Efficient two-level preconditionined conjugate gradient method on the GPU

Title

Efficient two-level preconditionined conjugate gradient method on the GPU

 Author

Gupta, R.
Van Gijzen, M.B.
Vuik, K.

Faculty

Electrical Engineering, Mathematics and Computer Science

 Date

2011-12-31

 Abstract

We present an implementation of Two-Level Preconditioned Conjugate Gradient Method for the GPU. We investigate a Truncated Neumann Series based preconditioner in combination with deflation and compare it with Block Incomplete Cholesky schemes. This combination exhibits fine-grain parallelism and hence we gain considerably in execution time. It’s numerical performance is also comparable to the Block Incomplete Cholesky approach. Our method provides a speedup of up to 16 times for a system of one million unknowns when compared to an optimized implementation on the CPU.

 To reference this document use:

http://resolver.tudelft.nl/uuid:73526ca9-769b-40b1-8552-d0cdc7fae344

 Publisher

Delft University of Technology, Faculty of Electrical Engineering, Mathematics and Computer Science, Delft Institute of Applied Mathematics

 ISSN

1389-6520

 Source

Reports of the Department of Applied Mathematical Analysis, 11-15

 Part of collection

Institutional Repository

 Document type

report

 Rights

(c)2011 Gupta, R., Van Gijzen, M.B., Vuik, K.
Files
PDF 11-15Gupta.pdf 886.06 KB
Close viewer