Title :
Research of Matrix Inversion Acceleration Method
Author :
He, Yingying ; Song, Yukun ; Du, Gaoming ; Zhang, Duoli
Author_Institution :
VLSI Res. Inst., Hefei Univ. of Technol., Hefei, China
Abstract :
Matrix inversion is widely used in scientific calculation and the computation complexity grows in exponent with the matrix dimension. In order to accelerate the computation speed, a block LU decomposition matrix inversion method based on multi-core processing is proposed. Parallel programs for 20*20 matrix inversion are developed and mapped onto an in-house developed multi-core platform, in which there are 4 cores and 8 cores respectively. In order to get the speed-up, the same workload is also ported to a single core environment. Experiments show that the maximum speed-ups are 2.5123 and 3.4936 respectively.
Keywords :
computational complexity; matrix inversion; parallel programming; block LU decomposition matrix inversion; computation complexity; matrix inversion acceleration; multi-core processing; parallel programs; scientific calculation; Acceleration; Helium; Integrated circuit interconnections; Matrix decomposition; Multicore processing; Network-on-a-chip; Parallel processing; Routing; Scalability; Very large scale integration;
Conference_Titel :
Computational Intelligence and Software Engineering, 2009. CiSE 2009. International Conference on
Conference_Location :
Wuhan
Print_ISBN :
978-1-4244-4507-3
Electronic_ISBN :
978-1-4244-4507-3
DOI :
10.1109/CISE.2009.5365280
