Pierre Boulet, Jack Dongarra, Yves Robert, and Fr ed eric Vivien. Static tiling for heterogeneous computing platforms. Parallel Computing, 25(5):547-- 568, 1999. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Source Code Transformations Strategies to.. - David, Genaud.. (2002) (Correct)
....execution. Providing a dedicated library to implement load balancing is also proposed by George [9] with the DParLib, who addresses the problem of dynamic load balancing via array distribution for the class of iterative algorithms with a Work is under progress to use theoretical results from [7] so as to nd the optimal number of emulated processes. SPMD implementation. The statistics about load balance produced dynamically during the execution can be used to redistribute arrays from one iteration to the other. However, the library does not take into account possible ....
Boulet, P., Dongarra, J., Robert, Y., Vivien, F.: Static tiling for heterogeneous computing platforms. Parallel Computing 25 (1999) 547568
Source code transformations strategies to.. - David, Genaud.. (2002) (Correct)
....speed ratings of table 1. We choose n 1 =1, n 2 =2, n 3 =4 and n 4 =4 meaning for instance that an Athlon is rated four times better than a R10k. The results reported below in table 2 show signi cant gains in performance. Note however that work is under study to use theoretical results from [BDRV99] so as to nd the optimal number of emulated processes. New experiments are currently undertaken to evaluate the extra gain that can be obtained with this method. In order to validate our code transformation, we measured the wall clock time of: 1. The initial application without load balancing, ....
Pierre Boulet, Jack Dongarra, Yves Robert, and Frdric Vivien. Static tiling for heterogeneous computing platforms. Parallel Computing, 25(5):547-568, 1999.
Matrix-Matrix Multiplication on Heterogeneous Platforms - Beaumont, Boudet.. (2000) (1 citation) (Correct)
....been little work devoted to the implementation of dense linear algebra kernels on heterogeneous platforms. Extensions of parallel libraries such as ScaLAPACK are not yet available, even for simple HNOWs. Preliminary results on implementing MMM and linear system solvers on a HNOWs are reported in [8, 6, 9]. Load balancing issues for heterogeneous 2D grids are studied by Kalinov and Lastovetky [23] in fact, they arrange the processors into columns. The load is rst balanced inside each column independently; next the load is balanced between columns, weighting each column by the inverse of the ....
Pierre Boulet, Jack Dongarra, Yves Robert, and Frederic Vivien. Static tiling for heterogeneous computing platforms. Parallel Computing, 25:547-568, 1999.
Algorithms and Tools for (Distributed) Heterogeneous.. - Mehaut, Robert (1999) (Correct)
....We describe NetSolve in Section 6. 4 Case Study: Cluster ScaLAPACK In this section we deal with our rst case study. We give a particular emphasis on the algorithmic aspects relevant to the design of a Cluster ScaLAPACK library in this section. Our discussion is based on [14, 15, 12, 13, 11, 10]. 4.1 Introduction Consider a heterogeneous NOW: whereas programming a large application made up of several loosely coupled tasks can be performed rather easily (because these tasks can be dispatched dynamically on the available processors) implementing a tightly coupled algorithm (such as a ....
....2 f1; pg; c i = 1 t i P p i=1 1 t i M : # Iteratively increment some c i until c 1 c 2 : c p = M for m = c 1 c 2 : c p to M nd k 2 f1; pg such that t k (c k 1) minft i (c i 1) g c k = c k 1 Algorithm 4. 2 gives the optimal allocation [14, 15]. This algorithm can only be applied to simple load balancing problems such as matrix matrix product on a processor ring. Indeed, such an algorithm can be decomposed into successive communication free steps. The communication between steps are reduced to a simple shift across the ring of ....
[Article contains additional citation context not shown here]
Pierre Boulet, Jack Dongarra, Yves Robert, and Frederic Vivien. Static tiling for heterogeneous computing platforms. Parallel Computing, 25:547-568, 1999.
Code Generation For General Loops Using Methods From - Andronikos, Ciorba.. (2004) (Correct)
No context found.
Pierre Boulet, Jack Dongarra, Yves Robert, and Fr ed eric Vivien. Static tiling for heterogeneous computing platforms. Parallel Computing, 25(5):547-- 568, 1999.
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