Numerics: Exact diagonalization
Iterative eigenvalue algorithm: Jacobi-Davidson (A. Basermann, NEC)
Performance&memory requirements determined by:
- sparse matrix-vector multiplication (JDS Format)
- BLAS-1 (Vector-add and vector mult-add)
Task: Compute lowest 10-100 eigenvalues of large (DMat=108) sparse matrices
Linear scaling, memory bound algorithm