Publication:
Adaptive level binning: a new algorithm for solving sparse triangular systems

dc.contributor.departmentDepartment of Computer Engineering
dc.contributor.departmentDepartment of Computer Engineering
dc.contributor.departmentN/A
dc.contributor.departmentDepartment of Computer Engineering
dc.contributor.kuauthorErten, Didem Unat
dc.contributor.kuauthorYılmaz, Buse
dc.contributor.kuauthorAhmad, Najeeb
dc.contributor.kuauthorSipahioğlu, Buğra
dc.contributor.kuprofileFaculty Member
dc.contributor.kuprofileResearcher
dc.contributor.kuprofilePhD Student
dc.contributor.kuprofileUndergraduate Student
dc.contributor.otherDepartment of Computer Engineering
dc.contributor.schoolcollegeinstituteCollege of Engineering
dc.contributor.schoolcollegeinstituteCollege of Engineering
dc.contributor.schoolcollegeinstituteGraduate School of Sciences and Engineering
dc.contributor.schoolcollegeinstituteCollege of Engineering
dc.contributor.yokid219274
dc.contributor.yokidN/A
dc.contributor.yokidN/A
dc.contributor.yokidN/A
dc.date.accessioned2024-11-09T23:29:37Z
dc.date.issued2020
dc.description.abstractSparse triangular solve (SpTRSV) is an important scientific kernel used in several applications such as preconditioners for Krylov methods. Parallelizing SpTRSV on multi-core systems is challenging since it exhibits limited parallelism due to computational dependencies and introduces high parallelization overhead due to finegrained and unbalanced nature of workloads. We propose a novel method, named Adaptive Level Binning (ALB), that addresses these challenges by eliminating redundant synchronization points and adapting the work granularity with an efficient load balancing strategy. Similar to the commonly used level-set methods for solving SpTRSV, ALB constructs level-sets of rows, where each level can be computed in parallel. Differently, ALB bins rows to levels adaptively and reduces redundant dependencies between rows. On an Intel® Xeon® Gold 6148 processor and NVIDIA® Tesla V100 GPU, ALB obtains 1.83x speedup on average and up to 5.28x speedup over Intel MKL and, over NVIDIA cuSPARSE, an average speedup of 2.80x and a maximum speedup of 39.40x for 29 matrices selected from Suite Sparse Matrix Collection.
dc.description.indexedbyWoS
dc.description.indexedbyScopus
dc.description.openaccessYES
dc.description.publisherscopeInternational
dc.identifier.doi10.1145/3368474.3368486
dc.identifier.isbn9781-4503-7236-7
dc.identifier.linkhttps://www.scopus.com/inward/record.uri?eid=2-s2.0-85094848543&doi=10.1145%2f3368474.3368486&partnerID=40&md5=c5f36ab61cae64633ae5011f71db70b1
dc.identifier.scopus2-s2.0-85094848543
dc.identifier.urihttps://dx.doi.org/10.1145/3368474.3368486
dc.identifier.urihttps://hdl.handle.net/20.500.14288/12075
dc.keywordsCPU
dc.keywordsFine-grained parallelism
dc.keywordsLevel-set
dc.keywordsSparse triangular solvers Matrix algebra
dc.keywordsAlgorithm for solving
dc.keywordsLevel Set method
dc.keywordsLimited parallelism
dc.keywordsLoad balancing strategy
dc.keywordsMulti-core systems
dc.keywordsParallelizations
dc.keywordsSynchronization points
dc.keywordsTriangular system
dc.keywordsNumerical methods
dc.languageEnglish
dc.publisherInformation Processing Society of Japan (IPSJ)
dc.sourceACM International Conference Proceeding Series
dc.subjectComputer science
dc.subjectInformation resources management
dc.subjectSoftware engineering
dc.titleAdaptive level binning: a new algorithm for solving sparse triangular systems
dc.typeConference proceeding
dspace.entity.typePublication
local.contributor.authorid0000-0002-2351-0770
local.contributor.authoridN/A
local.contributor.authorid0000-0002-3460-1256
local.contributor.authoridN/A
local.contributor.kuauthorErten, Didem Unat
local.contributor.kuauthorYılmaz, Buse
local.contributor.kuauthorAhmad, Najeeb
local.contributor.kuauthorSipahioğlu, Buğra
relation.isOrgUnitOfPublication89352e43-bf09-4ef4-82f6-6f9d0174ebae
relation.isOrgUnitOfPublication.latestForDiscovery89352e43-bf09-4ef4-82f6-6f9d0174ebae

Files