Dissertation/ Thesis

Cache locality exploiting methods and models for sparse matrix-vector multiplication

Λεπτομέρειες βιβλιογραφικής εγγραφής
Τίτλος: Cache locality exploiting methods and models for sparse matrix-vector multiplication
Συγγραφείς: Akbudak, Kadir
Έτος έκδοσης: 2009
Συλλογή: Bilkent University: Institutional Repository
Θεματικοί όροι: Cache locality, Sparse matrices, Matrix-vector multiplication, Matrix reordering, Computational hypergraph model, Hypergraph partitioning, Traveling salesman problem, QA188 .A53 2009, Sparse matrices--Data processing, Cache memory, Hypergraphs
Περιγραφή: Cataloged from PDF version of article. ; Includes bibliographical references leaves 52-56. ; The sparse matrix-vector multiplication (SpMxV) is an important kernel operation widely used in linear solvers. The same sparse matrix is multiplied by a dense vector repeatedly in these solvers to solve a system of linear equations. High performance gains can be obtained if we can take the advantage of today’s deep cache hierarchy in SpMxV operations. Matrices with irregular sparsity patterns make it difficult to utilize data locality effectively in SpMxV computations. Different techniques are proposed in the literature to utilize cache hierarchy effectively via exploiting data locality during SpMxV. In this work, we investigate two distinct frameworks for cacheaware/oblivious SpMxV: single matrix-vector multiply and multiple submatrix-vector multiplies. For the single matrix-vector multiply framework, we propose a cache-size aware top-down row/column-reordering approach based on 1D sparse matrix partitioning by utilizing the recently proposed appropriate hypergraph models of sparse matrices, and a cache oblivious bottom-up approach based on hierarchical clustering of rows/columns with similar sparsity patterns. We also propose a column compression scheme as a preprocessing step which makes these two approaches cache-line-size aware. The multiple submatrix-vector multiplies framework depends on the partitioning the matrix into multiple nonzero-disjoint submatrices. For an effective matrixto-submatrix partitioning required in this framework, we propose a cache-size aware top-down approach based on 2D sparse matrix partitioning by utilizing the recently proposed fine-grain hypergraph model. For this framework, we also propose a traveling salesman formulation for an effective ordering of individual submatrix-vector multiply operations. We evaluate the validity of our models and methods on a wide range of sparse matrices. Experimental results show that proposed methods and models outperforms state-of-the-art schemes. ...
Τύπος εγγράφου: thesis
Περιγραφή αρχείου: xiv, 56 leaves; application/pdf
Γλώσσα: English
Relation: https://hdl.handle.net/11693/15336; B118111
Διαθεσιμότητα: https://hdl.handle.net/11693/15336
Rights: info:eu-repo/semantics/openAccess
Αριθμός Καταχώρησης: edsbas.39D0640C
Βάση Δεδομένων: BASE
FullText Text:
  Availability: 0
CustomLinks:
  – Url: https://hdl.handle.net/11693/15336#
    Name: EDS - BASE (ns324271)
    Category: fullText
    Text: View record from BASE
Header DbId: edsbas
DbLabel: BASE
An: edsbas.39D0640C
RelevancyScore: 762
AccessLevel: 3
PubType: Dissertation/ Thesis
PubTypeId: dissertation
PreciseRelevancyScore: 761.994323730469
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: Cache locality exploiting methods and models for sparse matrix-vector multiplication
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Akbudak%2C+Kadir%22">Akbudak, Kadir</searchLink>
– Name: DatePubCY
  Label: Publication Year
  Group: Date
  Data: 2009
– Name: Subset
  Label: Collection
  Group: HoldingsInfo
  Data: Bilkent University: Institutional Repository
– Name: Subject
  Label: Subject Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22Cache+locality%22">Cache locality</searchLink><br /><searchLink fieldCode="DE" term="%22Sparse+matrices%22">Sparse matrices</searchLink><br /><searchLink fieldCode="DE" term="%22Matrix-vector+multiplication%22">Matrix-vector multiplication</searchLink><br /><searchLink fieldCode="DE" term="%22Matrix+reordering%22">Matrix reordering</searchLink><br /><searchLink fieldCode="DE" term="%22Computational+hypergraph+model%22">Computational hypergraph model</searchLink><br /><searchLink fieldCode="DE" term="%22Hypergraph+partitioning%22">Hypergraph partitioning</searchLink><br /><searchLink fieldCode="DE" term="%22Traveling+salesman+problem%22">Traveling salesman problem</searchLink><br /><searchLink fieldCode="DE" term="%22QA188+%2EA53+2009%22">QA188 .A53 2009</searchLink><br /><searchLink fieldCode="DE" term="%22Sparse+matrices--Data+processing%22">Sparse matrices--Data processing</searchLink><br /><searchLink fieldCode="DE" term="%22Cache+memory%22">Cache memory</searchLink><br /><searchLink fieldCode="DE" term="%22Hypergraphs%22">Hypergraphs</searchLink>
