On Parallelizing the MRRR Algorithm for Data-Parallel Coprocessors
نویسندگان
چکیده
The eigenvalues and eigenvectors of a symmetric matrix are needed in a myriad of applications in computational engineering and computational science. One of the fastest and most accurate eigensolvers is the Algorithm of Multiple Relatively Robust Representations (MRRR). This is the first stable algorithm that computes k eigenvalues and eigenvectors of a tridiagonal symmetric matrix in O(nk) time. We present a parallelization of the MRRR algorithm for data parallel coprocessors using the CUDA programming environment. The results clearly demonstrate the potential of data-parallel coprocessors for scientific computations: when comparing against routine sstemr, LAPACK’s implementation of MRRR, our parallel algorithm provides 10-fold speedups.
منابع مشابه
Parallelizing Assignment Problem with DNA Strands
Background:Many problems of combinatorial optimization, which are solvable only in exponential time, are known to be Non-Deterministic Polynomial hard (NP-hard). With the advent of parallel machines, new opportunities have been emerged to develop the effective solutions for NP-hard problems. However, solving these problems in polynomial time needs massive parallel machines and ...
متن کاملAn Implementation of the MRRR Algorithm on a Data-Parallel Coprocessor
The Algorithm of Multiple Relatively Robust Representations (MRRRR) is one of the most efficient and most accurate solvers for the symmetric tridiagonal eigenvalue problem. We present an implementation of the MRRR algorithm on a data-parallel coprocessor using the CUDA programming environment. We obtain up to 50-fold speedups over LAPACK’s MRRR implementation and demonstrate that the algorithm ...
متن کاملA New Approach to Solve N-Queen Problem with Parallel Genetic Algorithm
Over the past few decades great efforts were made to solve uncertain hybrid optimization problems. The n-Queen problem is one of such problems that many solutions have been proposed for. The traditional methods to solve this problem are exponential in terms of runtime and are not acceptable in terms of space and memory complexity. In this study, parallel genetic algorithms are proposed to solve...
متن کاملA Message-Passing Distributed Memory Parallel Algorithm for a Dual-Code Thin Layer, Parabolized Navier-Stokes Solver
In this study, the results of parallelization of a 3-D dual code (Thin Layer, Parabolized Navier-Stokes solver) for solving supersonic turbulent flow around body and wing-body combinations are presented. As a serial code, TLNS solver is very time consuming and takes a large part of memory due to the iterative and lengthy computations. Also for complicated geometries, an exceeding number of grid...
متن کاملLapack Working Note 195: Scalapack’s Mrrr Algorithm
The sequential algorithm of Multiple Relatively Robust Representations, MRRR, can compute numerically orthogonal eigenvectors of an unreduced symmetric tridiagonal matrix T ∈ R with O(n) cost. This paper describes the design of ScaLAPACK’s parallel MRRR algorithm. One emphasis is on the critical role of the representation tree in achieving both numerical accuracy and parallel scalability. A sec...
متن کامل