an extension of doubly lexical ordering of matrices to tensors. CANDECOMP/PARAFAC decomposition (CPD) is taken as an example to show their effect on existing three sparse tensor formats for CPUs: coordinate (COO), compressed sparse fiber (CSF), and hierarchical coordinate (HICOO). LEXIORDER obtains up to 4.14× speedup on sequential HICOOMTTKRP and 11.88× speedup on its parallel case. COO and CSFMTTKRP also achieves performance improvement in different degree. Our two reordering methods are more efficient and effective than other stateoftheart reordering methods used in SPLATT.},
doi = {10.1145/3330345.3330366},
journal = {},
number = ,
volume = ,
place = {United States},
year = {2019},
month = {6}
}