Sparse Eigensolver on GPU for 3D Lattice QCD
Lead PI:Dr. Mike Peardon
Abstract:This project is concerned with Linear Algebra on GPU with CUDA, and forms part of Lattice QCD research at TCD. 3D Lattice data is supplied, including periodic boundary conditions, producing Hermitian Matrices with sizes in the 10,000s (24x24x24). The goal is to use the GPU to efficiently calculate the lowest modes of the matrices and their corresponding eigenvectors (100s).This will involve implementing the Lanzcos algorithm for sparse matrices in CUDA.