2014
DOI: 10.1137/130919507
|View full text |Cite
|
Sign up to set email alerts
|

Adaptive Aggregation-Based Domain Decomposition Multigrid for the Lattice Wilson--Dirac Operator

Abstract: In lattice QCD computations a substantial amount of work is spent in solving linear systems arising in Wilson's discretization of the Dirac equations. We show first numerical results of the extension of the two-level DD-αAMG method to a true multilevel method based on our parallel MPI-C implementation. Using additional levels pays off, allowing to cut down the core minutes spent on one system solve by a factor of approximately 700 compared to standard Krylov subspace methods and yielding another speed-up of a … Show more

Help me understand this report

Search citation statements

Order By: Relevance

Paper Sections

Select...
1
1
1
1

Citation Types

2
200
0

Year Published

2015
2015
2023
2023

Publication Types

Select...
4
3
2

Relationship

0
9

Authors

Journals

citations
Cited by 156 publications
(204 citation statements)
references
References 41 publications
2
200
0
Order By: Relevance
“…[38] (see also Refs. [39,40] The distance | 0 | used to determine the improvement coefficients can in principle be chosen at will as long as | 0 | = | 0 | is kept (approximately) constant, to achieve the complete removal of order terms from continuum limit extrapolations of physical observables. We restricted ourselves to points where lattice artefacts on the coefficients are small -at least at tree-level.…”
Section: Acknowledgmentsmentioning
confidence: 99%
“…[38] (see also Refs. [39,40] The distance | 0 | used to determine the improvement coefficients can in principle be chosen at will as long as | 0 | = | 0 | is kept (approximately) constant, to achieve the complete removal of order terms from continuum limit extrapolations of physical observables. We restricted ourselves to points where lattice artefacts on the coefficients are small -at least at tree-level.…”
Section: Acknowledgmentsmentioning
confidence: 99%
“…[2] as a solver for the clover-improved Wilson Dirac operator D. In the DD-αAMG method a flexible iterative Krylov solver is preconditioned at every iteration step by a multigrid approach given by the error propagation…”
Section: Dd-αamg Methodsmentioning
confidence: 99%
“…This contains several algorithmic improvements, e.g., the Hasenbusch trick, higher order integrators, a multi-level integration scheme, a deflated solver [18,6] and twisted mass reweighting: in the light fermion part of the action a twisted mass term is introduced in order to push eigenvalues of the Dirac operator away from zero and, hence, increase the stability of the HMC simulation. This is then corrected for by reweighting the observables accordingly.…”
Section: Simulation Detailsmentioning
confidence: 99%
“…For the computation of the correlators a custom version of the CHROMA software package [4] including the LIBHADRONANALYSIS library has been developed where also the multigrid solver implementation of Refs. [5,6] is used.…”
Section: Measurement Detailsmentioning
confidence: 99%