| |
- L. Giannone, R. Fischer, K. Lackner
MPI/IPP, Garching
- M. D. Cerna, J. Nagle, M. M. Ravindran, Q. Ruan, D. Schmidt, A. Veeramani, A. Vrancic, L. Wenzel
National Instruments, Austin
- P. J. McCarthy
National University of Ireland, University College Cork
| |
Big physics control experiments require enormous computational power to solve large problems with demanding real-time constraints. Sensors are acquired in real-time to feed mathematical routines, which then generate control outputs to real-world processes. The underlying mathematics can be sophisticated, as even non-linear PDEs have to be solved thousands of times per second. Is low-level coding of highly specialized hardware required to meet the challenge? We report on an alternative approach based on LabVIEW that addresses demanding plasma shape control in tokamaks. A variety of input signals (magnetic coils, X-ray sensors) are combined and a constrained non-linear Grad-Shafranov PDE is solved to calculate the magnetic equilibrium in under 0.5 ms with low jitter. The off-the-shelf n-core hardware and graphical software architecture is described with a strong emphasis on the seamless interplay between development system and real-time target deployment. Numerous mathematical challenges were addressed and several generally-applicable numerical and mathematical strategies proved critical to the timing goals. Several benchmarks illustrate what can be achieved with this approach.
|
|
|
Poster
|
|
|
|