Twisted data layout

Shindo, Tatsuya; Iwashita, Hidetoshi; Kaneshiro, Shaun; Doi, Tsunehisa; Hagiwara, Junichi

doi:10.1145/181181.181570

Search citation statements

Order By: Relevance

Paper Sections

Select...

Adi Methods On Numa2

An Example Of Adi1

Citation Types

Supporting

Mentioning

Contrasting

Year Published

1995

2012

Publication Types

Select...

Other3

Book1

Relationship

Self Cite0

Independent4

Authors

Journals

Cited by 4 publications

(3 citation statements)

References 12 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The data distributions are shown for 3 nodes, each represented by a different shade. The first was generated by our Data Distributor [43]; while the second is the well-known twisted data layout [59]. In each distribution, all four 2D arrays (u, v, p, and q) are distributed as shown in the corresponding figure.…”

Section: An Example Of Adimentioning

confidence: 99%

Incremental Parallelization with Migration

Zhang

Pan

Shang

et al. 2012

2012 IEEE 10th International Symposium on Parallel and Distributed Processing With Applications

View full text Add to dashboard Cite

Section: An Example Of Adimentioning

confidence: 99%

Incremental Parallelization with Migration

Zhang

Pan

Shang

et al. 2012

2012 IEEE 10th International Symposium on Parallel and Distributed Processing With Applications

View full text Add to dashboard Cite

“…Thick arrows represent dependencies, and shaded plane represents the plane being updated. The outermost loop in the Fortran program implementation increases [(1) to (4)] then decreases [(5) to (8)] in k direction. When increasing k, an i-j plane is updated by using younger i-j plane, then forward/backward sweepings are done in i direction.…”

Section: Adi Methods On Numamentioning

confidence: 99%

“…To see the behavior with near-optimal NUMA decomposition of the array, we took twisted data layout [8] (or Multipartitioning in [7]) approach and adapted it for NUMA. Figure 7 shows the twisted data layout of i-j plane for 3 threads.…”

Section: Adi Methods On Numamentioning

confidence: 99%

Lessons learned from porting vector computer applications onto non-uniform memory access scalar machines

Hatazaki

Proceedings. Seventh International Conference on High Performance Computing and Grid in Asia Pacific Region, 2004.

View full text Add to dashboard Cite

Although recent large-scale scalar multiprocessor systems have good potential to overwhelm vector machines even in vector-specific application areas, the applicability has not been systematically studied. We ported 2 typical vector applications onto 2 different scalar NUMA platforms. We found that trivial array dimension reordering drastically affect performance. We also show that vectorspecific programming methods could hinder scalar/NUMA system's performance. A general workaround we developed is described along with a discussion for platforms' memory systems characteristics.

show abstract