Topic/Type: 2.4 Gyro-kinetic and gyro-fluid methods, Oral

gGEM: a gyrofluid model to be used on distributed platforms

M. Rodr?guez-Pascual1, B.D. Scott2, T.T. Ribeiro2, 3, F. Castej?n1, 4, R. Mayo1

1 CIEMAT
2 Max-Planck IPP
3 IPFN
4 CIEMAT

Gyrofluid models have been usually applied to tokamak core turbulence for incorporating finite ion gyroradius effects at arbitrary order into simple computations of turbulence occurring in largely 2D fluid experiments [1]. Then, the temperatures were incorporated in order to treat ion temperature gradient (ITG) turbulence and the model acquired several new advection terms, producing nonlinearities as well as drift frequency corrections, resulting from the effect of temperature fluctuations on the gyroaveraging operator [2].

Thus, the gyrofluid electromagnetic (GEM) model [3] was introduced in the context of edge turbulence studies [4], a matter of adding collisions and electromagnetic induction to the parallel dynamics of the ?standard? six-moment toroidal model previously used in core turbulence studies [2]. Several improvements have been done in the code since then and now it is able to describe the fluctuation free-energy conservation in a gyrofluid model. The polarization equation relates E x B flow and eddy energy to combinations of the potential and the density and perpendicular temperature

GEM has usually been run on high performance computers. In this work we describe how the application has been updated to run on standard X86 clusters, either with or without MPI support, so it can also be executed on Grid and distributed platforms in the framework of the EUFORIA Project [5]. In addition, a DRMAA API application has been developed to simplify the execution of GEM creating in this way the gGEM tool; for improving the performance, the GridWay metascheduler [6] has been also incorporated for maximizing the speed and reliability of executions on the Grid. The scalability and correctness of our solution has been evaluated in a local cluster. Fault tolerance and Grid suitability has been demonstrated by executing our application in the EUFORIA infrastructure.

With this work, a new strategy is open for the simulation of fusion plasmas since it can be coupled to other codes (European Transport Solver [7], Helena [8], etc.) and is ready to be executed on more powerful High Performance Computing architectures based on separated island of hundreds of cores.

[1] G. Knorr, F.R. Hansen, J.P. Lynov, H.L. P?cseli, and J.J. Rasmussen, Phys. Scr. 38, 829 (1988).

[2] M.A. Beer, and G. Hammett, Phys. Plasmas 3, 4046 (1996).

[3] B.D. Scott, Phys. Plasmas 12, 102307 (2005)

[4] B.D. Scott, Phys. Plasmas 7, 1845 (2000).

[5] The EUFORIA Project, available at http://www.euforia-project.eu/
[6] E. Huedo, R.S. Montero, I.M. Llorente, Scalable Computing-Practice & Experience 6, 1 (2005).

[7] D. Kalupin et al., Proc. 35th EPS Conference on Plasma Phys. 32D, P-5.027 (2008).

[8] G. Huysmans, J. Goedbloed, and W. Kerner, Proc CP90 Europhysics Conf. Computational Physics, 371 (1990).