– Name: Abstract
  Label: Description
  Group: Ab
  Data: Cataloged from PDF version of article. ; Includes bibliographical references leaves 52-56. ; The sparse matrix-vector multiplication (SpMxV) is an important kernel operation widely used in linear solvers. The same sparse matrix is multiplied by a dense vector repeatedly in these solvers to solve a system of linear equations. High performance gains can be obtained if we can take the advantage of today’s deep cache hierarchy in SpMxV operations. Matrices with irregular sparsity patterns make it difficult to utilize data locality effectively in SpMxV computations. Different techniques are proposed in the literature to utilize cache hierarchy effectively via exploiting data locality during SpMxV. In this work, we investigate two distinct frameworks for cacheaware/oblivious SpMxV: single matrix-vector multiply and multiple submatrix-vector multiplies. For the single matrix-vector multiply framework, we propose a cache-size aware top-down row/column-reordering approach based on 1D sparse matrix partitioning by utilizing the recently proposed appropriate hypergraph models of sparse matrices, and a cache oblivious bottom-up approach based on hierarchical clustering of rows/columns with similar sparsity patterns. We also propose a column compression scheme as a preprocessing step which makes these two approaches cache-line-size aware. The multiple submatrix-vector multiplies framework depends on the partitioning the matrix into multiple nonzero-disjoint submatrices. For an effective matrixto-submatrix partitioning required in this framework, we propose a cache-size aware top-down approach based on 2D sparse matrix partitioning by utilizing the recently proposed fine-grain hypergraph model. For this framework, we also propose a traveling salesman formulation for an effective ordering of individual submatrix-vector multiply operations. We evaluate the validity of our models and methods on a wide range of sparse matrices. Experimental results show that proposed methods and models outperforms state-of-the-art schemes. ...
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: thesis
– Name: Format
  Label: File Description
  Group: SrcInfo
  Data: xiv, 56 leaves; application/pdf
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: NoteTitleSource
  Label: Relation
  Group: SrcInfo
  Data: https://hdl.handle.net/11693/15336; B118111
– Name: URL
  Label: Availability
  Group: URL
  Data: https://hdl.handle.net/11693/15336
– Name: Copyright
  Label: Rights
  Group: Cpyrght
  Data: info:eu-repo/semantics/openAccess
– Name: AN
  Label: Accession Number
  Group: ID
  Data: edsbas.39D0640C
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=edsbas&AN=edsbas.39D0640C
RecordInfo BibRecord:
  BibEntity:
    Languages:
      – Text: English
    Subjects:
      – SubjectFull: Cache locality
        Type: general
      – SubjectFull: Sparse matrices
        Type: general
      – SubjectFull: Matrix-vector multiplication
        Type: general
      – SubjectFull: Matrix reordering
        Type: general
      – SubjectFull: Computational hypergraph model
        Type: general
      – SubjectFull: Hypergraph partitioning
        Type: general
      – SubjectFull: Traveling salesman problem
        Type: general
      – SubjectFull: QA188 .A53 2009
        Type: general
      – SubjectFull: Sparse matrices--Data processing
        Type: general
      – SubjectFull: Cache memory
        Type: general
      – SubjectFull: Hypergraphs
        Type: general
    Titles:
      – TitleFull: Cache locality exploiting methods and models for sparse matrix-vector multiplication
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Akbudak, Kadir
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 01
              Type: published
              Y: 2009
          Identifiers:
            – Type: issn-locals
              Value: edsbas
            – Type: issn-locals
              Value: edsbas.oa
ResultId 1