Function Description

The Kunpeng eigenvalue solving library (KML_EIGENSOLVER or KES) provides an iterative solver to obtain eigenvalues of large-scale matrices. It supports multiple MPI processes, user-defined matrix-vector multiplication, and user-defined preconditioners.

The solver provides the following methods to obtain eigenvalues:

• Locally optimal block preconditioned conjugate gradient (LOBPCG): supports data types including single-precision real number, single-precision complex number, double-precision real number, and double-precision complex number. It is applicable to scenarios where a small number of eigenvalues are required and the communication cost is low.
• Preconditioned conjugate projected gradient (PCPG): supports data types including single-precision real number, single-precision complex number, double-precision real number, and double-precision complex number. It is applicable to scenarios where a large number of eigenvalues are required, a lot of nodes are involved, and the communication cost is high.