Realistic Performance Prediction Tool for the Parallel Block LU Factorization Algorithm
نویسندگان
چکیده
This work describes a realistic performance prediction tool for the parallel block LU factorization algorithm. It takes into account the computational workload, communication costs and the overlapping of communications by useful computations. Estimation of the tool parameters and benchmarking are also discussed. Using this tool we develop a simple heuristic for scheduling LU factorization tasks. Results of numerical experiments are presented.
منابع مشابه
Computing a block incomplete LU preconditioner as the by-product of block left-looking A-biconjugation process
In this paper, we present a block version of incomplete LU preconditioner which is computed as the by-product of block A-biconjugation process. The pivot entries of this block preconditioner are one by one or two by two blocks. The L and U factors of this block preconditioner are computed separately. The block pivot selection of this preconditioner is inherited from one of the block versions of...
متن کاملCalculs pour les matrices denses : coût de communication et stabilité numérique. (Dense matrix computations : communication cost and numerical stability)
This dissertation focuses on a widely used linear algebra kernel to solve linear systems, that is the LU decomposition. Usually, to perform such a computation one uses the Gaussian elimination with partial pivoting (GEPP). The backward stability of GEPP depends on a quantity which is referred to as the growth factor, it is known that in general GEPP leads to modest element growth in practice. H...
متن کاملPoLAPACK: parallel factorization routines with algorithmic blocking
LU, QR, and Cholesky factorizations are the most widely used methods for solving dense linear systems of equations, and have been extensively studied and implemented on vector and parallel computers. Most of these factorization routines are implemented with blockpartitioned algorithms in order to perform matrix-matrix operations, that is, to obtain the highest performance by maximizing reuse of...
متن کاملDesign and Performance Modeling of Parallel Block Matrix Factorizations for Distributed Memory Multicomputers
EEcient and scalable parallel block algorithms for the LU factorization with partial pivoting, the Cholesky, and QR factorizations in a distributed memory multicomputer environment are presented. The distributed system is viewed as a ring of processors and the algorithms correspond to shared memory algorithms parallelized on block level (explicit parallelism). Performance of the algorithms are ...
متن کاملParallel Multilevel Block ILU Preconditioning Techniques for Large Sparse Linear Systems
We present a class of parallel preconditioning strategies built on a multilevel block incomplete LU (ILU) factorization technique to solve large sparse linear systems on distributed memory parallel computers. The preconditioners are constructed by using the concept of block independent sets. Two algorithms for constructing block independent sets of a distributed sparse matrix are proposed. We c...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
برای دانلود متن کامل این مقاله و بیش از 32 میلیون مقاله دیگر ابتدا ثبت نام کنید
ثبت ناماگر عضو سایت هستید لطفا وارد حساب کاربری خود شوید
ورودعنوان ژورنال:
- Informatica, Lith. Acad. Sci.
دوره 14 شماره
صفحات -
تاریخ انتشار 2003