0

I have two 50,000 x 50,000 symmetric real matrices A and B (double precision) which are non-sparse (55% non-zero). B is positive-definite.

I have this generalized eigenproblem: A v = ยต B v

I need to find 3 - 4 of the smallest algebraic eigenvalues (and if possible their associated eigenvectors).

Do I have still any options left to compute them on an average computer with 12 GB of RAM?

Any suggestions that I can either try without much effort or that definitely work. Thanks!

Ystar
  • 351
  • 1
  • 3
  • 8

1 Answers1

0

Your only options is to use out-of-core solvers.

Here is the scalapack reference: http://www.netlib.org/scalapack/prototype/readme.outofcore

But you have much time for this, isn't it? getting only few eignenvalues can render this problem feasible in a reasonable amount of time, but I would recommend to estimate the needed time before starting to code.

Sigi
  • 4,826
  • 1
  • 19
  • 23