Optimizing blocked-Davidson method based on adaptive algorithm using the new Sunway supercomputer

Chaochao Yang; Anjun Liu; Weifeng Liu; Jie Hou; Meng Guo

doi:10.1117/12.2673339

13 April 2023 Optimizing blocked-Davidson method based on adaptive algorithm using the new Sunway supercomputer

Chaochao Yang, Anjun Liu, Weifeng Liu, Jie Hou, Meng Guo

Proceedings Volume 12605, 2022 2nd Conference on High Performance Computing and Communication Engineering (HPCCE 2022); 1260512 (2023) https://doi.org/10.1117/12.2673339
Event: Second Conference on High Performance Computing and Communication Engineering, 2022, Harbin, China

Abstract

The Davidson method is a traditional large-scale iterative diagonalization method in computational quantum chemistry. The Blocked-Davidson method is a classical version of the Davidson method, which is an important part of VASP software, occupying 22.9% of the calculation time. When VASP is transplanted to the new Sunway supercomputer, it is incompatible with the many-core heterogeneous environment of Sunway processor. The Blocked-Davidson method does not use slave-core accelerators. This paper designs collaborative computing, double buffer, store variables in slave-core LDM, LDM level two access, and segmented transmission methods of Blocked-Davidson method based on Sunway many-core architecture. The test results showed that the function in the Blocked-Davidson method accelerated 30.65 times with a single core group in the silica test case. However, negative optimization of the Blocked-Davidson method happens in some test cases such as ZrTiAlVSiN5 and Hg50. Therefore, an adaptive algorithm of automatic selection of less time-consuming optimization paths is designed to solve the negative optimization. The maximum acceleration is 23.6 times compared with not using the adaptive algorithm.

Citation Download Citation

Chaochao Yang, Anjun Liu, Weifeng Liu, Jie Hou, and Meng Guo "Optimizing blocked-Davidson method based on adaptive algorithm using the new Sunway supercomputer", Proc. SPIE 12605, 2022 2nd Conference on High Performance Computing and Communication Engineering (HPCCE 2022), 1260512 (13 April 2023); https://doi.org/10.1117/12.2673339

ACCESS THE FULL ARTICLE

INSTITUTIONAL
Select your institution to access the SPIE Digital Library.

SELECT YOUR INSTITUTION

PERSONAL
Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.

PERSONAL SIGN IN

No SPIE Account? Create one

PURCHASE THIS CONTENT

SUBSCRIBE TO DIGITAL LIBRARY

50 downloads per 1-year subscription

Members: $195

Non-members: $335 ADD TO CART

25 downloads per 1 - year subscription

Members: $145

Non-members: $250 ADD TO CART

PURCHASE SINGLE ARTICLE

Includes PDF, HTML & Video, when available