Parallel beam-beam simulation incorporating multiple bunches and multiple interaction regions

The simulation code COMBI has been developed to enable the study of coherent beam-beam effects in the full collision scenario of the LHC, with multiple bunches interacting at multiple crossing points over many turns. The program structure and input are conceived in a general way which allows arbitrary numbers and placements of bunches and interaction points (IP's), together with procedural options for head-on and parasitic collisions (in the strong-strong sense), beam transport, statistics gathering, harmonic analysis, and periodic output of simulation data. The scale of this problem, once we go beyond the simplest case of a pair of bunches interacting once per turn, quickly escalates into the parallel computing arena, and herein we will describe the construction of an MPI-based version of COMBI able to utilize arbitrary numbers of processors to support efficient calculation of multi-bunch multi-IP interactions and transport. Implementing the parallel version did not require extensive disruption of the basic computational routines, since all MPI functionality is isolated in a separate layer of steering routines. After an overview of the basic methods and numerical components of the code, the computational framework will be described in detail and parallel efficiency and scalability of the code will be evaluated. Initial results are shown for an example 4-IP LHC collision scheme.