ANL, Argonne, Illinois
Fermilab, Batavia, Illinois
Large-scale beam dynamics simulations are important in accelerator design and optimization. With the new BG/P supercomputer installed at ANL, tera-scale computing can be easily accessed. In order to make use of this emerging technology to increase the speed and efficiency of accelerator simulations, we have systematized and upgraded our software. In this paper, we will first introduce the new version of the parallel beam dynamic code PTRACK [1] updated to run on BG/P with more than 104 processors. The new PTRACK includes possibility to track ~100,000,000 particles through multiple accelerator seeds in the presence of machine errors. An example of SNS linac simulations will be presented.
[1]. J. Xu, B. Mustapha, V.N. Aseev and P.N. Ostroumov, “Parallelization of a beam dynamics code and ***”, Physics Review Special Topic-Accelerator and Beams 10, 014201, 2007.
ANL, Argonne, Illinois
Matrix-based beam optics codes such as TRACE-3D are often used for small scale optimizations such as beam matching which involves a limited number of parameters. The limitation of such codes is further amplified for high-intensity and multiple charge state beams as their predictions start to deviate from the more realistic 3D particle tracking codes. For these reasons we have started developing large scale optimization tools for beam tracking codes. The large scale nature comes first from the possibility of optimizing a large number of parameters and second from the minimum number of particles to track especially for space charge dominated beams. The ultimate goal of these developments is not only to optimize the design of an accelerator but also to be able to use a beam dynamics code to operate it once built. A selected set of optimization options will be presented and discussed along with specific applications. We'll also emphasize the need for parallel computing to speed-up the optimization process.