Geoffrey C. Fox. Numerical Algorithms for Modern Parallel Computer Architectures, chapter A Graphical Approach to Load Balancing and Sparse Matrix Vector Multiplication on the Hypercube, pages 37-62. Springer-Verlag, 1988.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....load balancing, but they differ in the approach they take to providing physical locality. 8.3. 1 Partitioning Space: Orthogonal Recursive Bisection Orthogonal Recursire Bisection (ORB) is a technique for providing physical locality in a problem domain by explicitly partitioning the domain space [10]. It was first used for hierarchical N body problems in Salmon s message passing Barnes Hut implementation [23] The idea in ORB partitioning is to recursively divide the computational domain space into two subspaces with equal costs, until there is one subspace per processor (see Figure 11) To ....

Geoffrey C. Fox. Numerical Algorithms for Modern Parallel Computer Architectures, chapter A Graphical Approach to Load Balancing and Sparse Matrix Vector Multiplication on the Hypercube, pages 37-62. Springer-Verlag, 1988.

.... two dimensional grids have been developed [2, 3] Furthermore, adaptive local refinement coarsening of unstructured tetrahedral grids has been developed and implemented for complex, 3 D geometry flow simulations [4, 5] Relatively few parallel solvers for unstructured grids have been developed [6, 7, 8, 9]. These typically employ static grids, which are partitioned and allocated to 3 processors a priori by the user. Algorithms for load balancing in case of 2 D grids that do not change ( also referred as partitioning algorithms) have been presented in [9, 10, 11] Two approaches are simulated ....

G. C. Fox, "Numerical Algorithms for Modern Parallel Computers", ed. M. Schultz, Springer Verlag, Berlin, 1988.

....and hence the cell structure changes dynamically, the partitioning is redone every time step. 5.2. 1 Partitioning Space: Orthogonal Recursive Bisection Orthogonal Recursive Bisection (ORB) is a technique for providing physical locality in a problem domain by explicitly partitioning the domain space [5]. The idea here is to recursively divide space into two subspaces with equal costs, until there is one subspace per processor (see Figure 7) Initially, all processors are associated with the entire domain space. Every time a space is divided, half the processors associated with it are assigned to ....

Geoffrey C. Fox. Numerical Algorithms for Modern Parallel Computer Architectures, chapter A Graphical Approach to Load Balancing and Sparse Matrix Vector Multiplication on the Hypercube, pages 37--62. SpringerVerlag, 1988.

....list and force computation phases is mandatory to achieve parallel efficiency. To exploit spatial locality in IC CEDAR, we use a heuristic that distributes and groups objects based on their spatial coordinates. A common approach to decompose objects into clusters is Orthogonal Recursive Bisection [11], which recursively bisects the simulation space. However, unlike simulations involving only point objects in space, a large number of objects in IC CEDAR (such as spatial cells and charge groups) span a volume in space. Therefore, any recursive bisector of space is likely to intersect many ....

Geoffrey C. Fox. Numerical Algorithms for Modern Parallel Computer Architectures. Springer-Verlag, 1988.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC