As the elegant parallelization is an on-going project, we first parallelized
the tracking elements through partitioning the particles to multiple CPUs. We have
parallelized 95 out of 102 elements for tracking simulations. The elements
which have not been parallelized are listed at the end of
this manual. As most time-intensive elements have been
parallelized, we can expect a good speedup and efficiency for this type of simulations.
If a simulation is slow due to a particular element, the user is encouraged to send the input files to us for a performance study. After the tracking elements have been parallelized, I/O becomes the bottleneck of the simulation, especially for simulations requiring a very large number of particles. We developed parallel SDDS library [4] to meet the I/O requirement for large-scale computation. The parallelization of frequency map analysis, dynamic aperture search, and position-dependent momentum aperture determination is discussed in [5]. Recently, we also added several parallel optimization options to Pelegant, including Parallel Genetic Algorithm (PGA), Hybrid Parallel Simplex (HPS) and Parallel Particle Swarm Optimization (PPSO). The usage of these parallel optimization methods can be found in the parallel_optimization_setup section of User's Manual for elegant.