A range of applications and codes have been developed throughout Europe addressing different aspects of ITER plasmas. Full scale predictions for ITER sized fusion devices are currently only available at a comparatively low level of physics fidelity and advancement in this area is needed and requires improved access to computing facilities.
Current modelling activities relies on local or national computational resources an improved access to computing infrastructures will be instrumental in advancing a pan-European modelling activity for ITER to a very competitive status in relation to the ITER partners.
The aim of this project is thus to bring local code developments and modelling activities together into a single infrastructure expanding the user base of the modelling codes and at the same time bringing Grid- and High Performance Computing to the fusion modelling community.
This will in turn help advance the level of physics refinement and increase the modelling capabilities.
Within the fusion community, a large number of codes are in use to simulate various aspects of the plasma behaviour.With very few exceptions, these codes have been written by physicists with an emphasis on the physics, with a much smaller emphasis on using the latest technologies available from the computer science community.This proposal aims at improving the situation in a number of ways:
· Increasing the performance of key existing codes
· Enabling a subset of the existing codes for use on computational grids; in this context two subsets have been identified
* Grid-parallel (where the number of CPUs is expected to be in the range of 4-64)
· Parallelizing, or, in some cases, improving the parallelisation of a subset of the codes
· Implement a work flow environment that will be able to couple together a subset of the codes into an integrated simulation package, with some codes running on the workflow machine, some via the grid and some on high performance computers
· The integration of a visualization environment that can be used to monitor existing work-flows, as well as for post-processing.
In order to focus the effort, an important physics sub-domain was identified, that of transport and turbulence in the core and edge.Within this area, a subset of the available codes has been identified on the basis of:
· Interest from the code author/maintainer/responsible officer
· Willingness to make the code available to the wider EU fusion community after the effort finishes
Coverage of the physics as well as coverage of the various numerical techniques used to solve the problems".
The following codes have been identified as of potential interest for their adaptation:
- BIT1 Kinetic 1D3V (1D in usual and 3D in velocity space) code for simulation of the plasma edge.
- CENTORI Fully toroidal two-fluid, electromagnetic turbulence simulation code.
- COREDIV Transport of energy code, main ions and impurity ions in the core and the scrape of layer regions.
- EIRENE Kinetic neutral particle and line radiation transport code.
- ELMFIRE Gyro-kinetic full-f particle code, with mostly global emphasis.a
- ERO Gyro-kinetic code for impurity transport in plasma.
- ESEL Turbulence and profile evolution code at the outboard mid-plane in the SOL, using a fluid (ESEL) and gyrofluid (GESEL) approach.
- GEM Gyrofluid code (GEM is local, GEMX is nonlocal, same infrastructure, similar scheme and programming, somewhat differently formulated equations).
- GENE Nonlinear gyrokinetic code to investigate plasma turbulence.
- ISDEP Kinetic theory of transport code based on Langevin Equations.
- SOLPS Two codes (B2-Eirene) tightly coupled together based code.
- TECXY 2D multifluid plasma and impurity transport in the tokamak edge simulating code.
- TYR Drift Alfven plasma fluid turbulence and transport in flux-tube geometry code.The currently identified Grid-serial codes are BIT1, COREDIV, EIRENE, ERO, ESEL, ISDEP and SOLPS. Except for EIRENE, the rest of listed codes are being also considered for HPC parallelization. It is also possible and likely that additional codes might be identified. Of the above list of Grid-serial codes, all are also interested in Grid-parallel versions as well. In addition, the following codes are only of Grid interest in Grid-parallel versions: CENTORI, ELMFIRE, GEM, GENE, TECXY and TYR. All of them are also going to be studied and becoming candidates to be HPC ported.