The matrix is in the form of Ax=SBx where S is the eigen value. Both A and B are large sparse matrices. The number of elements are 4266*4266. I am using EISPACK for obtaining the eigen values but getting NAN or infinity as a solution. I have checked both A and B matrices and they were correct. Is it possible that my pc is unable to do such big calculations? Is parallel computing required for such large sparse matrices?
Hello,
If the program returns an answer, so there is no need for parallel computing. because the parallel computing only speeds up the program, nothing more. Besides, Fortran is much more faster than other programs like MATLAB.
I think, if there is a problem, it would be on the matrix B ( check B if it is singular ).
Check the condition number of the matrices and the ranks. If the ranks are not the same for A and B, you may have a rank deficiency problem. Second suggestion would be to use PARDISO of Intel MKL library. It should have a better and more direct solver choices.
I agree with Ramy, you may have a matrix conditioning issue that avoids convergence of the solver.
You may have rank deficiency or have small critical terms in your matrices that will get you close to the rank deficiency. In such case since we are working in finite precision resolution may fail. You should check the maximum and minimum terms in your matrix and their distribution to ensure that yo are not in this case.
- Badges
- Science topic
- Similar topics
- Engineering
- Mechanical Engineering