1994
DOI: 10.1007/3-540-57659-2_11
|View full text |Cite
|
Sign up to set email alerts
|

Automatic support for data distribution on distributed memory multiprocessor systems

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
2
1
1
1

Citation Types

0
10
0

Year Published

1996
1996
1998
1998

Publication Types

Select...
3
2

Relationship

0
5

Authors

Journals

citations
Cited by 17 publications
(10 citation statements)
references
References 14 publications
0
10
0
Order By: Relevance
“…Chapman, Fahringer, and Zima have noted that there are many known good data distributions for important numerical problems that are frequently used [6]. They describe the design of a distribution tool that makes use of performance prediction methods when possible, but also heuristically uses empirical performance data when available.…”
Section: Static Partitioningmentioning
confidence: 98%
See 1 more Smart Citation
“…Chapman, Fahringer, and Zima have noted that there are many known good data distributions for important numerical problems that are frequently used [6]. They describe the design of a distribution tool that makes use of performance prediction methods when possible, but also heuristically uses empirical performance data when available.…”
Section: Static Partitioningmentioning
confidence: 98%
“…This is accomplished in O (V 2 ) time 6 (where V is now the number of vertices in the phase transition graph) using Dijkstra's single source shortest path algorithm [38]. After the shortest path has been computed, the loop peeling performed on the PTG can be seen to be necessary to obtain the best solution if the peeled iteration has a different transition sequence than the remaining iterations.…”
Section: Phase and Phase Transition Selectionmentioning
confidence: 99%
“…An automatic tool has been proposed for implementation within the Vienna Fortran Compilation System, or VFCS [15]. The tool will rely heavily on heuristics in making distribution decisions, including data redistribution, and uses both program analysis and knowledge-based techniques.…”
Section: Dynamic Distribution For Dmmsmentioning
confidence: 99%
“…Chapman, Fahringer, and Zima have noted that there are many known good data distribu tions for important numerical problems that are frequently used [6]. They describe the design of a distribution tool that makes use of performance prediction methods when possible, but also heuristically uses empirical performance data when available.…”
Section: Dynamic Partitioningmentioning
confidence: 99%
“…For reference, the functions for an Intel Paragon and a Thinking Machines CM-5, corresponding to the communication primitives used in the edge weights, are shown in Table 1. 3T here is at m ost only one edge between two nodes for each array referenced in the statem en t. For m ultiple references using th e sam e array, th e edge weight is the sum of till com m unication for th at array. 6 (a) 100.000000 * (P2 > 1) * Shift(510.000000) (b) 3100.000000 * Transfer(510.000000) 4. The maximum cut is used to select the phase split; if there is more than one cut with the same maximum value, choose the first.…”
Section: Phase Decompositionmentioning
confidence: 99%