Karthik Swaminathan, Ramon Bertran, et al.
DSN 2023
We propose a distributed-memory parallel algorithm for computing some of the algebraically smallest eigenvalues (and corresponding eigenvectors) of a large, sparse, real symmetric positive definite matrix pencil that lie within a target interval. The algorithm is based on Chebyshev interpolation of the eigenvalues of the Schur complement (over the interface variables) of a domain decomposition reordering of the pencil and accordingly exposes two dimensions of parallelism: one derived from the reordering and one from the independence of the interpolation nodes. The new method demonstrates excellent parallel scalability, comparing favorably with PARPACK, and does not require factorization of the mass matrix, which significantly reduces memory consumption, especially for 3D problems. Our implementation is publicly available on GitHub.
Karthik Swaminathan, Ramon Bertran, et al.
DSN 2023
Leonard Dervishi, Xinyue Wang, et al.
NDSS 2023
Christoph Auer, Michele Dolfi, et al.
CLOUD 2022
Tengfei Ma, Trong Nghia Hoang, et al.
UAI 2023