Chebyshev Filter Diagonalization on Modern Manycore Processors and GPGPUs
Erlangen Regional Computing Center (RRZE), Friedrich-Alexander University of Erlangen-Nuremberg
arXiv:1803.02156 [cs.MS], (6 Mar 2018)
@article{kreutzer2018chebyshev,
title={Chebyshev Filter Diagonalization on Modern Manycore Processors and GPGPUs},
author={Kreutzer, Moritz and Hager, Georg and Ernst, Dominik and Fehske, Holger and Bishop, Alan R. and Wellein, Gerhard},
year={2018},
month={mar},
archivePrefix={"arXiv"},
primaryClass={cs.MS}
}
Chebyshev filter diagonalization is well established in quantum chemistry and quantum physics to compute bulks of eigenvalues of large sparse matrices. Choosing a block vector implementation, we investigate optimization opportunities on the new class of high-performance compute devices featuring both high-bandwidth and low-bandwidth memory. We focus on the transparent access to the full address space supported by both architectures under consideration: Intel Xeon Phi "Knights Landing" and Nvidia "Pascal." We propose two optimizations: (1) Subspace blocking is applied for improved performance and data access efficiency. We also show that it allows transparently handling problems much larger than the high-bandwidth memory without significant performance penalties. (2) Pipelining of communication and computation phases of successive subspaces is implemented to hide communication costs without extra memory traffic. As an application scenario we use filter diagonalization studies on topological insulator materials. Performance numbers on up to 512 nodes of the OakForest-PACS and Piz Daint supercomputers are presented, achieving beyond 100 Tflop/s for computing 100 inner eigenvalues of sparse matrices of dimension one billion.
March 10, 2018 by hgpu
