Implementations of main algorithms for generalized eigenproblem on GPU accelerator

  • Authors:
  • Yonghua Zhao;Jian Zhang;Xuebin Chi

  • Affiliations:
  • Supercomputing Center, Computer Network Information Center, Chinese Academy of Sciences, Beijing, China;Supercomputing Center, Computer Network Information Center, Chinese Academy of Sciences, Beijing, China;Supercomputing Center, Computer Network Information Center, Chinese Academy of Sciences, Beijing, China

  • Venue:
  • ICSI'12 Proceedings of the Third international conference on Advances in Swarm Intelligence - Volume Part II
  • Year:
  • 2012

Quantified Score

Hi-index 0.00

Visualization

Abstract

A generalized eigensystem problem is usually transformed, utilizing Cholesky decomposition, to a standard eigenproblem. The latter is then solved efficiently by a matrix reduction approach based on Householder tridiagonalization method. We present parallel implementation of an integrated transformation-reduction algorithm on GPU accelerator using CUBLAS. Experimental results clearly demonstrate the potential of data-parallel coprocessors for scientific computations. When comparing against the CPU implementation, the GPU implementations achieve above 16-fold and 26-fold speedups in double precision for reduction and transformation respectively.