Multigrid implementation of the Fourier acceleration method for Landau gauge fixing

Physical Review D - Particles, Fields, Gravitation and Cosmology

Volumn 57, Issue 7, 1998, Pages R3822-R3826

  Cucchieri, Attilio ^a   Mendes, Tereza ^a  

^a UNIVERSITY OF ROME TOR VERGATA   (Italy)

Author keywords

[No Author keywords available]

Indexed keywords

EID: 0000912913     PISSN: 15507998     EISSN: 15502368     Source Type: Journal    
DOI: 10.1103/PhysRevD.57.R3822     Document Type: Article

References (25)

1. For a review, see, G.C. Rossi, in Lattice ’96, Proceedings of the International Symposium, St. Louis, Missouri, edited by C. Barnard et al. [Nucl. Phys. B (Proc. Suppl.) 53, 3 (1997)].
2. See, for example, U. Wolff, in Lattice ’89, Proceedings of the International Symposium, Capri, Italy, edited by R. Petronzio et al. [Nucl. Phys. B (Proc. Suppl.) 17, 93 (1990)];
3. A.D. Sokal, in Lattice’90, Proceedings of International Symposium, Tallahasse, Florida, edited by U.M. Heller et al. [ibid. 20, 55 (1990)].
4. H. Suman and K. Schilling, Parallel Computing 20, 975 (1994).
5. C.T.H. Davies, Phys. Rev. D 37, 1581 (1988).
6. G.G. Batrouni, Phys. Rev. D 32, 2736 (1985);
7. Phys. Rev. DG. Katz, et al., 37, 1589 (1988);
8. E. Dagotto and J.B. Kogut, Phys. Rev. Lett. 58, 299 (1987);
9. E. Dagotto and J.B. Kogut, Nucl. Phys. B290, 451 (1987).
10. A. Cucchieri and T. Mendes, in Lattice’96 [1], p. 811;A. Cucchieri and T. Mendes (in preparation).
11. A. Cucchieri and T. Mendes, Nucl. Phys. B471, 263 (1996).
12. P. Marenzoni, G. Martinelli, and N. Stella, Nucl. Phys. B455, 339 (1995).
13. K.G. Wilson, Recent Developments in Gauge Theories, Proceedings of the NATO Advanced Study Institute, Carges̀e, 1979, edited by G. ’t Hooft (Plenum, New York, 1980);
14. J.E Mandula and M. Ogilvie, Phys. Lett. B 185, 127 (1987).
15. W.H. Press, Numerical Recipes in Fortran, 2nd ed. (Cambridge University Press, Cambridge, England, 1992).
16. Very recently, a new implementation of FFT for SIMD systems has been presented in Th. Lippert, hep-lat/9710060. The method is applied to a two-dimensional case and could in principle be extended to four dimensions.
17. A. Cucchieri and T. Mendes, hep-lat/9710040 (to be published in the Proceedings of the Lattice 97 Conference).
18. For details of the implementation and for an introduction to the recursive (deterministic) MG method, see, for example, Sec. II of J. Goodman and A.D. Sokal, Phys. Rev. D 40, 2035 (1989), and references therein.
19. We note that we use here the MG routine inside the FA method, as an alternative way of dealing with the Fourier transforms in Eq. (2). This should not be confused with the multigrid gauge-fixing method presented by Ph. de Forcrand and R. Gupta, in Lattice ’88, Proceedings of the International Symposium, Batavia, Illinois, edited by A. Kronfeld and P. MacKenzie [Nucl. Phys. B (Proc. Suppl.) 9, 516 (1989)];
20. A. Hulsebos, M.L. Laursen, and J. Smit, Phys. Lett. B 291, 431 (1992). In that case, multigrid is used directly in order to minimize (Formula presented) The method seems to reduce CSD significantly only if multigrid is combined with overrelaxation.
21. L. Stals, Parallel Implementation of Multigrid Methods, Proceedings of the Sixth Biennial Conference on Computational Techniques and Applications, Canberra, 1993, edited by D. Stewart (World Scientific, Singapore, 1993).
22. J.E. Mandula and M. Ogilvie, Phys. Lett. B 248, 156 (1990).
23. F. Gutbrod, “A Study of the Gluon Propagator in SU (2)Lattice Gauge Theory,” DESY-96-252 report.
24. J.E. Hetrick and Ph. de Forcrand, hep-lat/9710003 (to be published in the Proceedings of the Lattice 97 Conference).
25. On the coarsest grid, we have used Gauss-Seidel relaxation. In vector and parallel machine applications of multigrid one often uses a conjugate gradient algorithm to relax the solution on the coarsest grid [see, for example, S. Solomon and P.G. Lauwers, in Proceedings of the Workshop on Fermion Algorithms, Jülich, 1991, edited by H.J. Herrmann and F. Karsch (World Scientific, Singapore, 1991), p. 149]. In our case, even for our largest coarsest grid (Formula presented), we do not see an effective gain with respect to the simple Gauss-Seidel update (see Ref. 12).