Amestoy, P. R. and Puglisi, C. (2000), An unsymmetrized multifrontal LU factorization, Technical Report RT/APO/00/3, ENSEEIHT-IRIT. Also Lawrence Berkeley National Laboratory Report LBNL-46474. Home/Search Document Details and Download
Summary Related Articles Check |
This paper is cited in the following contexts:
Performance and Tuning of Two Distributed Memory Sparse.. - Amestoy, Duff.. (2001) Self-citation (Amestoy) (Correct)
....to one. We use the MC64 (Duff and Koster 1999) code of HSL (HSL 2000) to perform this preordering and scaling (Amestoy, Duff, L Excellent and Li 2000b) Both approaches use Level 3 BLAS to perform the elimination operations. However, in MUMPS the frontal matrices are always square. It is shown in Amestoy and Puglisi (2000) how one can detect and exploit sparsity within the frontal matrices but the present implementation takes no advantage of this sparsity and all the counts measured assume the frontal matrix is dense. In 1 SuperLU, advantage is taken of sparsity in the blocks and usually the dense matrix blocks ....
....computation and communication, and the algorithm s degree of parallelism. In the case of MUMPS, the porting of the code to the 512 processor CRAY T3E 900 gave us the opportunity to study the behaviour of the code on a larger number of processors than used in our previous work (Amestoy et al. 1999, Amestoy et al. 2000a) From our set of machine dependent parameters we chose appropriate parameters to address this issue. Other algorithmic modifications were motivated by having more processors available to us than formerly. The dynamic scheduling approach used in MUMPS was modified (see Amestoy et al. 2000b) to ....
[Article contains additional citation context not shown here]
Amestoy, P. R. and Puglisi, C. (2000), An unsymmetrized multifrontal LU factorization, Technical Report RT/APO/00/3, ENSEEIHT-IRIT. Also Lawrence Berkeley National Laboratory Report LBNL-46474.
Analysis and Comparison of Two General Sparse Solvers .. - Amestoy, Duff.. (2000) Self-citation (Amestoy) (Correct)
....In this section, we briefly describe the main characteristics of the algorithms used in the solvers and highlight the major differences between them. For a complete description of the algorithms, the reader should consult previous papers by the authors of these algorithms (Amestoy et al. 1999, Amestoy et al. 2000, Li and Demmel 1998, Li and Demmel 1999) Both algorithms can be described by a computational tree whose nodes represent computations and whose edges represent transfer of data. In the case of the multifrontal method, MUMPS, some steps of Gaussian elimination are performed on a dense frontal ....
....square. It is possible that there are zeros in the frontal matrix especially if there are delayed pivots or the matrix structure is markedly asymmetric but the present implementation takes no advantage of this sparsity and all the counts measured assume the frontal matrix is dense. It is shown in Amestoy and Puglisi (2000) that one can detect and exploit the structural asymmetry of the frontal matrices. With this new algorithm, significant gains both in memory and in time to perform the factorization can be obtained. For example, using MUMPS with the new algorithm, the number of operations to factorize matrices ....
[Article contains additional citation context not shown here]
Amestoy, P. R. and Puglisi, C. (2000), An unsymmetrized multifrontal LU factorization, Technical Report RT/APO/00/3, ENSEEIHT-IRIT. Also Lawrence Berkeley National Laboratory Report LBNL-46474.
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