The Plural Matrix Inversion Based on Heterogeneous Multi-core Reconfigurable System
Matrix inversion is a common compute-intensive signal processing algorithm, widely applied in kinds of digital signal processing fields with high real-time requirements.Currently, using a customized hardware accelerator to realize a fast solution is the most common method.With the development of multi-core technology, its powerful operation ability provides a new way to solve matrix inversion.The characteristic of strong data dependency in matrix inversion results in a great challenge to implement the algorithm in multi-core system which becomes more obviously when dealing with large dimension matrix.In this paper, combining the Gaussian elimination method with the LU decomposition (lower and upper triangular matrix decomposition) method, a method of multi-dimensional plural matrix inversion is mapped on a heterogeneous multi-core reconfigurable system based on NoC (Network on Chip).Considering the parallelism, storage, speed and so on, the inversion of any plural matrix under 128-order can be calculated through decomposition of parallel algorithm, recombination and task allocation.
Plural Matrix Inversion Heterogeneous Multi-core Reconfigurable System Gaussian Elimination LU Decomposition
Yuwen Luo Xiaoxia Sun Yukun Song
Institute of VLSI Design, Hefei University of Technology, Hefei 230009, Anhui, China
国际会议
三亚
英文
757-762
2015-12-26（万方平台首次上网日期，不代表论文的发表时间）